xref: /linux/net/tipc/socket.c (revision 6beeaf48db6c548fcfc2ad32739d33af2fef3a5b)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012-2019, Ericsson AB
5  * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6  * Copyright (c) 2020-2021, Red Hat Inc
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions are met:
11  *
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the names of the copyright holders nor the names of its
18  *    contributors may be used to endorse or promote products derived from
19  *    this software without specific prior written permission.
20  *
21  * Alternatively, this software may be distributed under the terms of the
22  * GNU General Public License ("GPL") version 2 as published by the Free
23  * Software Foundation.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
26  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
29  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35  * POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/rhashtable.h>
39 #include <linux/sched/signal.h>
40 
41 #include "core.h"
42 #include "name_table.h"
43 #include "node.h"
44 #include "link.h"
45 #include "name_distr.h"
46 #include "socket.h"
47 #include "bcast.h"
48 #include "netlink.h"
49 #include "group.h"
50 #include "trace.h"
51 
52 #define NAGLE_START_INIT	4
53 #define NAGLE_START_MAX		1024
54 #define CONN_TIMEOUT_DEFAULT    8000    /* default connect timeout = 8s */
55 #define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
56 #define TIPC_MAX_PORT		0xffffffff
57 #define TIPC_MIN_PORT		1
58 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of rcv window size */
59 
60 enum {
61 	TIPC_LISTEN = TCP_LISTEN,
62 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
63 	TIPC_OPEN = TCP_CLOSE,
64 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
65 	TIPC_CONNECTING = TCP_SYN_SENT,
66 };
67 
68 struct sockaddr_pair {
69 	struct sockaddr_tipc sock;
70 	struct sockaddr_tipc member;
71 };
72 
73 /**
74  * struct tipc_sock - TIPC socket structure
75  * @sk: socket - interacts with 'port' and with user via the socket API
76  * @max_pkt: maximum packet size "hint" used when building messages sent by port
77  * @maxnagle: maximum size of msg which can be subject to nagle
78  * @portid: unique port identity in TIPC socket hash table
79  * @phdr: preformatted message header used when sending messages
80  * @cong_links: list of congested links
81  * @publications: list of publications for port
82  * @blocking_link: address of the congested link we are currently sleeping on
83  * @pub_count: total # of publications port has made during its lifetime
84  * @conn_timeout: the time we can wait for an unresponded setup request
85  * @probe_unacked: probe has not received ack yet
86  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
87  * @cong_link_cnt: number of congested links
88  * @snt_unacked: # messages sent by socket, and not yet acked by peer
89  * @snd_win: send window size
90  * @peer_caps: peer capabilities mask
91  * @rcv_unacked: # messages read by user, but not yet acked back to peer
92  * @rcv_win: receive window size
93  * @peer: 'connected' peer for dgram/rdm
94  * @node: hash table node
95  * @mc_method: cookie for use between socket and broadcast layer
96  * @rcu: rcu struct for tipc_sock
97  * @group: TIPC communications group
98  * @oneway: message count in one direction (FIXME)
99  * @nagle_start: current nagle value
100  * @snd_backlog: send backlog count
101  * @msg_acc: messages accepted; used in managing backlog and nagle
102  * @pkt_cnt: TIPC socket packet count
103  * @expect_ack: whether this TIPC socket is expecting an ack
104  * @nodelay: setsockopt() TIPC_NODELAY setting
105  * @group_is_open: TIPC socket group is fully open (FIXME)
106  * @published: true if port has one or more associated names
107  * @conn_addrtype: address type used when establishing connection
108  */
109 struct tipc_sock {
110 	struct sock sk;
111 	u32 max_pkt;
112 	u32 maxnagle;
113 	u32 portid;
114 	struct tipc_msg phdr;
115 	struct list_head cong_links;
116 	struct list_head publications;
117 	u32 pub_count;
118 	atomic_t dupl_rcvcnt;
119 	u16 conn_timeout;
120 	bool probe_unacked;
121 	u16 cong_link_cnt;
122 	u16 snt_unacked;
123 	u16 snd_win;
124 	u16 peer_caps;
125 	u16 rcv_unacked;
126 	u16 rcv_win;
127 	struct sockaddr_tipc peer;
128 	struct rhash_head node;
129 	struct tipc_mc_method mc_method;
130 	struct rcu_head rcu;
131 	struct tipc_group *group;
132 	u32 oneway;
133 	u32 nagle_start;
134 	u16 snd_backlog;
135 	u16 msg_acc;
136 	u16 pkt_cnt;
137 	bool expect_ack;
138 	bool nodelay;
139 	bool group_is_open;
140 	bool published;
141 	u8 conn_addrtype;
142 };
143 
144 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
145 static void tipc_data_ready(struct sock *sk);
146 static void tipc_write_space(struct sock *sk);
147 static void tipc_sock_destruct(struct sock *sk);
148 static int tipc_release(struct socket *sock);
149 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
150 		       bool kern);
151 static void tipc_sk_timeout(struct timer_list *t);
152 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
153 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
154 static int tipc_sk_leave(struct tipc_sock *tsk);
155 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
156 static int tipc_sk_insert(struct tipc_sock *tsk);
157 static void tipc_sk_remove(struct tipc_sock *tsk);
158 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
159 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
160 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
161 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
162 
163 static const struct proto_ops packet_ops;
164 static const struct proto_ops stream_ops;
165 static const struct proto_ops msg_ops;
166 static struct proto tipc_proto;
167 static const struct rhashtable_params tsk_rht_params;
168 
169 static u32 tsk_own_node(struct tipc_sock *tsk)
170 {
171 	return msg_prevnode(&tsk->phdr);
172 }
173 
174 static u32 tsk_peer_node(struct tipc_sock *tsk)
175 {
176 	return msg_destnode(&tsk->phdr);
177 }
178 
179 static u32 tsk_peer_port(struct tipc_sock *tsk)
180 {
181 	return msg_destport(&tsk->phdr);
182 }
183 
184 static  bool tsk_unreliable(struct tipc_sock *tsk)
185 {
186 	return msg_src_droppable(&tsk->phdr) != 0;
187 }
188 
189 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
190 {
191 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
192 }
193 
194 static bool tsk_unreturnable(struct tipc_sock *tsk)
195 {
196 	return msg_dest_droppable(&tsk->phdr) != 0;
197 }
198 
199 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
200 {
201 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
202 }
203 
204 static int tsk_importance(struct tipc_sock *tsk)
205 {
206 	return msg_importance(&tsk->phdr);
207 }
208 
209 static struct tipc_sock *tipc_sk(const struct sock *sk)
210 {
211 	return container_of(sk, struct tipc_sock, sk);
212 }
213 
214 int tsk_set_importance(struct sock *sk, int imp)
215 {
216 	if (imp > TIPC_CRITICAL_IMPORTANCE)
217 		return -EINVAL;
218 	msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
219 	return 0;
220 }
221 
222 static bool tsk_conn_cong(struct tipc_sock *tsk)
223 {
224 	return tsk->snt_unacked > tsk->snd_win;
225 }
226 
227 static u16 tsk_blocks(int len)
228 {
229 	return ((len / FLOWCTL_BLK_SZ) + 1);
230 }
231 
232 /* tsk_blocks(): translate a buffer size in bytes to number of
233  * advertisable blocks, taking into account the ratio truesize(len)/len
234  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
235  */
236 static u16 tsk_adv_blocks(int len)
237 {
238 	return len / FLOWCTL_BLK_SZ / 4;
239 }
240 
241 /* tsk_inc(): increment counter for sent or received data
242  * - If block based flow control is not supported by peer we
243  *   fall back to message based ditto, incrementing the counter
244  */
245 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
246 {
247 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
248 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
249 	return 1;
250 }
251 
252 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
253  */
254 static void tsk_set_nagle(struct tipc_sock *tsk)
255 {
256 	struct sock *sk = &tsk->sk;
257 
258 	tsk->maxnagle = 0;
259 	if (sk->sk_type != SOCK_STREAM)
260 		return;
261 	if (tsk->nodelay)
262 		return;
263 	if (!(tsk->peer_caps & TIPC_NAGLE))
264 		return;
265 	/* Limit node local buffer size to avoid receive queue overflow */
266 	if (tsk->max_pkt == MAX_MSG_SIZE)
267 		tsk->maxnagle = 1500;
268 	else
269 		tsk->maxnagle = tsk->max_pkt;
270 }
271 
272 /**
273  * tsk_advance_rx_queue - discard first buffer in socket receive queue
274  * @sk: network socket
275  *
276  * Caller must hold socket lock
277  */
278 static void tsk_advance_rx_queue(struct sock *sk)
279 {
280 	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
281 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
282 }
283 
284 /* tipc_sk_respond() : send response message back to sender
285  */
286 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
287 {
288 	u32 selector;
289 	u32 dnode;
290 	u32 onode = tipc_own_addr(sock_net(sk));
291 
292 	if (!tipc_msg_reverse(onode, &skb, err))
293 		return;
294 
295 	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
296 	dnode = msg_destnode(buf_msg(skb));
297 	selector = msg_origport(buf_msg(skb));
298 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
299 }
300 
301 /**
302  * tsk_rej_rx_queue - reject all buffers in socket receive queue
303  * @sk: network socket
304  * @error: response error code
305  *
306  * Caller must hold socket lock
307  */
308 static void tsk_rej_rx_queue(struct sock *sk, int error)
309 {
310 	struct sk_buff *skb;
311 
312 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
313 		tipc_sk_respond(sk, skb, error);
314 }
315 
316 static bool tipc_sk_connected(struct sock *sk)
317 {
318 	return sk->sk_state == TIPC_ESTABLISHED;
319 }
320 
321 /* tipc_sk_type_connectionless - check if the socket is datagram socket
322  * @sk: socket
323  *
324  * Returns true if connection less, false otherwise
325  */
326 static bool tipc_sk_type_connectionless(struct sock *sk)
327 {
328 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
329 }
330 
331 /* tsk_peer_msg - verify if message was sent by connected port's peer
332  *
333  * Handles cases where the node's network address has changed from
334  * the default of <0.0.0> to its configured setting.
335  */
336 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
337 {
338 	struct sock *sk = &tsk->sk;
339 	u32 self = tipc_own_addr(sock_net(sk));
340 	u32 peer_port = tsk_peer_port(tsk);
341 	u32 orig_node, peer_node;
342 
343 	if (unlikely(!tipc_sk_connected(sk)))
344 		return false;
345 
346 	if (unlikely(msg_origport(msg) != peer_port))
347 		return false;
348 
349 	orig_node = msg_orignode(msg);
350 	peer_node = tsk_peer_node(tsk);
351 
352 	if (likely(orig_node == peer_node))
353 		return true;
354 
355 	if (!orig_node && peer_node == self)
356 		return true;
357 
358 	if (!peer_node && orig_node == self)
359 		return true;
360 
361 	return false;
362 }
363 
364 /* tipc_set_sk_state - set the sk_state of the socket
365  * @sk: socket
366  *
367  * Caller must hold socket lock
368  *
369  * Returns 0 on success, errno otherwise
370  */
371 static int tipc_set_sk_state(struct sock *sk, int state)
372 {
373 	int oldsk_state = sk->sk_state;
374 	int res = -EINVAL;
375 
376 	switch (state) {
377 	case TIPC_OPEN:
378 		res = 0;
379 		break;
380 	case TIPC_LISTEN:
381 	case TIPC_CONNECTING:
382 		if (oldsk_state == TIPC_OPEN)
383 			res = 0;
384 		break;
385 	case TIPC_ESTABLISHED:
386 		if (oldsk_state == TIPC_CONNECTING ||
387 		    oldsk_state == TIPC_OPEN)
388 			res = 0;
389 		break;
390 	case TIPC_DISCONNECTING:
391 		if (oldsk_state == TIPC_CONNECTING ||
392 		    oldsk_state == TIPC_ESTABLISHED)
393 			res = 0;
394 		break;
395 	}
396 
397 	if (!res)
398 		sk->sk_state = state;
399 
400 	return res;
401 }
402 
403 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
404 {
405 	struct sock *sk = sock->sk;
406 	int err = sock_error(sk);
407 	int typ = sock->type;
408 
409 	if (err)
410 		return err;
411 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
412 		if (sk->sk_state == TIPC_DISCONNECTING)
413 			return -EPIPE;
414 		else if (!tipc_sk_connected(sk))
415 			return -ENOTCONN;
416 	}
417 	if (!*timeout)
418 		return -EAGAIN;
419 	if (signal_pending(current))
420 		return sock_intr_errno(*timeout);
421 
422 	return 0;
423 }
424 
425 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
426 ({                                                                             \
427 	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
428 	struct sock *sk_;						       \
429 	int rc_;							       \
430 									       \
431 	while ((rc_ = !(condition_))) {					       \
432 		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
433 		smp_rmb();                                                     \
434 		sk_ = (sock_)->sk;					       \
435 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
436 		if (rc_)						       \
437 			break;						       \
438 		add_wait_queue(sk_sleep(sk_), &wait_);                         \
439 		release_sock(sk_);					       \
440 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
441 		sched_annotate_sleep();				               \
442 		lock_sock(sk_);						       \
443 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
444 	}								       \
445 	rc_;								       \
446 })
447 
448 /**
449  * tipc_sk_create - create a TIPC socket
450  * @net: network namespace (must be default network)
451  * @sock: pre-allocated socket structure
452  * @protocol: protocol indicator (must be 0)
453  * @kern: caused by kernel or by userspace?
454  *
455  * This routine creates additional data structures used by the TIPC socket,
456  * initializes them, and links them together.
457  *
458  * Return: 0 on success, errno otherwise
459  */
460 static int tipc_sk_create(struct net *net, struct socket *sock,
461 			  int protocol, int kern)
462 {
463 	const struct proto_ops *ops;
464 	struct sock *sk;
465 	struct tipc_sock *tsk;
466 	struct tipc_msg *msg;
467 
468 	/* Validate arguments */
469 	if (unlikely(protocol != 0))
470 		return -EPROTONOSUPPORT;
471 
472 	switch (sock->type) {
473 	case SOCK_STREAM:
474 		ops = &stream_ops;
475 		break;
476 	case SOCK_SEQPACKET:
477 		ops = &packet_ops;
478 		break;
479 	case SOCK_DGRAM:
480 	case SOCK_RDM:
481 		ops = &msg_ops;
482 		break;
483 	default:
484 		return -EPROTOTYPE;
485 	}
486 
487 	/* Allocate socket's protocol area */
488 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
489 	if (sk == NULL)
490 		return -ENOMEM;
491 
492 	tsk = tipc_sk(sk);
493 	tsk->max_pkt = MAX_PKT_DEFAULT;
494 	tsk->maxnagle = 0;
495 	tsk->nagle_start = NAGLE_START_INIT;
496 	INIT_LIST_HEAD(&tsk->publications);
497 	INIT_LIST_HEAD(&tsk->cong_links);
498 	msg = &tsk->phdr;
499 
500 	/* Finish initializing socket data structures */
501 	sock->ops = ops;
502 	sock_init_data(sock, sk);
503 	tipc_set_sk_state(sk, TIPC_OPEN);
504 	if (tipc_sk_insert(tsk)) {
505 		pr_warn("Socket create failed; port number exhausted\n");
506 		return -EINVAL;
507 	}
508 
509 	/* Ensure tsk is visible before we read own_addr. */
510 	smp_mb();
511 
512 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
513 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
514 
515 	msg_set_origport(msg, tsk->portid);
516 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
517 	sk->sk_shutdown = 0;
518 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
519 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
520 	sk->sk_data_ready = tipc_data_ready;
521 	sk->sk_write_space = tipc_write_space;
522 	sk->sk_destruct = tipc_sock_destruct;
523 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
524 	tsk->group_is_open = true;
525 	atomic_set(&tsk->dupl_rcvcnt, 0);
526 
527 	/* Start out with safe limits until we receive an advertised window */
528 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
529 	tsk->rcv_win = tsk->snd_win;
530 
531 	if (tipc_sk_type_connectionless(sk)) {
532 		tsk_set_unreturnable(tsk, true);
533 		if (sock->type == SOCK_DGRAM)
534 			tsk_set_unreliable(tsk, true);
535 	}
536 	__skb_queue_head_init(&tsk->mc_method.deferredq);
537 	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
538 	return 0;
539 }
540 
541 static void tipc_sk_callback(struct rcu_head *head)
542 {
543 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
544 
545 	sock_put(&tsk->sk);
546 }
547 
548 /* Caller should hold socket lock for the socket. */
549 static void __tipc_shutdown(struct socket *sock, int error)
550 {
551 	struct sock *sk = sock->sk;
552 	struct tipc_sock *tsk = tipc_sk(sk);
553 	struct net *net = sock_net(sk);
554 	long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
555 	u32 dnode = tsk_peer_node(tsk);
556 	struct sk_buff *skb;
557 
558 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
559 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
560 					    !tsk_conn_cong(tsk)));
561 
562 	/* Push out delayed messages if in Nagle mode */
563 	tipc_sk_push_backlog(tsk, false);
564 	/* Remove pending SYN */
565 	__skb_queue_purge(&sk->sk_write_queue);
566 
567 	/* Remove partially received buffer if any */
568 	skb = skb_peek(&sk->sk_receive_queue);
569 	if (skb && TIPC_SKB_CB(skb)->bytes_read) {
570 		__skb_unlink(skb, &sk->sk_receive_queue);
571 		kfree_skb(skb);
572 	}
573 
574 	/* Reject all unreceived messages if connectionless */
575 	if (tipc_sk_type_connectionless(sk)) {
576 		tsk_rej_rx_queue(sk, error);
577 		return;
578 	}
579 
580 	switch (sk->sk_state) {
581 	case TIPC_CONNECTING:
582 	case TIPC_ESTABLISHED:
583 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
584 		tipc_node_remove_conn(net, dnode, tsk->portid);
585 		/* Send a FIN+/- to its peer */
586 		skb = __skb_dequeue(&sk->sk_receive_queue);
587 		if (skb) {
588 			__skb_queue_purge(&sk->sk_receive_queue);
589 			tipc_sk_respond(sk, skb, error);
590 			break;
591 		}
592 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
593 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
594 				      tsk_own_node(tsk), tsk_peer_port(tsk),
595 				      tsk->portid, error);
596 		if (skb)
597 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
598 		break;
599 	case TIPC_LISTEN:
600 		/* Reject all SYN messages */
601 		tsk_rej_rx_queue(sk, error);
602 		break;
603 	default:
604 		__skb_queue_purge(&sk->sk_receive_queue);
605 		break;
606 	}
607 }
608 
609 /**
610  * tipc_release - destroy a TIPC socket
611  * @sock: socket to destroy
612  *
613  * This routine cleans up any messages that are still queued on the socket.
614  * For DGRAM and RDM socket types, all queued messages are rejected.
615  * For SEQPACKET and STREAM socket types, the first message is rejected
616  * and any others are discarded.  (If the first message on a STREAM socket
617  * is partially-read, it is discarded and the next one is rejected instead.)
618  *
619  * NOTE: Rejected messages are not necessarily returned to the sender!  They
620  * are returned or discarded according to the "destination droppable" setting
621  * specified for the message by the sender.
622  *
623  * Return: 0 on success, errno otherwise
624  */
625 static int tipc_release(struct socket *sock)
626 {
627 	struct sock *sk = sock->sk;
628 	struct tipc_sock *tsk;
629 
630 	/*
631 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
632 	 * releases a pre-allocated child socket that was never used)
633 	 */
634 	if (sk == NULL)
635 		return 0;
636 
637 	tsk = tipc_sk(sk);
638 	lock_sock(sk);
639 
640 	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
641 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
642 	sk->sk_shutdown = SHUTDOWN_MASK;
643 	tipc_sk_leave(tsk);
644 	tipc_sk_withdraw(tsk, NULL);
645 	__skb_queue_purge(&tsk->mc_method.deferredq);
646 	sk_stop_timer(sk, &sk->sk_timer);
647 	tipc_sk_remove(tsk);
648 
649 	sock_orphan(sk);
650 	/* Reject any messages that accumulated in backlog queue */
651 	release_sock(sk);
652 	tipc_dest_list_purge(&tsk->cong_links);
653 	tsk->cong_link_cnt = 0;
654 	call_rcu(&tsk->rcu, tipc_sk_callback);
655 	sock->sk = NULL;
656 
657 	return 0;
658 }
659 
660 /**
661  * __tipc_bind - associate or disassocate TIPC name(s) with a socket
662  * @sock: socket structure
663  * @skaddr: socket address describing name(s) and desired operation
664  * @alen: size of socket address data structure
665  *
666  * Name and name sequence binding are indicated using a positive scope value;
667  * a negative scope value unbinds the specified name.  Specifying no name
668  * (i.e. a socket address length of 0) unbinds all names from the socket.
669  *
670  * Return: 0 on success, errno otherwise
671  *
672  * NOTE: This routine doesn't need to take the socket lock since it doesn't
673  *       access any non-constant socket information.
674  */
675 static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
676 {
677 	struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
678 	struct tipc_sock *tsk = tipc_sk(sock->sk);
679 	bool unbind = false;
680 
681 	if (unlikely(!alen))
682 		return tipc_sk_withdraw(tsk, NULL);
683 
684 	if (ua->addrtype == TIPC_SERVICE_ADDR) {
685 		ua->addrtype = TIPC_SERVICE_RANGE;
686 		ua->sr.upper = ua->sr.lower;
687 	}
688 	if (ua->scope < 0) {
689 		unbind = true;
690 		ua->scope = -ua->scope;
691 	}
692 	/* Users may still use deprecated TIPC_ZONE_SCOPE */
693 	if (ua->scope != TIPC_NODE_SCOPE)
694 		ua->scope = TIPC_CLUSTER_SCOPE;
695 
696 	if (tsk->group)
697 		return -EACCES;
698 
699 	if (unbind)
700 		return tipc_sk_withdraw(tsk, ua);
701 	return tipc_sk_publish(tsk, ua);
702 }
703 
704 int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
705 {
706 	int res;
707 
708 	lock_sock(sock->sk);
709 	res = __tipc_bind(sock, skaddr, alen);
710 	release_sock(sock->sk);
711 	return res;
712 }
713 
714 static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
715 {
716 	struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
717 	u32 atype = ua->addrtype;
718 
719 	if (alen) {
720 		if (!tipc_uaddr_valid(ua, alen))
721 			return -EINVAL;
722 		if (atype == TIPC_SOCKET_ADDR)
723 			return -EAFNOSUPPORT;
724 		if (ua->sr.type < TIPC_RESERVED_TYPES) {
725 			pr_warn_once("Can't bind to reserved service type %u\n",
726 				     ua->sr.type);
727 			return -EACCES;
728 		}
729 	}
730 	return tipc_sk_bind(sock, skaddr, alen);
731 }
732 
733 /**
734  * tipc_getname - get port ID of socket or peer socket
735  * @sock: socket structure
736  * @uaddr: area for returned socket address
737  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
738  *
739  * Return: 0 on success, errno otherwise
740  *
741  * NOTE: This routine doesn't need to take the socket lock since it only
742  *       accesses socket information that is unchanging (or which changes in
743  *       a completely predictable manner).
744  */
745 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
746 			int peer)
747 {
748 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
749 	struct sock *sk = sock->sk;
750 	struct tipc_sock *tsk = tipc_sk(sk);
751 
752 	memset(addr, 0, sizeof(*addr));
753 	if (peer) {
754 		if ((!tipc_sk_connected(sk)) &&
755 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
756 			return -ENOTCONN;
757 		addr->addr.id.ref = tsk_peer_port(tsk);
758 		addr->addr.id.node = tsk_peer_node(tsk);
759 	} else {
760 		addr->addr.id.ref = tsk->portid;
761 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
762 	}
763 
764 	addr->addrtype = TIPC_SOCKET_ADDR;
765 	addr->family = AF_TIPC;
766 	addr->scope = 0;
767 	addr->addr.name.domain = 0;
768 
769 	return sizeof(*addr);
770 }
771 
772 /**
773  * tipc_poll - read and possibly block on pollmask
774  * @file: file structure associated with the socket
775  * @sock: socket for which to calculate the poll bits
776  * @wait: ???
777  *
778  * Return: pollmask value
779  *
780  * COMMENTARY:
781  * It appears that the usual socket locking mechanisms are not useful here
782  * since the pollmask info is potentially out-of-date the moment this routine
783  * exits.  TCP and other protocols seem to rely on higher level poll routines
784  * to handle any preventable race conditions, so TIPC will do the same ...
785  *
786  * IMPORTANT: The fact that a read or write operation is indicated does NOT
787  * imply that the operation will succeed, merely that it should be performed
788  * and will not block.
789  */
790 static __poll_t tipc_poll(struct file *file, struct socket *sock,
791 			      poll_table *wait)
792 {
793 	struct sock *sk = sock->sk;
794 	struct tipc_sock *tsk = tipc_sk(sk);
795 	__poll_t revents = 0;
796 
797 	sock_poll_wait(file, sock, wait);
798 	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
799 
800 	if (sk->sk_shutdown & RCV_SHUTDOWN)
801 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
802 	if (sk->sk_shutdown == SHUTDOWN_MASK)
803 		revents |= EPOLLHUP;
804 
805 	switch (sk->sk_state) {
806 	case TIPC_ESTABLISHED:
807 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
808 			revents |= EPOLLOUT;
809 		fallthrough;
810 	case TIPC_LISTEN:
811 	case TIPC_CONNECTING:
812 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
813 			revents |= EPOLLIN | EPOLLRDNORM;
814 		break;
815 	case TIPC_OPEN:
816 		if (tsk->group_is_open && !tsk->cong_link_cnt)
817 			revents |= EPOLLOUT;
818 		if (!tipc_sk_type_connectionless(sk))
819 			break;
820 		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
821 			break;
822 		revents |= EPOLLIN | EPOLLRDNORM;
823 		break;
824 	case TIPC_DISCONNECTING:
825 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
826 		break;
827 	}
828 	return revents;
829 }
830 
831 /**
832  * tipc_sendmcast - send multicast message
833  * @sock: socket structure
834  * @ua: destination address struct
835  * @msg: message to send
836  * @dlen: length of data to send
837  * @timeout: timeout to wait for wakeup
838  *
839  * Called from function tipc_sendmsg(), which has done all sanity checks
840  * Return: the number of bytes sent on success, or errno
841  */
842 static int tipc_sendmcast(struct  socket *sock, struct tipc_uaddr *ua,
843 			  struct msghdr *msg, size_t dlen, long timeout)
844 {
845 	struct sock *sk = sock->sk;
846 	struct tipc_sock *tsk = tipc_sk(sk);
847 	struct tipc_msg *hdr = &tsk->phdr;
848 	struct net *net = sock_net(sk);
849 	int mtu = tipc_bcast_get_mtu(net);
850 	struct sk_buff_head pkts;
851 	struct tipc_nlist dsts;
852 	int rc;
853 
854 	if (tsk->group)
855 		return -EACCES;
856 
857 	/* Block or return if any destination link is congested */
858 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
859 	if (unlikely(rc))
860 		return rc;
861 
862 	/* Lookup destination nodes */
863 	tipc_nlist_init(&dsts, tipc_own_addr(net));
864 	tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
865 	if (!dsts.local && !dsts.remote)
866 		return -EHOSTUNREACH;
867 
868 	/* Build message header */
869 	msg_set_type(hdr, TIPC_MCAST_MSG);
870 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
871 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
872 	msg_set_destport(hdr, 0);
873 	msg_set_destnode(hdr, 0);
874 	msg_set_nametype(hdr, ua->sr.type);
875 	msg_set_namelower(hdr, ua->sr.lower);
876 	msg_set_nameupper(hdr, ua->sr.upper);
877 
878 	/* Build message as chain of buffers */
879 	__skb_queue_head_init(&pkts);
880 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
881 
882 	/* Send message if build was successful */
883 	if (unlikely(rc == dlen)) {
884 		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
885 					TIPC_DUMP_SK_SNDQ, " ");
886 		rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
887 				     &tsk->cong_link_cnt);
888 	}
889 
890 	tipc_nlist_purge(&dsts);
891 
892 	return rc ? rc : dlen;
893 }
894 
895 /**
896  * tipc_send_group_msg - send a message to a member in the group
897  * @net: network namespace
898  * @tsk: tipc socket
899  * @m: message to send
900  * @mb: group member
901  * @dnode: destination node
902  * @dport: destination port
903  * @dlen: total length of message data
904  */
905 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
906 			       struct msghdr *m, struct tipc_member *mb,
907 			       u32 dnode, u32 dport, int dlen)
908 {
909 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
910 	struct tipc_mc_method *method = &tsk->mc_method;
911 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
912 	struct tipc_msg *hdr = &tsk->phdr;
913 	struct sk_buff_head pkts;
914 	int mtu, rc;
915 
916 	/* Complete message header */
917 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
918 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
919 	msg_set_destport(hdr, dport);
920 	msg_set_destnode(hdr, dnode);
921 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
922 
923 	/* Build message as chain of buffers */
924 	__skb_queue_head_init(&pkts);
925 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
926 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
927 	if (unlikely(rc != dlen))
928 		return rc;
929 
930 	/* Send message */
931 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
932 	if (unlikely(rc == -ELINKCONG)) {
933 		tipc_dest_push(&tsk->cong_links, dnode, 0);
934 		tsk->cong_link_cnt++;
935 	}
936 
937 	/* Update send window */
938 	tipc_group_update_member(mb, blks);
939 
940 	/* A broadcast sent within next EXPIRE period must follow same path */
941 	method->rcast = true;
942 	method->mandatory = true;
943 	return dlen;
944 }
945 
946 /**
947  * tipc_send_group_unicast - send message to a member in the group
948  * @sock: socket structure
949  * @m: message to send
950  * @dlen: total length of message data
951  * @timeout: timeout to wait for wakeup
952  *
953  * Called from function tipc_sendmsg(), which has done all sanity checks
954  * Return: the number of bytes sent on success, or errno
955  */
956 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
957 				   int dlen, long timeout)
958 {
959 	struct sock *sk = sock->sk;
960 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
961 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
962 	struct tipc_sock *tsk = tipc_sk(sk);
963 	struct net *net = sock_net(sk);
964 	struct tipc_member *mb = NULL;
965 	u32 node, port;
966 	int rc;
967 
968 	node = ua->sk.node;
969 	port = ua->sk.ref;
970 	if (!port && !node)
971 		return -EHOSTUNREACH;
972 
973 	/* Block or return if destination link or member is congested */
974 	rc = tipc_wait_for_cond(sock, &timeout,
975 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
976 				tsk->group &&
977 				!tipc_group_cong(tsk->group, node, port, blks,
978 						 &mb));
979 	if (unlikely(rc))
980 		return rc;
981 
982 	if (unlikely(!mb))
983 		return -EHOSTUNREACH;
984 
985 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
986 
987 	return rc ? rc : dlen;
988 }
989 
990 /**
991  * tipc_send_group_anycast - send message to any member with given identity
992  * @sock: socket structure
993  * @m: message to send
994  * @dlen: total length of message data
995  * @timeout: timeout to wait for wakeup
996  *
997  * Called from function tipc_sendmsg(), which has done all sanity checks
998  * Return: the number of bytes sent on success, or errno
999  */
1000 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1001 				   int dlen, long timeout)
1002 {
1003 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1004 	struct sock *sk = sock->sk;
1005 	struct tipc_sock *tsk = tipc_sk(sk);
1006 	struct list_head *cong_links = &tsk->cong_links;
1007 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1008 	struct tipc_msg *hdr = &tsk->phdr;
1009 	struct tipc_member *first = NULL;
1010 	struct tipc_member *mbr = NULL;
1011 	struct net *net = sock_net(sk);
1012 	u32 node, port, exclude;
1013 	struct list_head dsts;
1014 	int lookups = 0;
1015 	int dstcnt, rc;
1016 	bool cong;
1017 
1018 	INIT_LIST_HEAD(&dsts);
1019 	ua->sa.type = msg_nametype(hdr);
1020 	ua->scope = msg_lookup_scope(hdr);
1021 
1022 	while (++lookups < 4) {
1023 		exclude = tipc_group_exclude(tsk->group);
1024 
1025 		first = NULL;
1026 
1027 		/* Look for a non-congested destination member, if any */
1028 		while (1) {
1029 			if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1030 						       exclude, false))
1031 				return -EHOSTUNREACH;
1032 			tipc_dest_pop(&dsts, &node, &port);
1033 			cong = tipc_group_cong(tsk->group, node, port, blks,
1034 					       &mbr);
1035 			if (!cong)
1036 				break;
1037 			if (mbr == first)
1038 				break;
1039 			if (!first)
1040 				first = mbr;
1041 		}
1042 
1043 		/* Start over if destination was not in member list */
1044 		if (unlikely(!mbr))
1045 			continue;
1046 
1047 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1048 			break;
1049 
1050 		/* Block or return if destination link or member is congested */
1051 		rc = tipc_wait_for_cond(sock, &timeout,
1052 					!tipc_dest_find(cong_links, node, 0) &&
1053 					tsk->group &&
1054 					!tipc_group_cong(tsk->group, node, port,
1055 							 blks, &mbr));
1056 		if (unlikely(rc))
1057 			return rc;
1058 
1059 		/* Send, unless destination disappeared while waiting */
1060 		if (likely(mbr))
1061 			break;
1062 	}
1063 
1064 	if (unlikely(lookups >= 4))
1065 		return -EHOSTUNREACH;
1066 
1067 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1068 
1069 	return rc ? rc : dlen;
1070 }
1071 
1072 /**
1073  * tipc_send_group_bcast - send message to all members in communication group
1074  * @sock: socket structure
1075  * @m: message to send
1076  * @dlen: total length of message data
1077  * @timeout: timeout to wait for wakeup
1078  *
1079  * Called from function tipc_sendmsg(), which has done all sanity checks
1080  * Return: the number of bytes sent on success, or errno
1081  */
1082 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1083 				 int dlen, long timeout)
1084 {
1085 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1086 	struct sock *sk = sock->sk;
1087 	struct net *net = sock_net(sk);
1088 	struct tipc_sock *tsk = tipc_sk(sk);
1089 	struct tipc_nlist *dsts;
1090 	struct tipc_mc_method *method = &tsk->mc_method;
1091 	bool ack = method->mandatory && method->rcast;
1092 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1093 	struct tipc_msg *hdr = &tsk->phdr;
1094 	int mtu = tipc_bcast_get_mtu(net);
1095 	struct sk_buff_head pkts;
1096 	int rc = -EHOSTUNREACH;
1097 
1098 	/* Block or return if any destination link or member is congested */
1099 	rc = tipc_wait_for_cond(sock, &timeout,
1100 				!tsk->cong_link_cnt && tsk->group &&
1101 				!tipc_group_bc_cong(tsk->group, blks));
1102 	if (unlikely(rc))
1103 		return rc;
1104 
1105 	dsts = tipc_group_dests(tsk->group);
1106 	if (!dsts->local && !dsts->remote)
1107 		return -EHOSTUNREACH;
1108 
1109 	/* Complete message header */
1110 	if (ua) {
1111 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1112 		msg_set_nameinst(hdr, ua->sa.instance);
1113 	} else {
1114 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1115 		msg_set_nameinst(hdr, 0);
1116 	}
1117 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1118 	msg_set_destport(hdr, 0);
1119 	msg_set_destnode(hdr, 0);
1120 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1121 
1122 	/* Avoid getting stuck with repeated forced replicasts */
1123 	msg_set_grp_bc_ack_req(hdr, ack);
1124 
1125 	/* Build message as chain of buffers */
1126 	__skb_queue_head_init(&pkts);
1127 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1128 	if (unlikely(rc != dlen))
1129 		return rc;
1130 
1131 	/* Send message */
1132 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1133 	if (unlikely(rc))
1134 		return rc;
1135 
1136 	/* Update broadcast sequence number and send windows */
1137 	tipc_group_update_bc_members(tsk->group, blks, ack);
1138 
1139 	/* Broadcast link is now free to choose method for next broadcast */
1140 	method->mandatory = false;
1141 	method->expires = jiffies;
1142 
1143 	return dlen;
1144 }
1145 
1146 /**
1147  * tipc_send_group_mcast - send message to all members with given identity
1148  * @sock: socket structure
1149  * @m: message to send
1150  * @dlen: total length of message data
1151  * @timeout: timeout to wait for wakeup
1152  *
1153  * Called from function tipc_sendmsg(), which has done all sanity checks
1154  * Return: the number of bytes sent on success, or errno
1155  */
1156 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1157 				 int dlen, long timeout)
1158 {
1159 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1160 	struct sock *sk = sock->sk;
1161 	struct tipc_sock *tsk = tipc_sk(sk);
1162 	struct tipc_group *grp = tsk->group;
1163 	struct tipc_msg *hdr = &tsk->phdr;
1164 	struct net *net = sock_net(sk);
1165 	struct list_head dsts;
1166 	u32 dstcnt, exclude;
1167 
1168 	INIT_LIST_HEAD(&dsts);
1169 	ua->sa.type = msg_nametype(hdr);
1170 	ua->scope = msg_lookup_scope(hdr);
1171 	exclude = tipc_group_exclude(grp);
1172 
1173 	if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1174 		return -EHOSTUNREACH;
1175 
1176 	if (dstcnt == 1) {
1177 		tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1178 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1179 	}
1180 
1181 	tipc_dest_list_purge(&dsts);
1182 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1183 }
1184 
1185 /**
1186  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1187  * @net: the associated network namespace
1188  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1189  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1190  *
1191  * Multi-threaded: parallel calls with reference to same queues may occur
1192  */
1193 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1194 		       struct sk_buff_head *inputq)
1195 {
1196 	u32 self = tipc_own_addr(net);
1197 	struct sk_buff *skb, *_skb;
1198 	u32 portid, onode;
1199 	struct sk_buff_head tmpq;
1200 	struct list_head dports;
1201 	struct tipc_msg *hdr;
1202 	struct tipc_uaddr ua;
1203 	int user, mtyp, hlen;
1204 
1205 	__skb_queue_head_init(&tmpq);
1206 	INIT_LIST_HEAD(&dports);
1207 	ua.addrtype = TIPC_SERVICE_RANGE;
1208 
1209 	/* tipc_skb_peek() increments the head skb's reference counter */
1210 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1211 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1212 		hdr = buf_msg(skb);
1213 		user = msg_user(hdr);
1214 		mtyp = msg_type(hdr);
1215 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1216 		onode = msg_orignode(hdr);
1217 		ua.sr.type = msg_nametype(hdr);
1218 		ua.sr.lower = msg_namelower(hdr);
1219 		ua.sr.upper = msg_nameupper(hdr);
1220 		if (onode == self)
1221 			ua.scope = TIPC_ANY_SCOPE;
1222 		else
1223 			ua.scope = TIPC_CLUSTER_SCOPE;
1224 
1225 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1226 			spin_lock_bh(&inputq->lock);
1227 			if (skb_peek(arrvq) == skb) {
1228 				__skb_dequeue(arrvq);
1229 				__skb_queue_tail(inputq, skb);
1230 			}
1231 			kfree_skb(skb);
1232 			spin_unlock_bh(&inputq->lock);
1233 			continue;
1234 		}
1235 
1236 		/* Group messages require exact scope match */
1237 		if (msg_in_group(hdr)) {
1238 			ua.sr.lower = 0;
1239 			ua.sr.upper = ~0;
1240 			ua.scope = msg_lookup_scope(hdr);
1241 		}
1242 
1243 		/* Create destination port list: */
1244 		tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports);
1245 
1246 		/* Clone message per destination */
1247 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1248 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1249 			if (_skb) {
1250 				msg_set_destport(buf_msg(_skb), portid);
1251 				__skb_queue_tail(&tmpq, _skb);
1252 				continue;
1253 			}
1254 			pr_warn("Failed to clone mcast rcv buffer\n");
1255 		}
1256 		/* Append clones to inputq only if skb is still head of arrvq */
1257 		spin_lock_bh(&inputq->lock);
1258 		if (skb_peek(arrvq) == skb) {
1259 			skb_queue_splice_tail_init(&tmpq, inputq);
1260 			/* Decrement the skb's refcnt */
1261 			kfree_skb(__skb_dequeue(arrvq));
1262 		}
1263 		spin_unlock_bh(&inputq->lock);
1264 		__skb_queue_purge(&tmpq);
1265 		kfree_skb(skb);
1266 	}
1267 	tipc_sk_rcv(net, inputq);
1268 }
1269 
1270 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1271  *                         when socket is in Nagle mode
1272  */
1273 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1274 {
1275 	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1276 	struct sk_buff *skb = skb_peek_tail(txq);
1277 	struct net *net = sock_net(&tsk->sk);
1278 	u32 dnode = tsk_peer_node(tsk);
1279 	int rc;
1280 
1281 	if (nagle_ack) {
1282 		tsk->pkt_cnt += skb_queue_len(txq);
1283 		if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1284 			tsk->oneway = 0;
1285 			if (tsk->nagle_start < NAGLE_START_MAX)
1286 				tsk->nagle_start *= 2;
1287 			tsk->expect_ack = false;
1288 			pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1289 				 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1290 				 tsk->nagle_start);
1291 		} else {
1292 			tsk->nagle_start = NAGLE_START_INIT;
1293 			if (skb) {
1294 				msg_set_ack_required(buf_msg(skb));
1295 				tsk->expect_ack = true;
1296 			} else {
1297 				tsk->expect_ack = false;
1298 			}
1299 		}
1300 		tsk->msg_acc = 0;
1301 		tsk->pkt_cnt = 0;
1302 	}
1303 
1304 	if (!skb || tsk->cong_link_cnt)
1305 		return;
1306 
1307 	/* Do not send SYN again after congestion */
1308 	if (msg_is_syn(buf_msg(skb)))
1309 		return;
1310 
1311 	if (tsk->msg_acc)
1312 		tsk->pkt_cnt += skb_queue_len(txq);
1313 	tsk->snt_unacked += tsk->snd_backlog;
1314 	tsk->snd_backlog = 0;
1315 	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1316 	if (rc == -ELINKCONG)
1317 		tsk->cong_link_cnt = 1;
1318 }
1319 
1320 /**
1321  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1322  * @tsk: receiving socket
1323  * @skb: pointer to message buffer.
1324  * @inputq: buffer list containing the buffers
1325  * @xmitq: output message area
1326  */
1327 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1328 				   struct sk_buff_head *inputq,
1329 				   struct sk_buff_head *xmitq)
1330 {
1331 	struct tipc_msg *hdr = buf_msg(skb);
1332 	u32 onode = tsk_own_node(tsk);
1333 	struct sock *sk = &tsk->sk;
1334 	int mtyp = msg_type(hdr);
1335 	bool was_cong;
1336 
1337 	/* Ignore if connection cannot be validated: */
1338 	if (!tsk_peer_msg(tsk, hdr)) {
1339 		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1340 		goto exit;
1341 	}
1342 
1343 	if (unlikely(msg_errcode(hdr))) {
1344 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1345 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1346 				      tsk_peer_port(tsk));
1347 		sk->sk_state_change(sk);
1348 
1349 		/* State change is ignored if socket already awake,
1350 		 * - convert msg to abort msg and add to inqueue
1351 		 */
1352 		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1353 		msg_set_type(hdr, TIPC_CONN_MSG);
1354 		msg_set_size(hdr, BASIC_H_SIZE);
1355 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1356 		__skb_queue_tail(inputq, skb);
1357 		return;
1358 	}
1359 
1360 	tsk->probe_unacked = false;
1361 
1362 	if (mtyp == CONN_PROBE) {
1363 		msg_set_type(hdr, CONN_PROBE_REPLY);
1364 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1365 			__skb_queue_tail(xmitq, skb);
1366 		return;
1367 	} else if (mtyp == CONN_ACK) {
1368 		was_cong = tsk_conn_cong(tsk);
1369 		tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1370 		tsk->snt_unacked -= msg_conn_ack(hdr);
1371 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1372 			tsk->snd_win = msg_adv_win(hdr);
1373 		if (was_cong && !tsk_conn_cong(tsk))
1374 			sk->sk_write_space(sk);
1375 	} else if (mtyp != CONN_PROBE_REPLY) {
1376 		pr_warn("Received unknown CONN_PROTO msg\n");
1377 	}
1378 exit:
1379 	kfree_skb(skb);
1380 }
1381 
1382 /**
1383  * tipc_sendmsg - send message in connectionless manner
1384  * @sock: socket structure
1385  * @m: message to send
1386  * @dsz: amount of user data to be sent
1387  *
1388  * Message must have an destination specified explicitly.
1389  * Used for SOCK_RDM and SOCK_DGRAM messages,
1390  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1391  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1392  *
1393  * Return: the number of bytes sent on success, or errno otherwise
1394  */
1395 static int tipc_sendmsg(struct socket *sock,
1396 			struct msghdr *m, size_t dsz)
1397 {
1398 	struct sock *sk = sock->sk;
1399 	int ret;
1400 
1401 	lock_sock(sk);
1402 	ret = __tipc_sendmsg(sock, m, dsz);
1403 	release_sock(sk);
1404 
1405 	return ret;
1406 }
1407 
1408 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1409 {
1410 	struct sock *sk = sock->sk;
1411 	struct net *net = sock_net(sk);
1412 	struct tipc_sock *tsk = tipc_sk(sk);
1413 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1414 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1415 	struct list_head *clinks = &tsk->cong_links;
1416 	bool syn = !tipc_sk_type_connectionless(sk);
1417 	struct tipc_group *grp = tsk->group;
1418 	struct tipc_msg *hdr = &tsk->phdr;
1419 	struct tipc_socket_addr skaddr;
1420 	struct sk_buff_head pkts;
1421 	int atype, mtu, rc;
1422 
1423 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1424 		return -EMSGSIZE;
1425 
1426 	if (ua) {
1427 		if (!tipc_uaddr_valid(ua, m->msg_namelen))
1428 			return -EINVAL;
1429 		atype = ua->addrtype;
1430 	}
1431 
1432 	/* If socket belongs to a communication group follow other paths */
1433 	if (grp) {
1434 		if (!ua)
1435 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1436 		if (atype == TIPC_SERVICE_ADDR)
1437 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1438 		if (atype == TIPC_SOCKET_ADDR)
1439 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1440 		if (atype == TIPC_SERVICE_RANGE)
1441 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1442 		return -EINVAL;
1443 	}
1444 
1445 	if (!ua) {
1446 		ua = (struct tipc_uaddr *)&tsk->peer;
1447 		if (!syn && ua->family != AF_TIPC)
1448 			return -EDESTADDRREQ;
1449 		atype = ua->addrtype;
1450 	}
1451 
1452 	if (unlikely(syn)) {
1453 		if (sk->sk_state == TIPC_LISTEN)
1454 			return -EPIPE;
1455 		if (sk->sk_state != TIPC_OPEN)
1456 			return -EISCONN;
1457 		if (tsk->published)
1458 			return -EOPNOTSUPP;
1459 		if (atype == TIPC_SERVICE_ADDR)
1460 			tsk->conn_addrtype = atype;
1461 		msg_set_syn(hdr, 1);
1462 	}
1463 
1464 	/* Determine destination */
1465 	if (atype == TIPC_SERVICE_RANGE) {
1466 		return tipc_sendmcast(sock, ua, m, dlen, timeout);
1467 	} else if (atype == TIPC_SERVICE_ADDR) {
1468 		skaddr.node = ua->lookup_node;
1469 		ua->scope = tipc_node2scope(skaddr.node);
1470 		if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1471 			return -EHOSTUNREACH;
1472 	} else if (atype == TIPC_SOCKET_ADDR) {
1473 		skaddr = ua->sk;
1474 	} else {
1475 		return -EINVAL;
1476 	}
1477 
1478 	/* Block or return if destination link is congested */
1479 	rc = tipc_wait_for_cond(sock, &timeout,
1480 				!tipc_dest_find(clinks, skaddr.node, 0));
1481 	if (unlikely(rc))
1482 		return rc;
1483 
1484 	/* Finally build message header */
1485 	msg_set_destnode(hdr, skaddr.node);
1486 	msg_set_destport(hdr, skaddr.ref);
1487 	if (atype == TIPC_SERVICE_ADDR) {
1488 		msg_set_type(hdr, TIPC_NAMED_MSG);
1489 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1490 		msg_set_nametype(hdr, ua->sa.type);
1491 		msg_set_nameinst(hdr, ua->sa.instance);
1492 		msg_set_lookup_scope(hdr, ua->scope);
1493 	} else { /* TIPC_SOCKET_ADDR */
1494 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1495 		msg_set_lookup_scope(hdr, 0);
1496 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1497 	}
1498 
1499 	/* Add message body */
1500 	__skb_queue_head_init(&pkts);
1501 	mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1502 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1503 	if (unlikely(rc != dlen))
1504 		return rc;
1505 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1506 		__skb_queue_purge(&pkts);
1507 		return -ENOMEM;
1508 	}
1509 
1510 	/* Send message */
1511 	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1512 	rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1513 	if (unlikely(rc == -ELINKCONG)) {
1514 		tipc_dest_push(clinks, skaddr.node, 0);
1515 		tsk->cong_link_cnt++;
1516 		rc = 0;
1517 	}
1518 
1519 	if (unlikely(syn && !rc)) {
1520 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1521 		if (dlen && timeout) {
1522 			timeout = msecs_to_jiffies(timeout);
1523 			tipc_wait_for_connect(sock, &timeout);
1524 		}
1525 	}
1526 
1527 	return rc ? rc : dlen;
1528 }
1529 
1530 /**
1531  * tipc_sendstream - send stream-oriented data
1532  * @sock: socket structure
1533  * @m: data to send
1534  * @dsz: total length of data to be transmitted
1535  *
1536  * Used for SOCK_STREAM data.
1537  *
1538  * Return: the number of bytes sent on success (or partial success),
1539  * or errno if no data sent
1540  */
1541 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1542 {
1543 	struct sock *sk = sock->sk;
1544 	int ret;
1545 
1546 	lock_sock(sk);
1547 	ret = __tipc_sendstream(sock, m, dsz);
1548 	release_sock(sk);
1549 
1550 	return ret;
1551 }
1552 
1553 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1554 {
1555 	struct sock *sk = sock->sk;
1556 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1557 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1558 	struct sk_buff_head *txq = &sk->sk_write_queue;
1559 	struct tipc_sock *tsk = tipc_sk(sk);
1560 	struct tipc_msg *hdr = &tsk->phdr;
1561 	struct net *net = sock_net(sk);
1562 	struct sk_buff *skb;
1563 	u32 dnode = tsk_peer_node(tsk);
1564 	int maxnagle = tsk->maxnagle;
1565 	int maxpkt = tsk->max_pkt;
1566 	int send, sent = 0;
1567 	int blocks, rc = 0;
1568 
1569 	if (unlikely(dlen > INT_MAX))
1570 		return -EMSGSIZE;
1571 
1572 	/* Handle implicit connection setup */
1573 	if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1574 		rc = __tipc_sendmsg(sock, m, dlen);
1575 		if (dlen && dlen == rc) {
1576 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1577 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1578 		}
1579 		return rc;
1580 	}
1581 
1582 	do {
1583 		rc = tipc_wait_for_cond(sock, &timeout,
1584 					(!tsk->cong_link_cnt &&
1585 					 !tsk_conn_cong(tsk) &&
1586 					 tipc_sk_connected(sk)));
1587 		if (unlikely(rc))
1588 			break;
1589 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1590 		blocks = tsk->snd_backlog;
1591 		if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1592 		    send <= maxnagle) {
1593 			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1594 			if (unlikely(rc < 0))
1595 				break;
1596 			blocks += rc;
1597 			tsk->msg_acc++;
1598 			if (blocks <= 64 && tsk->expect_ack) {
1599 				tsk->snd_backlog = blocks;
1600 				sent += send;
1601 				break;
1602 			} else if (blocks > 64) {
1603 				tsk->pkt_cnt += skb_queue_len(txq);
1604 			} else {
1605 				skb = skb_peek_tail(txq);
1606 				if (skb) {
1607 					msg_set_ack_required(buf_msg(skb));
1608 					tsk->expect_ack = true;
1609 				} else {
1610 					tsk->expect_ack = false;
1611 				}
1612 				tsk->msg_acc = 0;
1613 				tsk->pkt_cnt = 0;
1614 			}
1615 		} else {
1616 			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1617 			if (unlikely(rc != send))
1618 				break;
1619 			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1620 		}
1621 		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1622 					 TIPC_DUMP_SK_SNDQ, " ");
1623 		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1624 		if (unlikely(rc == -ELINKCONG)) {
1625 			tsk->cong_link_cnt = 1;
1626 			rc = 0;
1627 		}
1628 		if (likely(!rc)) {
1629 			tsk->snt_unacked += blocks;
1630 			tsk->snd_backlog = 0;
1631 			sent += send;
1632 		}
1633 	} while (sent < dlen && !rc);
1634 
1635 	return sent ? sent : rc;
1636 }
1637 
1638 /**
1639  * tipc_send_packet - send a connection-oriented message
1640  * @sock: socket structure
1641  * @m: message to send
1642  * @dsz: length of data to be transmitted
1643  *
1644  * Used for SOCK_SEQPACKET messages.
1645  *
1646  * Return: the number of bytes sent on success, or errno otherwise
1647  */
1648 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1649 {
1650 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1651 		return -EMSGSIZE;
1652 
1653 	return tipc_sendstream(sock, m, dsz);
1654 }
1655 
1656 /* tipc_sk_finish_conn - complete the setup of a connection
1657  */
1658 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1659 				u32 peer_node)
1660 {
1661 	struct sock *sk = &tsk->sk;
1662 	struct net *net = sock_net(sk);
1663 	struct tipc_msg *msg = &tsk->phdr;
1664 
1665 	msg_set_syn(msg, 0);
1666 	msg_set_destnode(msg, peer_node);
1667 	msg_set_destport(msg, peer_port);
1668 	msg_set_type(msg, TIPC_CONN_MSG);
1669 	msg_set_lookup_scope(msg, 0);
1670 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1671 
1672 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1673 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1674 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1675 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1676 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1677 	tsk_set_nagle(tsk);
1678 	__skb_queue_purge(&sk->sk_write_queue);
1679 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1680 		return;
1681 
1682 	/* Fall back to message based flow control */
1683 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1684 	tsk->snd_win = FLOWCTL_MSG_WIN;
1685 }
1686 
1687 /**
1688  * tipc_sk_set_orig_addr - capture sender's address for received message
1689  * @m: descriptor for message info
1690  * @skb: received message
1691  *
1692  * Note: Address is not captured if not requested by receiver.
1693  */
1694 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1695 {
1696 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1697 	struct tipc_msg *hdr = buf_msg(skb);
1698 
1699 	if (!srcaddr)
1700 		return;
1701 
1702 	srcaddr->sock.family = AF_TIPC;
1703 	srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1704 	srcaddr->sock.scope = 0;
1705 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1706 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1707 	srcaddr->sock.addr.name.domain = 0;
1708 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1709 
1710 	if (!msg_in_group(hdr))
1711 		return;
1712 
1713 	/* Group message users may also want to know sending member's id */
1714 	srcaddr->member.family = AF_TIPC;
1715 	srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1716 	srcaddr->member.scope = 0;
1717 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1718 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1719 	srcaddr->member.addr.name.domain = 0;
1720 	m->msg_namelen = sizeof(*srcaddr);
1721 }
1722 
1723 /**
1724  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1725  * @m: descriptor for message info
1726  * @skb: received message buffer
1727  * @tsk: TIPC port associated with message
1728  *
1729  * Note: Ancillary data is not captured if not requested by receiver.
1730  *
1731  * Return: 0 if successful, otherwise errno
1732  */
1733 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1734 				 struct tipc_sock *tsk)
1735 {
1736 	struct tipc_msg *hdr;
1737 	u32 data[3] = {0,};
1738 	bool has_addr;
1739 	int dlen, rc;
1740 
1741 	if (likely(m->msg_controllen == 0))
1742 		return 0;
1743 
1744 	hdr = buf_msg(skb);
1745 	dlen = msg_data_sz(hdr);
1746 
1747 	/* Capture errored message object, if any */
1748 	if (msg_errcode(hdr)) {
1749 		if (skb_linearize(skb))
1750 			return -ENOMEM;
1751 		hdr = buf_msg(skb);
1752 		data[0] = msg_errcode(hdr);
1753 		data[1] = dlen;
1754 		rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1755 		if (rc || !dlen)
1756 			return rc;
1757 		rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1758 		if (rc)
1759 			return rc;
1760 	}
1761 
1762 	/* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1763 	switch (msg_type(hdr)) {
1764 	case TIPC_NAMED_MSG:
1765 		has_addr = true;
1766 		data[0] = msg_nametype(hdr);
1767 		data[1] = msg_namelower(hdr);
1768 		data[2] = data[1];
1769 		break;
1770 	case TIPC_MCAST_MSG:
1771 		has_addr = true;
1772 		data[0] = msg_nametype(hdr);
1773 		data[1] = msg_namelower(hdr);
1774 		data[2] = msg_nameupper(hdr);
1775 		break;
1776 	case TIPC_CONN_MSG:
1777 		has_addr = !!tsk->conn_addrtype;
1778 		data[0] = msg_nametype(&tsk->phdr);
1779 		data[1] = msg_nameinst(&tsk->phdr);
1780 		data[2] = data[1];
1781 		break;
1782 	default:
1783 		has_addr = false;
1784 	}
1785 	if (!has_addr)
1786 		return 0;
1787 	return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
1788 }
1789 
1790 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1791 {
1792 	struct sock *sk = &tsk->sk;
1793 	struct sk_buff *skb = NULL;
1794 	struct tipc_msg *msg;
1795 	u32 peer_port = tsk_peer_port(tsk);
1796 	u32 dnode = tsk_peer_node(tsk);
1797 
1798 	if (!tipc_sk_connected(sk))
1799 		return NULL;
1800 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1801 			      dnode, tsk_own_node(tsk), peer_port,
1802 			      tsk->portid, TIPC_OK);
1803 	if (!skb)
1804 		return NULL;
1805 	msg = buf_msg(skb);
1806 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1807 	tsk->rcv_unacked = 0;
1808 
1809 	/* Adjust to and advertize the correct window limit */
1810 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1811 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1812 		msg_set_adv_win(msg, tsk->rcv_win);
1813 	}
1814 	return skb;
1815 }
1816 
1817 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1818 {
1819 	struct sk_buff *skb;
1820 
1821 	skb = tipc_sk_build_ack(tsk);
1822 	if (!skb)
1823 		return;
1824 
1825 	tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1826 			   msg_link_selector(buf_msg(skb)));
1827 }
1828 
1829 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1830 {
1831 	struct sock *sk = sock->sk;
1832 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1833 	long timeo = *timeop;
1834 	int err = sock_error(sk);
1835 
1836 	if (err)
1837 		return err;
1838 
1839 	for (;;) {
1840 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1841 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1842 				err = -ENOTCONN;
1843 				break;
1844 			}
1845 			add_wait_queue(sk_sleep(sk), &wait);
1846 			release_sock(sk);
1847 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1848 			sched_annotate_sleep();
1849 			lock_sock(sk);
1850 			remove_wait_queue(sk_sleep(sk), &wait);
1851 		}
1852 		err = 0;
1853 		if (!skb_queue_empty(&sk->sk_receive_queue))
1854 			break;
1855 		err = -EAGAIN;
1856 		if (!timeo)
1857 			break;
1858 		err = sock_intr_errno(timeo);
1859 		if (signal_pending(current))
1860 			break;
1861 
1862 		err = sock_error(sk);
1863 		if (err)
1864 			break;
1865 	}
1866 	*timeop = timeo;
1867 	return err;
1868 }
1869 
1870 /**
1871  * tipc_recvmsg - receive packet-oriented message
1872  * @sock: network socket
1873  * @m: descriptor for message info
1874  * @buflen: length of user buffer area
1875  * @flags: receive flags
1876  *
1877  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1878  * If the complete message doesn't fit in user area, truncate it.
1879  *
1880  * Return: size of returned message data, errno otherwise
1881  */
1882 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1883 			size_t buflen,	int flags)
1884 {
1885 	struct sock *sk = sock->sk;
1886 	bool connected = !tipc_sk_type_connectionless(sk);
1887 	struct tipc_sock *tsk = tipc_sk(sk);
1888 	int rc, err, hlen, dlen, copy;
1889 	struct tipc_skb_cb *skb_cb;
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 		skb_cb = TIPC_SKB_CB(skb);
1914 		hdr = buf_msg(skb);
1915 		dlen = msg_data_sz(hdr);
1916 		hlen = msg_hdr_sz(hdr);
1917 		err = msg_errcode(hdr);
1918 		grp_evt = msg_is_grp_evt(hdr);
1919 		if (likely(dlen || err))
1920 			break;
1921 		tsk_advance_rx_queue(sk);
1922 	} while (1);
1923 
1924 	/* Collect msg meta data, including error code and rejected data */
1925 	tipc_sk_set_orig_addr(m, skb);
1926 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1927 	if (unlikely(rc))
1928 		goto exit;
1929 	hdr = buf_msg(skb);
1930 
1931 	/* Capture data if non-error msg, otherwise just set return value */
1932 	if (likely(!err)) {
1933 		int offset = skb_cb->bytes_read;
1934 
1935 		copy = min_t(int, dlen - offset, buflen);
1936 		rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1937 		if (unlikely(rc))
1938 			goto exit;
1939 		if (unlikely(offset + copy < dlen)) {
1940 			if (flags & MSG_EOR) {
1941 				if (!(flags & MSG_PEEK))
1942 					skb_cb->bytes_read = offset + copy;
1943 			} else {
1944 				m->msg_flags |= MSG_TRUNC;
1945 				skb_cb->bytes_read = 0;
1946 			}
1947 		} else {
1948 			if (flags & MSG_EOR)
1949 				m->msg_flags |= MSG_EOR;
1950 			skb_cb->bytes_read = 0;
1951 		}
1952 	} else {
1953 		copy = 0;
1954 		rc = 0;
1955 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1956 			rc = -ECONNRESET;
1957 			goto exit;
1958 		}
1959 	}
1960 
1961 	/* Mark message as group event if applicable */
1962 	if (unlikely(grp_evt)) {
1963 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1964 			m->msg_flags |= MSG_EOR;
1965 		m->msg_flags |= MSG_OOB;
1966 		copy = 0;
1967 	}
1968 
1969 	/* Caption of data or error code/rejected data was successful */
1970 	if (unlikely(flags & MSG_PEEK))
1971 		goto exit;
1972 
1973 	/* Send group flow control advertisement when applicable */
1974 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1975 		__skb_queue_head_init(&xmitq);
1976 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1977 					  msg_orignode(hdr), msg_origport(hdr),
1978 					  &xmitq);
1979 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1980 	}
1981 
1982 	if (skb_cb->bytes_read)
1983 		goto exit;
1984 
1985 	tsk_advance_rx_queue(sk);
1986 
1987 	if (likely(!connected))
1988 		goto exit;
1989 
1990 	/* Send connection flow control advertisement when applicable */
1991 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1992 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1993 		tipc_sk_send_ack(tsk);
1994 exit:
1995 	release_sock(sk);
1996 	return rc ? rc : copy;
1997 }
1998 
1999 /**
2000  * tipc_recvstream - receive stream-oriented data
2001  * @sock: network socket
2002  * @m: descriptor for message info
2003  * @buflen: total size of user buffer area
2004  * @flags: receive flags
2005  *
2006  * Used for SOCK_STREAM messages only.  If not enough data is available
2007  * will optionally wait for more; never truncates data.
2008  *
2009  * Return: size of returned message data, errno otherwise
2010  */
2011 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2012 			   size_t buflen, int flags)
2013 {
2014 	struct sock *sk = sock->sk;
2015 	struct tipc_sock *tsk = tipc_sk(sk);
2016 	struct sk_buff *skb;
2017 	struct tipc_msg *hdr;
2018 	struct tipc_skb_cb *skb_cb;
2019 	bool peek = flags & MSG_PEEK;
2020 	int offset, required, copy, copied = 0;
2021 	int hlen, dlen, err, rc;
2022 	long timeout;
2023 
2024 	/* Catch invalid receive attempts */
2025 	if (unlikely(!buflen))
2026 		return -EINVAL;
2027 
2028 	lock_sock(sk);
2029 
2030 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
2031 		rc = -ENOTCONN;
2032 		goto exit;
2033 	}
2034 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2035 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2036 
2037 	do {
2038 		/* Look at first msg in receive queue; wait if necessary */
2039 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
2040 		if (unlikely(rc))
2041 			break;
2042 		skb = skb_peek(&sk->sk_receive_queue);
2043 		skb_cb = TIPC_SKB_CB(skb);
2044 		hdr = buf_msg(skb);
2045 		dlen = msg_data_sz(hdr);
2046 		hlen = msg_hdr_sz(hdr);
2047 		err = msg_errcode(hdr);
2048 
2049 		/* Discard any empty non-errored (SYN-) message */
2050 		if (unlikely(!dlen && !err)) {
2051 			tsk_advance_rx_queue(sk);
2052 			continue;
2053 		}
2054 
2055 		/* Collect msg meta data, incl. error code and rejected data */
2056 		if (!copied) {
2057 			tipc_sk_set_orig_addr(m, skb);
2058 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
2059 			if (rc)
2060 				break;
2061 			hdr = buf_msg(skb);
2062 		}
2063 
2064 		/* Copy data if msg ok, otherwise return error/partial data */
2065 		if (likely(!err)) {
2066 			offset = skb_cb->bytes_read;
2067 			copy = min_t(int, dlen - offset, buflen - copied);
2068 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2069 			if (unlikely(rc))
2070 				break;
2071 			copied += copy;
2072 			offset += copy;
2073 			if (unlikely(offset < dlen)) {
2074 				if (!peek)
2075 					skb_cb->bytes_read = offset;
2076 				break;
2077 			}
2078 		} else {
2079 			rc = 0;
2080 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2081 				rc = -ECONNRESET;
2082 			if (copied || rc)
2083 				break;
2084 		}
2085 
2086 		if (unlikely(peek))
2087 			break;
2088 
2089 		tsk_advance_rx_queue(sk);
2090 
2091 		/* Send connection flow control advertisement when applicable */
2092 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2093 		if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2094 			tipc_sk_send_ack(tsk);
2095 
2096 		/* Exit if all requested data or FIN/error received */
2097 		if (copied == buflen || err)
2098 			break;
2099 
2100 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2101 exit:
2102 	release_sock(sk);
2103 	return copied ? copied : rc;
2104 }
2105 
2106 /**
2107  * tipc_write_space - wake up thread if port congestion is released
2108  * @sk: socket
2109  */
2110 static void tipc_write_space(struct sock *sk)
2111 {
2112 	struct socket_wq *wq;
2113 
2114 	rcu_read_lock();
2115 	wq = rcu_dereference(sk->sk_wq);
2116 	if (skwq_has_sleeper(wq))
2117 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2118 						EPOLLWRNORM | EPOLLWRBAND);
2119 	rcu_read_unlock();
2120 }
2121 
2122 /**
2123  * tipc_data_ready - wake up threads to indicate messages have been received
2124  * @sk: socket
2125  */
2126 static void tipc_data_ready(struct sock *sk)
2127 {
2128 	struct socket_wq *wq;
2129 
2130 	rcu_read_lock();
2131 	wq = rcu_dereference(sk->sk_wq);
2132 	if (skwq_has_sleeper(wq))
2133 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2134 						EPOLLRDNORM | EPOLLRDBAND);
2135 	rcu_read_unlock();
2136 }
2137 
2138 static void tipc_sock_destruct(struct sock *sk)
2139 {
2140 	__skb_queue_purge(&sk->sk_receive_queue);
2141 }
2142 
2143 static void tipc_sk_proto_rcv(struct sock *sk,
2144 			      struct sk_buff_head *inputq,
2145 			      struct sk_buff_head *xmitq)
2146 {
2147 	struct sk_buff *skb = __skb_dequeue(inputq);
2148 	struct tipc_sock *tsk = tipc_sk(sk);
2149 	struct tipc_msg *hdr = buf_msg(skb);
2150 	struct tipc_group *grp = tsk->group;
2151 	bool wakeup = false;
2152 
2153 	switch (msg_user(hdr)) {
2154 	case CONN_MANAGER:
2155 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2156 		return;
2157 	case SOCK_WAKEUP:
2158 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2159 		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2160 		smp_wmb();
2161 		tsk->cong_link_cnt--;
2162 		wakeup = true;
2163 		tipc_sk_push_backlog(tsk, false);
2164 		break;
2165 	case GROUP_PROTOCOL:
2166 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2167 		break;
2168 	case TOP_SRV:
2169 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2170 				      hdr, inputq, xmitq);
2171 		break;
2172 	default:
2173 		break;
2174 	}
2175 
2176 	if (wakeup)
2177 		sk->sk_write_space(sk);
2178 
2179 	kfree_skb(skb);
2180 }
2181 
2182 /**
2183  * tipc_sk_filter_connect - check incoming message for a connection-based socket
2184  * @tsk: TIPC socket
2185  * @skb: pointer to message buffer.
2186  * @xmitq: for Nagle ACK if any
2187  * Return: true if message should be added to receive queue, false otherwise
2188  */
2189 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2190 				   struct sk_buff_head *xmitq)
2191 {
2192 	struct sock *sk = &tsk->sk;
2193 	struct net *net = sock_net(sk);
2194 	struct tipc_msg *hdr = buf_msg(skb);
2195 	bool con_msg = msg_connected(hdr);
2196 	u32 pport = tsk_peer_port(tsk);
2197 	u32 pnode = tsk_peer_node(tsk);
2198 	u32 oport = msg_origport(hdr);
2199 	u32 onode = msg_orignode(hdr);
2200 	int err = msg_errcode(hdr);
2201 	unsigned long delay;
2202 
2203 	if (unlikely(msg_mcast(hdr)))
2204 		return false;
2205 	tsk->oneway = 0;
2206 
2207 	switch (sk->sk_state) {
2208 	case TIPC_CONNECTING:
2209 		/* Setup ACK */
2210 		if (likely(con_msg)) {
2211 			if (err)
2212 				break;
2213 			tipc_sk_finish_conn(tsk, oport, onode);
2214 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2215 			/* ACK+ message with data is added to receive queue */
2216 			if (msg_data_sz(hdr))
2217 				return true;
2218 			/* Empty ACK-, - wake up sleeping connect() and drop */
2219 			sk->sk_state_change(sk);
2220 			msg_set_dest_droppable(hdr, 1);
2221 			return false;
2222 		}
2223 		/* Ignore connectionless message if not from listening socket */
2224 		if (oport != pport || onode != pnode)
2225 			return false;
2226 
2227 		/* Rejected SYN */
2228 		if (err != TIPC_ERR_OVERLOAD)
2229 			break;
2230 
2231 		/* Prepare for new setup attempt if we have a SYN clone */
2232 		if (skb_queue_empty(&sk->sk_write_queue))
2233 			break;
2234 		get_random_bytes(&delay, 2);
2235 		delay %= (tsk->conn_timeout / 4);
2236 		delay = msecs_to_jiffies(delay + 100);
2237 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2238 		return false;
2239 	case TIPC_OPEN:
2240 	case TIPC_DISCONNECTING:
2241 		return false;
2242 	case TIPC_LISTEN:
2243 		/* Accept only SYN message */
2244 		if (!msg_is_syn(hdr) &&
2245 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2246 			return false;
2247 		if (!con_msg && !err)
2248 			return true;
2249 		return false;
2250 	case TIPC_ESTABLISHED:
2251 		if (!skb_queue_empty(&sk->sk_write_queue))
2252 			tipc_sk_push_backlog(tsk, false);
2253 		/* Accept only connection-based messages sent by peer */
2254 		if (likely(con_msg && !err && pport == oport &&
2255 			   pnode == onode)) {
2256 			if (msg_ack_required(hdr)) {
2257 				struct sk_buff *skb;
2258 
2259 				skb = tipc_sk_build_ack(tsk);
2260 				if (skb) {
2261 					msg_set_nagle_ack(buf_msg(skb));
2262 					__skb_queue_tail(xmitq, skb);
2263 				}
2264 			}
2265 			return true;
2266 		}
2267 		if (!tsk_peer_msg(tsk, hdr))
2268 			return false;
2269 		if (!err)
2270 			return true;
2271 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2272 		tipc_node_remove_conn(net, pnode, tsk->portid);
2273 		sk->sk_state_change(sk);
2274 		return true;
2275 	default:
2276 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2277 	}
2278 	/* Abort connection setup attempt */
2279 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2280 	sk->sk_err = ECONNREFUSED;
2281 	sk->sk_state_change(sk);
2282 	return true;
2283 }
2284 
2285 /**
2286  * rcvbuf_limit - get proper overload limit of socket receive queue
2287  * @sk: socket
2288  * @skb: message
2289  *
2290  * For connection oriented messages, irrespective of importance,
2291  * default queue limit is 2 MB.
2292  *
2293  * For connectionless messages, queue limits are based on message
2294  * importance as follows:
2295  *
2296  * TIPC_LOW_IMPORTANCE       (2 MB)
2297  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2298  * TIPC_HIGH_IMPORTANCE      (8 MB)
2299  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2300  *
2301  * Return: overload limit according to corresponding message importance
2302  */
2303 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2304 {
2305 	struct tipc_sock *tsk = tipc_sk(sk);
2306 	struct tipc_msg *hdr = buf_msg(skb);
2307 
2308 	if (unlikely(msg_in_group(hdr)))
2309 		return READ_ONCE(sk->sk_rcvbuf);
2310 
2311 	if (unlikely(!msg_connected(hdr)))
2312 		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2313 
2314 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2315 		return READ_ONCE(sk->sk_rcvbuf);
2316 
2317 	return FLOWCTL_MSG_LIM;
2318 }
2319 
2320 /**
2321  * tipc_sk_filter_rcv - validate incoming message
2322  * @sk: socket
2323  * @skb: pointer to message.
2324  * @xmitq: output message area (FIXME)
2325  *
2326  * Enqueues message on receive queue if acceptable; optionally handles
2327  * disconnect indication for a connected socket.
2328  *
2329  * Called with socket lock already taken
2330  */
2331 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2332 			       struct sk_buff_head *xmitq)
2333 {
2334 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2335 	struct tipc_sock *tsk = tipc_sk(sk);
2336 	struct tipc_group *grp = tsk->group;
2337 	struct tipc_msg *hdr = buf_msg(skb);
2338 	struct net *net = sock_net(sk);
2339 	struct sk_buff_head inputq;
2340 	int mtyp = msg_type(hdr);
2341 	int limit, err = TIPC_OK;
2342 
2343 	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2344 	TIPC_SKB_CB(skb)->bytes_read = 0;
2345 	__skb_queue_head_init(&inputq);
2346 	__skb_queue_tail(&inputq, skb);
2347 
2348 	if (unlikely(!msg_isdata(hdr)))
2349 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2350 
2351 	if (unlikely(grp))
2352 		tipc_group_filter_msg(grp, &inputq, xmitq);
2353 
2354 	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2355 		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2356 
2357 	/* Validate and add to receive buffer if there is space */
2358 	while ((skb = __skb_dequeue(&inputq))) {
2359 		hdr = buf_msg(skb);
2360 		limit = rcvbuf_limit(sk, skb);
2361 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2362 		    (!sk_conn && msg_connected(hdr)) ||
2363 		    (!grp && msg_in_group(hdr)))
2364 			err = TIPC_ERR_NO_PORT;
2365 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2366 			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2367 					   "err_overload2!");
2368 			atomic_inc(&sk->sk_drops);
2369 			err = TIPC_ERR_OVERLOAD;
2370 		}
2371 
2372 		if (unlikely(err)) {
2373 			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2374 				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2375 						      "@filter_rcv!");
2376 				__skb_queue_tail(xmitq, skb);
2377 			}
2378 			err = TIPC_OK;
2379 			continue;
2380 		}
2381 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2382 		skb_set_owner_r(skb, sk);
2383 		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2384 					 "rcvq >90% allocated!");
2385 		sk->sk_data_ready(sk);
2386 	}
2387 }
2388 
2389 /**
2390  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2391  * @sk: socket
2392  * @skb: message
2393  *
2394  * Caller must hold socket lock
2395  */
2396 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2397 {
2398 	unsigned int before = sk_rmem_alloc_get(sk);
2399 	struct sk_buff_head xmitq;
2400 	unsigned int added;
2401 
2402 	__skb_queue_head_init(&xmitq);
2403 
2404 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2405 	added = sk_rmem_alloc_get(sk) - before;
2406 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2407 
2408 	/* Send pending response/rejected messages, if any */
2409 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2410 	return 0;
2411 }
2412 
2413 /**
2414  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2415  *                   inputq and try adding them to socket or backlog queue
2416  * @inputq: list of incoming buffers with potentially different destinations
2417  * @sk: socket where the buffers should be enqueued
2418  * @dport: port number for the socket
2419  * @xmitq: output queue
2420  *
2421  * Caller must hold socket lock
2422  */
2423 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2424 			    u32 dport, struct sk_buff_head *xmitq)
2425 {
2426 	unsigned long time_limit = jiffies + 2;
2427 	struct sk_buff *skb;
2428 	unsigned int lim;
2429 	atomic_t *dcnt;
2430 	u32 onode;
2431 
2432 	while (skb_queue_len(inputq)) {
2433 		if (unlikely(time_after_eq(jiffies, time_limit)))
2434 			return;
2435 
2436 		skb = tipc_skb_dequeue(inputq, dport);
2437 		if (unlikely(!skb))
2438 			return;
2439 
2440 		/* Add message directly to receive queue if possible */
2441 		if (!sock_owned_by_user(sk)) {
2442 			tipc_sk_filter_rcv(sk, skb, xmitq);
2443 			continue;
2444 		}
2445 
2446 		/* Try backlog, compensating for double-counted bytes */
2447 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2448 		if (!sk->sk_backlog.len)
2449 			atomic_set(dcnt, 0);
2450 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2451 		if (likely(!sk_add_backlog(sk, skb, lim))) {
2452 			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2453 						 "bklg & rcvq >90% allocated!");
2454 			continue;
2455 		}
2456 
2457 		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2458 		/* Overload => reject message back to sender */
2459 		onode = tipc_own_addr(sock_net(sk));
2460 		atomic_inc(&sk->sk_drops);
2461 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2462 			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2463 					      "@sk_enqueue!");
2464 			__skb_queue_tail(xmitq, skb);
2465 		}
2466 		break;
2467 	}
2468 }
2469 
2470 /**
2471  * tipc_sk_rcv - handle a chain of incoming buffers
2472  * @net: the associated network namespace
2473  * @inputq: buffer list containing the buffers
2474  * Consumes all buffers in list until inputq is empty
2475  * Note: may be called in multiple threads referring to the same queue
2476  */
2477 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2478 {
2479 	struct sk_buff_head xmitq;
2480 	u32 dnode, dport = 0;
2481 	int err;
2482 	struct tipc_sock *tsk;
2483 	struct sock *sk;
2484 	struct sk_buff *skb;
2485 
2486 	__skb_queue_head_init(&xmitq);
2487 	while (skb_queue_len(inputq)) {
2488 		dport = tipc_skb_peek_port(inputq, dport);
2489 		tsk = tipc_sk_lookup(net, dport);
2490 
2491 		if (likely(tsk)) {
2492 			sk = &tsk->sk;
2493 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2494 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2495 				spin_unlock_bh(&sk->sk_lock.slock);
2496 			}
2497 			/* Send pending response/rejected messages, if any */
2498 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2499 			sock_put(sk);
2500 			continue;
2501 		}
2502 		/* No destination socket => dequeue skb if still there */
2503 		skb = tipc_skb_dequeue(inputq, dport);
2504 		if (!skb)
2505 			return;
2506 
2507 		/* Try secondary lookup if unresolved named message */
2508 		err = TIPC_ERR_NO_PORT;
2509 		if (tipc_msg_lookup_dest(net, skb, &err))
2510 			goto xmit;
2511 
2512 		/* Prepare for message rejection */
2513 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2514 			continue;
2515 
2516 		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2517 xmit:
2518 		dnode = msg_destnode(buf_msg(skb));
2519 		tipc_node_xmit_skb(net, skb, dnode, dport);
2520 	}
2521 }
2522 
2523 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2524 {
2525 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2526 	struct sock *sk = sock->sk;
2527 	int done;
2528 
2529 	do {
2530 		int err = sock_error(sk);
2531 		if (err)
2532 			return err;
2533 		if (!*timeo_p)
2534 			return -ETIMEDOUT;
2535 		if (signal_pending(current))
2536 			return sock_intr_errno(*timeo_p);
2537 		if (sk->sk_state == TIPC_DISCONNECTING)
2538 			break;
2539 
2540 		add_wait_queue(sk_sleep(sk), &wait);
2541 		done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2542 				     &wait);
2543 		remove_wait_queue(sk_sleep(sk), &wait);
2544 	} while (!done);
2545 	return 0;
2546 }
2547 
2548 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2549 {
2550 	if (addr->family != AF_TIPC)
2551 		return false;
2552 	if (addr->addrtype == TIPC_SERVICE_RANGE)
2553 		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2554 	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2555 		addr->addrtype == TIPC_SOCKET_ADDR);
2556 }
2557 
2558 /**
2559  * tipc_connect - establish a connection to another TIPC port
2560  * @sock: socket structure
2561  * @dest: socket address for destination port
2562  * @destlen: size of socket address data structure
2563  * @flags: file-related flags associated with socket
2564  *
2565  * Return: 0 on success, errno otherwise
2566  */
2567 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2568 			int destlen, int flags)
2569 {
2570 	struct sock *sk = sock->sk;
2571 	struct tipc_sock *tsk = tipc_sk(sk);
2572 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2573 	struct msghdr m = {NULL,};
2574 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2575 	int previous;
2576 	int res = 0;
2577 
2578 	if (destlen != sizeof(struct sockaddr_tipc))
2579 		return -EINVAL;
2580 
2581 	lock_sock(sk);
2582 
2583 	if (tsk->group) {
2584 		res = -EINVAL;
2585 		goto exit;
2586 	}
2587 
2588 	if (dst->family == AF_UNSPEC) {
2589 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2590 		if (!tipc_sk_type_connectionless(sk))
2591 			res = -EINVAL;
2592 		goto exit;
2593 	}
2594 	if (!tipc_sockaddr_is_sane(dst)) {
2595 		res = -EINVAL;
2596 		goto exit;
2597 	}
2598 	/* DGRAM/RDM connect(), just save the destaddr */
2599 	if (tipc_sk_type_connectionless(sk)) {
2600 		memcpy(&tsk->peer, dest, destlen);
2601 		goto exit;
2602 	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2603 		res = -EINVAL;
2604 		goto exit;
2605 	}
2606 
2607 	previous = sk->sk_state;
2608 
2609 	switch (sk->sk_state) {
2610 	case TIPC_OPEN:
2611 		/* Send a 'SYN-' to destination */
2612 		m.msg_name = dest;
2613 		m.msg_namelen = destlen;
2614 
2615 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2616 		 * indicate send_msg() is never blocked.
2617 		 */
2618 		if (!timeout)
2619 			m.msg_flags = MSG_DONTWAIT;
2620 
2621 		res = __tipc_sendmsg(sock, &m, 0);
2622 		if ((res < 0) && (res != -EWOULDBLOCK))
2623 			goto exit;
2624 
2625 		/* Just entered TIPC_CONNECTING state; the only
2626 		 * difference is that return value in non-blocking
2627 		 * case is EINPROGRESS, rather than EALREADY.
2628 		 */
2629 		res = -EINPROGRESS;
2630 		fallthrough;
2631 	case TIPC_CONNECTING:
2632 		if (!timeout) {
2633 			if (previous == TIPC_CONNECTING)
2634 				res = -EALREADY;
2635 			goto exit;
2636 		}
2637 		timeout = msecs_to_jiffies(timeout);
2638 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2639 		res = tipc_wait_for_connect(sock, &timeout);
2640 		break;
2641 	case TIPC_ESTABLISHED:
2642 		res = -EISCONN;
2643 		break;
2644 	default:
2645 		res = -EINVAL;
2646 	}
2647 
2648 exit:
2649 	release_sock(sk);
2650 	return res;
2651 }
2652 
2653 /**
2654  * tipc_listen - allow socket to listen for incoming connections
2655  * @sock: socket structure
2656  * @len: (unused)
2657  *
2658  * Return: 0 on success, errno otherwise
2659  */
2660 static int tipc_listen(struct socket *sock, int len)
2661 {
2662 	struct sock *sk = sock->sk;
2663 	int res;
2664 
2665 	lock_sock(sk);
2666 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2667 	release_sock(sk);
2668 
2669 	return res;
2670 }
2671 
2672 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2673 {
2674 	struct sock *sk = sock->sk;
2675 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2676 	int err;
2677 
2678 	/* True wake-one mechanism for incoming connections: only
2679 	 * one process gets woken up, not the 'whole herd'.
2680 	 * Since we do not 'race & poll' for established sockets
2681 	 * anymore, the common case will execute the loop only once.
2682 	*/
2683 	for (;;) {
2684 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2685 			add_wait_queue(sk_sleep(sk), &wait);
2686 			release_sock(sk);
2687 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2688 			lock_sock(sk);
2689 			remove_wait_queue(sk_sleep(sk), &wait);
2690 		}
2691 		err = 0;
2692 		if (!skb_queue_empty(&sk->sk_receive_queue))
2693 			break;
2694 		err = -EAGAIN;
2695 		if (!timeo)
2696 			break;
2697 		err = sock_intr_errno(timeo);
2698 		if (signal_pending(current))
2699 			break;
2700 	}
2701 	return err;
2702 }
2703 
2704 /**
2705  * tipc_accept - wait for connection request
2706  * @sock: listening socket
2707  * @new_sock: new socket that is to be connected
2708  * @flags: file-related flags associated with socket
2709  * @kern: caused by kernel or by userspace?
2710  *
2711  * Return: 0 on success, errno otherwise
2712  */
2713 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2714 		       bool kern)
2715 {
2716 	struct sock *new_sk, *sk = sock->sk;
2717 	struct tipc_sock *new_tsock;
2718 	struct msghdr m = {NULL,};
2719 	struct tipc_msg *msg;
2720 	struct sk_buff *buf;
2721 	long timeo;
2722 	int res;
2723 
2724 	lock_sock(sk);
2725 
2726 	if (sk->sk_state != TIPC_LISTEN) {
2727 		res = -EINVAL;
2728 		goto exit;
2729 	}
2730 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2731 	res = tipc_wait_for_accept(sock, timeo);
2732 	if (res)
2733 		goto exit;
2734 
2735 	buf = skb_peek(&sk->sk_receive_queue);
2736 
2737 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2738 	if (res)
2739 		goto exit;
2740 	security_sk_clone(sock->sk, new_sock->sk);
2741 
2742 	new_sk = new_sock->sk;
2743 	new_tsock = tipc_sk(new_sk);
2744 	msg = buf_msg(buf);
2745 
2746 	/* we lock on new_sk; but lockdep sees the lock on sk */
2747 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2748 
2749 	/*
2750 	 * Reject any stray messages received by new socket
2751 	 * before the socket lock was taken (very, very unlikely)
2752 	 */
2753 	tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2754 
2755 	/* Connect new socket to it's peer */
2756 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2757 
2758 	tsk_set_importance(new_sk, msg_importance(msg));
2759 	if (msg_named(msg)) {
2760 		new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2761 		msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2762 		msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2763 	}
2764 
2765 	/*
2766 	 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2767 	 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2768 	 */
2769 	if (!msg_data_sz(msg)) {
2770 		tsk_advance_rx_queue(sk);
2771 	} else {
2772 		__skb_dequeue(&sk->sk_receive_queue);
2773 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2774 		skb_set_owner_r(buf, new_sk);
2775 	}
2776 	__tipc_sendstream(new_sock, &m, 0);
2777 	release_sock(new_sk);
2778 exit:
2779 	release_sock(sk);
2780 	return res;
2781 }
2782 
2783 /**
2784  * tipc_shutdown - shutdown socket connection
2785  * @sock: socket structure
2786  * @how: direction to close (must be SHUT_RDWR)
2787  *
2788  * Terminates connection (if necessary), then purges socket's receive queue.
2789  *
2790  * Return: 0 on success, errno otherwise
2791  */
2792 static int tipc_shutdown(struct socket *sock, int how)
2793 {
2794 	struct sock *sk = sock->sk;
2795 	int res;
2796 
2797 	if (how != SHUT_RDWR)
2798 		return -EINVAL;
2799 
2800 	lock_sock(sk);
2801 
2802 	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2803 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2804 	sk->sk_shutdown = SHUTDOWN_MASK;
2805 
2806 	if (sk->sk_state == TIPC_DISCONNECTING) {
2807 		/* Discard any unreceived messages */
2808 		__skb_queue_purge(&sk->sk_receive_queue);
2809 
2810 		res = 0;
2811 	} else {
2812 		res = -ENOTCONN;
2813 	}
2814 	/* Wake up anyone sleeping in poll. */
2815 	sk->sk_state_change(sk);
2816 
2817 	release_sock(sk);
2818 	return res;
2819 }
2820 
2821 static void tipc_sk_check_probing_state(struct sock *sk,
2822 					struct sk_buff_head *list)
2823 {
2824 	struct tipc_sock *tsk = tipc_sk(sk);
2825 	u32 pnode = tsk_peer_node(tsk);
2826 	u32 pport = tsk_peer_port(tsk);
2827 	u32 self = tsk_own_node(tsk);
2828 	u32 oport = tsk->portid;
2829 	struct sk_buff *skb;
2830 
2831 	if (tsk->probe_unacked) {
2832 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2833 		sk->sk_err = ECONNABORTED;
2834 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2835 		sk->sk_state_change(sk);
2836 		return;
2837 	}
2838 	/* Prepare new probe */
2839 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2840 			      pnode, self, pport, oport, TIPC_OK);
2841 	if (skb)
2842 		__skb_queue_tail(list, skb);
2843 	tsk->probe_unacked = true;
2844 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2845 }
2846 
2847 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2848 {
2849 	struct tipc_sock *tsk = tipc_sk(sk);
2850 
2851 	/* Try again later if dest link is congested */
2852 	if (tsk->cong_link_cnt) {
2853 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2854 		return;
2855 	}
2856 	/* Prepare SYN for retransmit */
2857 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2858 }
2859 
2860 static void tipc_sk_timeout(struct timer_list *t)
2861 {
2862 	struct sock *sk = from_timer(sk, t, sk_timer);
2863 	struct tipc_sock *tsk = tipc_sk(sk);
2864 	u32 pnode = tsk_peer_node(tsk);
2865 	struct sk_buff_head list;
2866 	int rc = 0;
2867 
2868 	__skb_queue_head_init(&list);
2869 	bh_lock_sock(sk);
2870 
2871 	/* Try again later if socket is busy */
2872 	if (sock_owned_by_user(sk)) {
2873 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2874 		bh_unlock_sock(sk);
2875 		sock_put(sk);
2876 		return;
2877 	}
2878 
2879 	if (sk->sk_state == TIPC_ESTABLISHED)
2880 		tipc_sk_check_probing_state(sk, &list);
2881 	else if (sk->sk_state == TIPC_CONNECTING)
2882 		tipc_sk_retry_connect(sk, &list);
2883 
2884 	bh_unlock_sock(sk);
2885 
2886 	if (!skb_queue_empty(&list))
2887 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2888 
2889 	/* SYN messages may cause link congestion */
2890 	if (rc == -ELINKCONG) {
2891 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2892 		tsk->cong_link_cnt = 1;
2893 	}
2894 	sock_put(sk);
2895 }
2896 
2897 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2898 {
2899 	struct sock *sk = &tsk->sk;
2900 	struct net *net = sock_net(sk);
2901 	struct tipc_socket_addr skaddr;
2902 	struct publication *p;
2903 	u32 key;
2904 
2905 	if (tipc_sk_connected(sk))
2906 		return -EINVAL;
2907 	key = tsk->portid + tsk->pub_count + 1;
2908 	if (key == tsk->portid)
2909 		return -EADDRINUSE;
2910 	skaddr.ref = tsk->portid;
2911 	skaddr.node = tipc_own_addr(net);
2912 	p = tipc_nametbl_publish(net, ua, &skaddr, key);
2913 	if (unlikely(!p))
2914 		return -EINVAL;
2915 
2916 	list_add(&p->binding_sock, &tsk->publications);
2917 	tsk->pub_count++;
2918 	tsk->published = true;
2919 	return 0;
2920 }
2921 
2922 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2923 {
2924 	struct net *net = sock_net(&tsk->sk);
2925 	struct publication *safe, *p;
2926 	struct tipc_uaddr _ua;
2927 	int rc = -EINVAL;
2928 
2929 	list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2930 		if (!ua) {
2931 			tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2932 				   p->sr.type, p->sr.lower, p->sr.upper);
2933 			tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2934 			continue;
2935 		}
2936 		/* Unbind specific publication */
2937 		if (p->scope != ua->scope)
2938 			continue;
2939 		if (p->sr.type != ua->sr.type)
2940 			continue;
2941 		if (p->sr.lower != ua->sr.lower)
2942 			continue;
2943 		if (p->sr.upper != ua->sr.upper)
2944 			break;
2945 		tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2946 		rc = 0;
2947 		break;
2948 	}
2949 	if (list_empty(&tsk->publications)) {
2950 		tsk->published = 0;
2951 		rc = 0;
2952 	}
2953 	return rc;
2954 }
2955 
2956 /* tipc_sk_reinit: set non-zero address in all existing sockets
2957  *                 when we go from standalone to network mode.
2958  */
2959 void tipc_sk_reinit(struct net *net)
2960 {
2961 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2962 	struct rhashtable_iter iter;
2963 	struct tipc_sock *tsk;
2964 	struct tipc_msg *msg;
2965 
2966 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2967 
2968 	do {
2969 		rhashtable_walk_start(&iter);
2970 
2971 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2972 			sock_hold(&tsk->sk);
2973 			rhashtable_walk_stop(&iter);
2974 			lock_sock(&tsk->sk);
2975 			msg = &tsk->phdr;
2976 			msg_set_prevnode(msg, tipc_own_addr(net));
2977 			msg_set_orignode(msg, tipc_own_addr(net));
2978 			release_sock(&tsk->sk);
2979 			rhashtable_walk_start(&iter);
2980 			sock_put(&tsk->sk);
2981 		}
2982 
2983 		rhashtable_walk_stop(&iter);
2984 	} while (tsk == ERR_PTR(-EAGAIN));
2985 
2986 	rhashtable_walk_exit(&iter);
2987 }
2988 
2989 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2990 {
2991 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2992 	struct tipc_sock *tsk;
2993 
2994 	rcu_read_lock();
2995 	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2996 	if (tsk)
2997 		sock_hold(&tsk->sk);
2998 	rcu_read_unlock();
2999 
3000 	return tsk;
3001 }
3002 
3003 static int tipc_sk_insert(struct tipc_sock *tsk)
3004 {
3005 	struct sock *sk = &tsk->sk;
3006 	struct net *net = sock_net(sk);
3007 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3008 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3009 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
3010 
3011 	while (remaining--) {
3012 		portid++;
3013 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3014 			portid = TIPC_MIN_PORT;
3015 		tsk->portid = portid;
3016 		sock_hold(&tsk->sk);
3017 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3018 						   tsk_rht_params))
3019 			return 0;
3020 		sock_put(&tsk->sk);
3021 	}
3022 
3023 	return -1;
3024 }
3025 
3026 static void tipc_sk_remove(struct tipc_sock *tsk)
3027 {
3028 	struct sock *sk = &tsk->sk;
3029 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3030 
3031 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3032 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3033 		__sock_put(sk);
3034 	}
3035 }
3036 
3037 static const struct rhashtable_params tsk_rht_params = {
3038 	.nelem_hint = 192,
3039 	.head_offset = offsetof(struct tipc_sock, node),
3040 	.key_offset = offsetof(struct tipc_sock, portid),
3041 	.key_len = sizeof(u32), /* portid */
3042 	.max_size = 1048576,
3043 	.min_size = 256,
3044 	.automatic_shrinking = true,
3045 };
3046 
3047 int tipc_sk_rht_init(struct net *net)
3048 {
3049 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3050 
3051 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3052 }
3053 
3054 void tipc_sk_rht_destroy(struct net *net)
3055 {
3056 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3057 
3058 	/* Wait for socket readers to complete */
3059 	synchronize_net();
3060 
3061 	rhashtable_destroy(&tn->sk_rht);
3062 }
3063 
3064 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3065 {
3066 	struct net *net = sock_net(&tsk->sk);
3067 	struct tipc_group *grp = tsk->group;
3068 	struct tipc_msg *hdr = &tsk->phdr;
3069 	struct tipc_uaddr ua;
3070 	int rc;
3071 
3072 	if (mreq->type < TIPC_RESERVED_TYPES)
3073 		return -EACCES;
3074 	if (mreq->scope > TIPC_NODE_SCOPE)
3075 		return -EINVAL;
3076 	if (mreq->scope != TIPC_NODE_SCOPE)
3077 		mreq->scope = TIPC_CLUSTER_SCOPE;
3078 	if (grp)
3079 		return -EACCES;
3080 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3081 	if (!grp)
3082 		return -ENOMEM;
3083 	tsk->group = grp;
3084 	msg_set_lookup_scope(hdr, mreq->scope);
3085 	msg_set_nametype(hdr, mreq->type);
3086 	msg_set_dest_droppable(hdr, true);
3087 	tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3088 		   mreq->type, mreq->instance, mreq->instance);
3089 	tipc_nametbl_build_group(net, grp, &ua);
3090 	rc = tipc_sk_publish(tsk, &ua);
3091 	if (rc) {
3092 		tipc_group_delete(net, grp);
3093 		tsk->group = NULL;
3094 		return rc;
3095 	}
3096 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
3097 	tsk->mc_method.rcast = true;
3098 	tsk->mc_method.mandatory = true;
3099 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3100 	return rc;
3101 }
3102 
3103 static int tipc_sk_leave(struct tipc_sock *tsk)
3104 {
3105 	struct net *net = sock_net(&tsk->sk);
3106 	struct tipc_group *grp = tsk->group;
3107 	struct tipc_uaddr ua;
3108 	int scope;
3109 
3110 	if (!grp)
3111 		return -EINVAL;
3112 	ua.addrtype = TIPC_SERVICE_RANGE;
3113 	tipc_group_self(grp, &ua.sr, &scope);
3114 	ua.scope = scope;
3115 	tipc_group_delete(net, grp);
3116 	tsk->group = NULL;
3117 	tipc_sk_withdraw(tsk, &ua);
3118 	return 0;
3119 }
3120 
3121 /**
3122  * tipc_setsockopt - set socket option
3123  * @sock: socket structure
3124  * @lvl: option level
3125  * @opt: option identifier
3126  * @ov: pointer to new option value
3127  * @ol: length of option value
3128  *
3129  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3130  * (to ease compatibility).
3131  *
3132  * Return: 0 on success, errno otherwise
3133  */
3134 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3135 			   sockptr_t ov, unsigned int ol)
3136 {
3137 	struct sock *sk = sock->sk;
3138 	struct tipc_sock *tsk = tipc_sk(sk);
3139 	struct tipc_group_req mreq;
3140 	u32 value = 0;
3141 	int res = 0;
3142 
3143 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3144 		return 0;
3145 	if (lvl != SOL_TIPC)
3146 		return -ENOPROTOOPT;
3147 
3148 	switch (opt) {
3149 	case TIPC_IMPORTANCE:
3150 	case TIPC_SRC_DROPPABLE:
3151 	case TIPC_DEST_DROPPABLE:
3152 	case TIPC_CONN_TIMEOUT:
3153 	case TIPC_NODELAY:
3154 		if (ol < sizeof(value))
3155 			return -EINVAL;
3156 		if (copy_from_sockptr(&value, ov, sizeof(u32)))
3157 			return -EFAULT;
3158 		break;
3159 	case TIPC_GROUP_JOIN:
3160 		if (ol < sizeof(mreq))
3161 			return -EINVAL;
3162 		if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3163 			return -EFAULT;
3164 		break;
3165 	default:
3166 		if (!sockptr_is_null(ov) || ol)
3167 			return -EINVAL;
3168 	}
3169 
3170 	lock_sock(sk);
3171 
3172 	switch (opt) {
3173 	case TIPC_IMPORTANCE:
3174 		res = tsk_set_importance(sk, value);
3175 		break;
3176 	case TIPC_SRC_DROPPABLE:
3177 		if (sock->type != SOCK_STREAM)
3178 			tsk_set_unreliable(tsk, value);
3179 		else
3180 			res = -ENOPROTOOPT;
3181 		break;
3182 	case TIPC_DEST_DROPPABLE:
3183 		tsk_set_unreturnable(tsk, value);
3184 		break;
3185 	case TIPC_CONN_TIMEOUT:
3186 		tipc_sk(sk)->conn_timeout = value;
3187 		break;
3188 	case TIPC_MCAST_BROADCAST:
3189 		tsk->mc_method.rcast = false;
3190 		tsk->mc_method.mandatory = true;
3191 		break;
3192 	case TIPC_MCAST_REPLICAST:
3193 		tsk->mc_method.rcast = true;
3194 		tsk->mc_method.mandatory = true;
3195 		break;
3196 	case TIPC_GROUP_JOIN:
3197 		res = tipc_sk_join(tsk, &mreq);
3198 		break;
3199 	case TIPC_GROUP_LEAVE:
3200 		res = tipc_sk_leave(tsk);
3201 		break;
3202 	case TIPC_NODELAY:
3203 		tsk->nodelay = !!value;
3204 		tsk_set_nagle(tsk);
3205 		break;
3206 	default:
3207 		res = -EINVAL;
3208 	}
3209 
3210 	release_sock(sk);
3211 
3212 	return res;
3213 }
3214 
3215 /**
3216  * tipc_getsockopt - get socket option
3217  * @sock: socket structure
3218  * @lvl: option level
3219  * @opt: option identifier
3220  * @ov: receptacle for option value
3221  * @ol: receptacle for length of option value
3222  *
3223  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3224  * (to ease compatibility).
3225  *
3226  * Return: 0 on success, errno otherwise
3227  */
3228 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3229 			   char __user *ov, int __user *ol)
3230 {
3231 	struct sock *sk = sock->sk;
3232 	struct tipc_sock *tsk = tipc_sk(sk);
3233 	struct tipc_service_range seq;
3234 	int len, scope;
3235 	u32 value;
3236 	int res;
3237 
3238 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3239 		return put_user(0, ol);
3240 	if (lvl != SOL_TIPC)
3241 		return -ENOPROTOOPT;
3242 	res = get_user(len, ol);
3243 	if (res)
3244 		return res;
3245 
3246 	lock_sock(sk);
3247 
3248 	switch (opt) {
3249 	case TIPC_IMPORTANCE:
3250 		value = tsk_importance(tsk);
3251 		break;
3252 	case TIPC_SRC_DROPPABLE:
3253 		value = tsk_unreliable(tsk);
3254 		break;
3255 	case TIPC_DEST_DROPPABLE:
3256 		value = tsk_unreturnable(tsk);
3257 		break;
3258 	case TIPC_CONN_TIMEOUT:
3259 		value = tsk->conn_timeout;
3260 		/* no need to set "res", since already 0 at this point */
3261 		break;
3262 	case TIPC_NODE_RECVQ_DEPTH:
3263 		value = 0; /* was tipc_queue_size, now obsolete */
3264 		break;
3265 	case TIPC_SOCK_RECVQ_DEPTH:
3266 		value = skb_queue_len(&sk->sk_receive_queue);
3267 		break;
3268 	case TIPC_SOCK_RECVQ_USED:
3269 		value = sk_rmem_alloc_get(sk);
3270 		break;
3271 	case TIPC_GROUP_JOIN:
3272 		seq.type = 0;
3273 		if (tsk->group)
3274 			tipc_group_self(tsk->group, &seq, &scope);
3275 		value = seq.type;
3276 		break;
3277 	default:
3278 		res = -EINVAL;
3279 	}
3280 
3281 	release_sock(sk);
3282 
3283 	if (res)
3284 		return res;	/* "get" failed */
3285 
3286 	if (len < sizeof(value))
3287 		return -EINVAL;
3288 
3289 	if (copy_to_user(ov, &value, sizeof(value)))
3290 		return -EFAULT;
3291 
3292 	return put_user(sizeof(value), ol);
3293 }
3294 
3295 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3296 {
3297 	struct net *net = sock_net(sock->sk);
3298 	struct tipc_sioc_nodeid_req nr = {0};
3299 	struct tipc_sioc_ln_req lnr;
3300 	void __user *argp = (void __user *)arg;
3301 
3302 	switch (cmd) {
3303 	case SIOCGETLINKNAME:
3304 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3305 			return -EFAULT;
3306 		if (!tipc_node_get_linkname(net,
3307 					    lnr.bearer_id & 0xffff, lnr.peer,
3308 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3309 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3310 				return -EFAULT;
3311 			return 0;
3312 		}
3313 		return -EADDRNOTAVAIL;
3314 	case SIOCGETNODEID:
3315 		if (copy_from_user(&nr, argp, sizeof(nr)))
3316 			return -EFAULT;
3317 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3318 			return -EADDRNOTAVAIL;
3319 		if (copy_to_user(argp, &nr, sizeof(nr)))
3320 			return -EFAULT;
3321 		return 0;
3322 	default:
3323 		return -ENOIOCTLCMD;
3324 	}
3325 }
3326 
3327 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3328 {
3329 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3330 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3331 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3332 
3333 	tsk1->peer.family = AF_TIPC;
3334 	tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3335 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3336 	tsk1->peer.addr.id.ref = tsk2->portid;
3337 	tsk1->peer.addr.id.node = onode;
3338 	tsk2->peer.family = AF_TIPC;
3339 	tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3340 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3341 	tsk2->peer.addr.id.ref = tsk1->portid;
3342 	tsk2->peer.addr.id.node = onode;
3343 
3344 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3345 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3346 	return 0;
3347 }
3348 
3349 /* Protocol switches for the various types of TIPC sockets */
3350 
3351 static const struct proto_ops msg_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		= sock_no_accept,
3359 	.getname	= tipc_getname,
3360 	.poll		= tipc_poll,
3361 	.ioctl		= tipc_ioctl,
3362 	.listen		= sock_no_listen,
3363 	.shutdown	= tipc_shutdown,
3364 	.setsockopt	= tipc_setsockopt,
3365 	.getsockopt	= tipc_getsockopt,
3366 	.sendmsg	= tipc_sendmsg,
3367 	.recvmsg	= tipc_recvmsg,
3368 	.mmap		= sock_no_mmap,
3369 	.sendpage	= sock_no_sendpage
3370 };
3371 
3372 static const struct proto_ops packet_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_send_packet,
3388 	.recvmsg	= tipc_recvmsg,
3389 	.mmap		= sock_no_mmap,
3390 	.sendpage	= sock_no_sendpage
3391 };
3392 
3393 static const struct proto_ops stream_ops = {
3394 	.owner		= THIS_MODULE,
3395 	.family		= AF_TIPC,
3396 	.release	= tipc_release,
3397 	.bind		= tipc_bind,
3398 	.connect	= tipc_connect,
3399 	.socketpair	= tipc_socketpair,
3400 	.accept		= tipc_accept,
3401 	.getname	= tipc_getname,
3402 	.poll		= tipc_poll,
3403 	.ioctl		= tipc_ioctl,
3404 	.listen		= tipc_listen,
3405 	.shutdown	= tipc_shutdown,
3406 	.setsockopt	= tipc_setsockopt,
3407 	.getsockopt	= tipc_getsockopt,
3408 	.sendmsg	= tipc_sendstream,
3409 	.recvmsg	= tipc_recvstream,
3410 	.mmap		= sock_no_mmap,
3411 	.sendpage	= sock_no_sendpage
3412 };
3413 
3414 static const struct net_proto_family tipc_family_ops = {
3415 	.owner		= THIS_MODULE,
3416 	.family		= AF_TIPC,
3417 	.create		= tipc_sk_create
3418 };
3419 
3420 static struct proto tipc_proto = {
3421 	.name		= "TIPC",
3422 	.owner		= THIS_MODULE,
3423 	.obj_size	= sizeof(struct tipc_sock),
3424 	.sysctl_rmem	= sysctl_tipc_rmem
3425 };
3426 
3427 /**
3428  * tipc_socket_init - initialize TIPC socket interface
3429  *
3430  * Return: 0 on success, errno otherwise
3431  */
3432 int tipc_socket_init(void)
3433 {
3434 	int res;
3435 
3436 	res = proto_register(&tipc_proto, 1);
3437 	if (res) {
3438 		pr_err("Failed to register TIPC protocol type\n");
3439 		goto out;
3440 	}
3441 
3442 	res = sock_register(&tipc_family_ops);
3443 	if (res) {
3444 		pr_err("Failed to register TIPC socket type\n");
3445 		proto_unregister(&tipc_proto);
3446 		goto out;
3447 	}
3448  out:
3449 	return res;
3450 }
3451 
3452 /**
3453  * tipc_socket_stop - stop TIPC socket interface
3454  */
3455 void tipc_socket_stop(void)
3456 {
3457 	sock_unregister(tipc_family_ops.family);
3458 	proto_unregister(&tipc_proto);
3459 }
3460 
3461 /* Caller should hold socket lock for the passed tipc socket. */
3462 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3463 {
3464 	u32 peer_node, peer_port;
3465 	u32 conn_type, conn_instance;
3466 	struct nlattr *nest;
3467 
3468 	peer_node = tsk_peer_node(tsk);
3469 	peer_port = tsk_peer_port(tsk);
3470 	conn_type = msg_nametype(&tsk->phdr);
3471 	conn_instance = msg_nameinst(&tsk->phdr);
3472 	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3473 	if (!nest)
3474 		return -EMSGSIZE;
3475 
3476 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3477 		goto msg_full;
3478 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3479 		goto msg_full;
3480 
3481 	if (tsk->conn_addrtype != 0) {
3482 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3483 			goto msg_full;
3484 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3485 			goto msg_full;
3486 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3487 			goto msg_full;
3488 	}
3489 	nla_nest_end(skb, nest);
3490 
3491 	return 0;
3492 
3493 msg_full:
3494 	nla_nest_cancel(skb, nest);
3495 
3496 	return -EMSGSIZE;
3497 }
3498 
3499 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3500 			  *tsk)
3501 {
3502 	struct net *net = sock_net(skb->sk);
3503 	struct sock *sk = &tsk->sk;
3504 
3505 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3506 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3507 		return -EMSGSIZE;
3508 
3509 	if (tipc_sk_connected(sk)) {
3510 		if (__tipc_nl_add_sk_con(skb, tsk))
3511 			return -EMSGSIZE;
3512 	} else if (!list_empty(&tsk->publications)) {
3513 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3514 			return -EMSGSIZE;
3515 	}
3516 	return 0;
3517 }
3518 
3519 /* Caller should hold socket lock for the passed tipc socket. */
3520 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3521 			    struct tipc_sock *tsk)
3522 {
3523 	struct nlattr *attrs;
3524 	void *hdr;
3525 
3526 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3527 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3528 	if (!hdr)
3529 		goto msg_cancel;
3530 
3531 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3532 	if (!attrs)
3533 		goto genlmsg_cancel;
3534 
3535 	if (__tipc_nl_add_sk_info(skb, tsk))
3536 		goto attr_msg_cancel;
3537 
3538 	nla_nest_end(skb, attrs);
3539 	genlmsg_end(skb, hdr);
3540 
3541 	return 0;
3542 
3543 attr_msg_cancel:
3544 	nla_nest_cancel(skb, attrs);
3545 genlmsg_cancel:
3546 	genlmsg_cancel(skb, hdr);
3547 msg_cancel:
3548 	return -EMSGSIZE;
3549 }
3550 
3551 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3552 		    int (*skb_handler)(struct sk_buff *skb,
3553 				       struct netlink_callback *cb,
3554 				       struct tipc_sock *tsk))
3555 {
3556 	struct rhashtable_iter *iter = (void *)cb->args[4];
3557 	struct tipc_sock *tsk;
3558 	int err;
3559 
3560 	rhashtable_walk_start(iter);
3561 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3562 		if (IS_ERR(tsk)) {
3563 			err = PTR_ERR(tsk);
3564 			if (err == -EAGAIN) {
3565 				err = 0;
3566 				continue;
3567 			}
3568 			break;
3569 		}
3570 
3571 		sock_hold(&tsk->sk);
3572 		rhashtable_walk_stop(iter);
3573 		lock_sock(&tsk->sk);
3574 		err = skb_handler(skb, cb, tsk);
3575 		if (err) {
3576 			release_sock(&tsk->sk);
3577 			sock_put(&tsk->sk);
3578 			goto out;
3579 		}
3580 		release_sock(&tsk->sk);
3581 		rhashtable_walk_start(iter);
3582 		sock_put(&tsk->sk);
3583 	}
3584 	rhashtable_walk_stop(iter);
3585 out:
3586 	return skb->len;
3587 }
3588 EXPORT_SYMBOL(tipc_nl_sk_walk);
3589 
3590 int tipc_dump_start(struct netlink_callback *cb)
3591 {
3592 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3593 }
3594 EXPORT_SYMBOL(tipc_dump_start);
3595 
3596 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3597 {
3598 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3599 	struct rhashtable_iter *iter = (void *)cb->args[4];
3600 	struct tipc_net *tn = tipc_net(net);
3601 
3602 	if (!iter) {
3603 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3604 		if (!iter)
3605 			return -ENOMEM;
3606 
3607 		cb->args[4] = (long)iter;
3608 	}
3609 
3610 	rhashtable_walk_enter(&tn->sk_rht, iter);
3611 	return 0;
3612 }
3613 
3614 int tipc_dump_done(struct netlink_callback *cb)
3615 {
3616 	struct rhashtable_iter *hti = (void *)cb->args[4];
3617 
3618 	rhashtable_walk_exit(hti);
3619 	kfree(hti);
3620 	return 0;
3621 }
3622 EXPORT_SYMBOL(tipc_dump_done);
3623 
3624 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3625 			   struct tipc_sock *tsk, u32 sk_filter_state,
3626 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3627 {
3628 	struct sock *sk = &tsk->sk;
3629 	struct nlattr *attrs;
3630 	struct nlattr *stat;
3631 
3632 	/*filter response w.r.t sk_state*/
3633 	if (!(sk_filter_state & (1 << sk->sk_state)))
3634 		return 0;
3635 
3636 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3637 	if (!attrs)
3638 		goto msg_cancel;
3639 
3640 	if (__tipc_nl_add_sk_info(skb, tsk))
3641 		goto attr_msg_cancel;
3642 
3643 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3644 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3645 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3646 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3647 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3648 					 sock_i_uid(sk))) ||
3649 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3650 			      tipc_diag_gen_cookie(sk),
3651 			      TIPC_NLA_SOCK_PAD))
3652 		goto attr_msg_cancel;
3653 
3654 	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3655 	if (!stat)
3656 		goto attr_msg_cancel;
3657 
3658 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3659 			skb_queue_len(&sk->sk_receive_queue)) ||
3660 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3661 			skb_queue_len(&sk->sk_write_queue)) ||
3662 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3663 			atomic_read(&sk->sk_drops)))
3664 		goto stat_msg_cancel;
3665 
3666 	if (tsk->cong_link_cnt &&
3667 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3668 		goto stat_msg_cancel;
3669 
3670 	if (tsk_conn_cong(tsk) &&
3671 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3672 		goto stat_msg_cancel;
3673 
3674 	nla_nest_end(skb, stat);
3675 
3676 	if (tsk->group)
3677 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3678 			goto stat_msg_cancel;
3679 
3680 	nla_nest_end(skb, attrs);
3681 
3682 	return 0;
3683 
3684 stat_msg_cancel:
3685 	nla_nest_cancel(skb, stat);
3686 attr_msg_cancel:
3687 	nla_nest_cancel(skb, attrs);
3688 msg_cancel:
3689 	return -EMSGSIZE;
3690 }
3691 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3692 
3693 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3694 {
3695 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3696 }
3697 
3698 /* Caller should hold socket lock for the passed tipc socket. */
3699 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3700 				 struct netlink_callback *cb,
3701 				 struct publication *publ)
3702 {
3703 	void *hdr;
3704 	struct nlattr *attrs;
3705 
3706 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3707 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3708 	if (!hdr)
3709 		goto msg_cancel;
3710 
3711 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3712 	if (!attrs)
3713 		goto genlmsg_cancel;
3714 
3715 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3716 		goto attr_msg_cancel;
3717 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3718 		goto attr_msg_cancel;
3719 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3720 		goto attr_msg_cancel;
3721 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3722 		goto attr_msg_cancel;
3723 
3724 	nla_nest_end(skb, attrs);
3725 	genlmsg_end(skb, hdr);
3726 
3727 	return 0;
3728 
3729 attr_msg_cancel:
3730 	nla_nest_cancel(skb, attrs);
3731 genlmsg_cancel:
3732 	genlmsg_cancel(skb, hdr);
3733 msg_cancel:
3734 	return -EMSGSIZE;
3735 }
3736 
3737 /* Caller should hold socket lock for the passed tipc socket. */
3738 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3739 				  struct netlink_callback *cb,
3740 				  struct tipc_sock *tsk, u32 *last_publ)
3741 {
3742 	int err;
3743 	struct publication *p;
3744 
3745 	if (*last_publ) {
3746 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3747 			if (p->key == *last_publ)
3748 				break;
3749 		}
3750 		if (p->key != *last_publ) {
3751 			/* We never set seq or call nl_dump_check_consistent()
3752 			 * this means that setting prev_seq here will cause the
3753 			 * consistence check to fail in the netlink callback
3754 			 * handler. Resulting in the last NLMSG_DONE message
3755 			 * having the NLM_F_DUMP_INTR flag set.
3756 			 */
3757 			cb->prev_seq = 1;
3758 			*last_publ = 0;
3759 			return -EPIPE;
3760 		}
3761 	} else {
3762 		p = list_first_entry(&tsk->publications, struct publication,
3763 				     binding_sock);
3764 	}
3765 
3766 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3767 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3768 		if (err) {
3769 			*last_publ = p->key;
3770 			return err;
3771 		}
3772 	}
3773 	*last_publ = 0;
3774 
3775 	return 0;
3776 }
3777 
3778 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3779 {
3780 	int err;
3781 	u32 tsk_portid = cb->args[0];
3782 	u32 last_publ = cb->args[1];
3783 	u32 done = cb->args[2];
3784 	struct net *net = sock_net(skb->sk);
3785 	struct tipc_sock *tsk;
3786 
3787 	if (!tsk_portid) {
3788 		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3789 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3790 
3791 		if (!attrs[TIPC_NLA_SOCK])
3792 			return -EINVAL;
3793 
3794 		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3795 						  attrs[TIPC_NLA_SOCK],
3796 						  tipc_nl_sock_policy, NULL);
3797 		if (err)
3798 			return err;
3799 
3800 		if (!sock[TIPC_NLA_SOCK_REF])
3801 			return -EINVAL;
3802 
3803 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3804 	}
3805 
3806 	if (done)
3807 		return 0;
3808 
3809 	tsk = tipc_sk_lookup(net, tsk_portid);
3810 	if (!tsk)
3811 		return -EINVAL;
3812 
3813 	lock_sock(&tsk->sk);
3814 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3815 	if (!err)
3816 		done = 1;
3817 	release_sock(&tsk->sk);
3818 	sock_put(&tsk->sk);
3819 
3820 	cb->args[0] = tsk_portid;
3821 	cb->args[1] = last_publ;
3822 	cb->args[2] = done;
3823 
3824 	return skb->len;
3825 }
3826 
3827 /**
3828  * tipc_sk_filtering - check if a socket should be traced
3829  * @sk: the socket to be examined
3830  *
3831  * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3832  * (portid, sock type, name type, name lower, name upper)
3833  *
3834  * Return: true if the socket meets the socket tuple data
3835  * (value 0 = 'any') or when there is no tuple set (all = 0),
3836  * otherwise false
3837  */
3838 bool tipc_sk_filtering(struct sock *sk)
3839 {
3840 	struct tipc_sock *tsk;
3841 	struct publication *p;
3842 	u32 _port, _sktype, _type, _lower, _upper;
3843 	u32 type = 0, lower = 0, upper = 0;
3844 
3845 	if (!sk)
3846 		return true;
3847 
3848 	tsk = tipc_sk(sk);
3849 
3850 	_port = sysctl_tipc_sk_filter[0];
3851 	_sktype = sysctl_tipc_sk_filter[1];
3852 	_type = sysctl_tipc_sk_filter[2];
3853 	_lower = sysctl_tipc_sk_filter[3];
3854 	_upper = sysctl_tipc_sk_filter[4];
3855 
3856 	if (!_port && !_sktype && !_type && !_lower && !_upper)
3857 		return true;
3858 
3859 	if (_port)
3860 		return (_port == tsk->portid);
3861 
3862 	if (_sktype && _sktype != sk->sk_type)
3863 		return false;
3864 
3865 	if (tsk->published) {
3866 		p = list_first_entry_or_null(&tsk->publications,
3867 					     struct publication, binding_sock);
3868 		if (p) {
3869 			type = p->sr.type;
3870 			lower = p->sr.lower;
3871 			upper = p->sr.upper;
3872 		}
3873 	}
3874 
3875 	if (!tipc_sk_type_connectionless(sk)) {
3876 		type = msg_nametype(&tsk->phdr);
3877 		lower = msg_nameinst(&tsk->phdr);
3878 		upper = lower;
3879 	}
3880 
3881 	if ((_type && _type != type) || (_lower && _lower != lower) ||
3882 	    (_upper && _upper != upper))
3883 		return false;
3884 
3885 	return true;
3886 }
3887 
3888 u32 tipc_sock_get_portid(struct sock *sk)
3889 {
3890 	return (sk) ? (tipc_sk(sk))->portid : 0;
3891 }
3892 
3893 /**
3894  * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3895  *			both the rcv and backlog queues are considered
3896  * @sk: tipc sk to be checked
3897  * @skb: tipc msg to be checked
3898  *
3899  * Return: true if the socket rx queue allocation is > 90%, otherwise false
3900  */
3901 
3902 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3903 {
3904 	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3905 	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3906 	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3907 
3908 	return (qsize > lim * 90 / 100);
3909 }
3910 
3911 /**
3912  * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3913  *			only the rcv queue is considered
3914  * @sk: tipc sk to be checked
3915  * @skb: tipc msg to be checked
3916  *
3917  * Return: true if the socket rx queue allocation is > 90%, otherwise false
3918  */
3919 
3920 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3921 {
3922 	unsigned int lim = rcvbuf_limit(sk, skb);
3923 	unsigned int qsize = sk_rmem_alloc_get(sk);
3924 
3925 	return (qsize > lim * 90 / 100);
3926 }
3927 
3928 /**
3929  * tipc_sk_dump - dump TIPC socket
3930  * @sk: tipc sk to be dumped
3931  * @dqueues: bitmask to decide if any socket queue to be dumped?
3932  *           - TIPC_DUMP_NONE: don't dump socket queues
3933  *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3934  *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3935  *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3936  *           - TIPC_DUMP_ALL: dump all the socket queues above
3937  * @buf: returned buffer of dump data in format
3938  */
3939 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3940 {
3941 	int i = 0;
3942 	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3943 	u32 conn_type, conn_instance;
3944 	struct tipc_sock *tsk;
3945 	struct publication *p;
3946 	bool tsk_connected;
3947 
3948 	if (!sk) {
3949 		i += scnprintf(buf, sz, "sk data: (null)\n");
3950 		return i;
3951 	}
3952 
3953 	tsk = tipc_sk(sk);
3954 	tsk_connected = !tipc_sk_type_connectionless(sk);
3955 
3956 	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3957 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3958 	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3959 	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3960 	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3961 	if (tsk_connected) {
3962 		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3963 		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3964 		conn_type = msg_nametype(&tsk->phdr);
3965 		conn_instance = msg_nameinst(&tsk->phdr);
3966 		i += scnprintf(buf + i, sz - i, " %u", conn_type);
3967 		i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3968 	}
3969 	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3970 	if (tsk->published) {
3971 		p = list_first_entry_or_null(&tsk->publications,
3972 					     struct publication, binding_sock);
3973 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3974 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3975 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3976 	}
3977 	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3978 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3979 	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3980 	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3981 	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3982 	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3983 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3984 	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3985 	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3986 	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3987 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3988 	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3989 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3990 	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3991 
3992 	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3993 		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3994 		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3995 	}
3996 
3997 	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3998 		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3999 		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
4000 	}
4001 
4002 	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
4003 		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
4004 		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
4005 		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4006 			i += scnprintf(buf + i, sz - i, "  tail ");
4007 			i += tipc_skb_dump(sk->sk_backlog.tail, false,
4008 					   buf + i);
4009 		}
4010 	}
4011 
4012 	return i;
4013 }
4014