xref: /kernel/linux/linux-5.10/net/tipc/socket.c (revision 8c2ecf20)
1/*
2 * net/tipc/socket.c: TIPC socket API
3 *
4 * Copyright (c) 2001-2007, 2012-2017, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 *    contributors may be used to endorse or promote products derived from
18 *    this software without specific prior written permission.
19 *
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
35 */
36
37#include <linux/rhashtable.h>
38#include <linux/sched/signal.h>
39
40#include "core.h"
41#include "name_table.h"
42#include "node.h"
43#include "link.h"
44#include "name_distr.h"
45#include "socket.h"
46#include "bcast.h"
47#include "netlink.h"
48#include "group.h"
49#include "trace.h"
50
51#define NAGLE_START_INIT	4
52#define NAGLE_START_MAX		1024
53#define CONN_TIMEOUT_DEFAULT    8000    /* default connect timeout = 8s */
54#define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
55#define TIPC_MAX_PORT		0xffffffff
56#define TIPC_MIN_PORT		1
57#define TIPC_ACK_RATE		4       /* ACK at 1/4 of rcv window size */
58
59enum {
60	TIPC_LISTEN = TCP_LISTEN,
61	TIPC_ESTABLISHED = TCP_ESTABLISHED,
62	TIPC_OPEN = TCP_CLOSE,
63	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
64	TIPC_CONNECTING = TCP_SYN_SENT,
65};
66
67struct sockaddr_pair {
68	struct sockaddr_tipc sock;
69	struct sockaddr_tipc member;
70};
71
72/**
73 * struct tipc_sock - TIPC socket structure
74 * @sk: socket - interacts with 'port' and with user via the socket API
75 * @conn_type: TIPC type used when connection was established
76 * @conn_instance: TIPC instance used when connection was established
77 * @published: non-zero if port has one or more associated names
78 * @max_pkt: maximum packet size "hint" used when building messages sent by port
79 * @maxnagle: maximum size of msg which can be subject to nagle
80 * @portid: unique port identity in TIPC socket hash table
81 * @phdr: preformatted message header used when sending messages
82 * #cong_links: list of congested links
83 * @publications: list of publications for port
84 * @blocking_link: address of the congested link we are currently sleeping on
85 * @pub_count: total # of publications port has made during its lifetime
86 * @conn_timeout: the time we can wait for an unresponded setup request
87 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
88 * @cong_link_cnt: number of congested links
89 * @snt_unacked: # messages sent by socket, and not yet acked by peer
90 * @rcv_unacked: # messages read by user, but not yet acked back to peer
91 * @peer: 'connected' peer for dgram/rdm
92 * @node: hash table node
93 * @mc_method: cookie for use between socket and broadcast layer
94 * @rcu: rcu struct for tipc_sock
95 */
96struct tipc_sock {
97	struct sock sk;
98	u32 conn_type;
99	u32 conn_instance;
100	int published;
101	u32 max_pkt;
102	u32 maxnagle;
103	u32 portid;
104	struct tipc_msg phdr;
105	struct list_head cong_links;
106	struct list_head publications;
107	u32 pub_count;
108	atomic_t dupl_rcvcnt;
109	u16 conn_timeout;
110	bool probe_unacked;
111	u16 cong_link_cnt;
112	u16 snt_unacked;
113	u16 snd_win;
114	u16 peer_caps;
115	u16 rcv_unacked;
116	u16 rcv_win;
117	struct sockaddr_tipc peer;
118	struct rhash_head node;
119	struct tipc_mc_method mc_method;
120	struct rcu_head rcu;
121	struct tipc_group *group;
122	u32 oneway;
123	u32 nagle_start;
124	u16 snd_backlog;
125	u16 msg_acc;
126	u16 pkt_cnt;
127	bool expect_ack;
128	bool nodelay;
129	bool group_is_open;
130};
131
132static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
133static void tipc_data_ready(struct sock *sk);
134static void tipc_write_space(struct sock *sk);
135static void tipc_sock_destruct(struct sock *sk);
136static int tipc_release(struct socket *sock);
137static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
138		       bool kern);
139static void tipc_sk_timeout(struct timer_list *t);
140static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
141			   struct tipc_name_seq const *seq);
142static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
143			    struct tipc_name_seq const *seq);
144static int tipc_sk_leave(struct tipc_sock *tsk);
145static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
146static int tipc_sk_insert(struct tipc_sock *tsk);
147static void tipc_sk_remove(struct tipc_sock *tsk);
148static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
149static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
150static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
151static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
152
153static const struct proto_ops packet_ops;
154static const struct proto_ops stream_ops;
155static const struct proto_ops msg_ops;
156static struct proto tipc_proto;
157static const struct rhashtable_params tsk_rht_params;
158
159static u32 tsk_own_node(struct tipc_sock *tsk)
160{
161	return msg_prevnode(&tsk->phdr);
162}
163
164static u32 tsk_peer_node(struct tipc_sock *tsk)
165{
166	return msg_destnode(&tsk->phdr);
167}
168
169static u32 tsk_peer_port(struct tipc_sock *tsk)
170{
171	return msg_destport(&tsk->phdr);
172}
173
174static  bool tsk_unreliable(struct tipc_sock *tsk)
175{
176	return msg_src_droppable(&tsk->phdr) != 0;
177}
178
179static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
180{
181	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
182}
183
184static bool tsk_unreturnable(struct tipc_sock *tsk)
185{
186	return msg_dest_droppable(&tsk->phdr) != 0;
187}
188
189static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
190{
191	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
192}
193
194static int tsk_importance(struct tipc_sock *tsk)
195{
196	return msg_importance(&tsk->phdr);
197}
198
199static struct tipc_sock *tipc_sk(const struct sock *sk)
200{
201	return container_of(sk, struct tipc_sock, sk);
202}
203
204int tsk_set_importance(struct sock *sk, int imp)
205{
206	if (imp > TIPC_CRITICAL_IMPORTANCE)
207		return -EINVAL;
208	msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
209	return 0;
210}
211
212static bool tsk_conn_cong(struct tipc_sock *tsk)
213{
214	return tsk->snt_unacked > tsk->snd_win;
215}
216
217static u16 tsk_blocks(int len)
218{
219	return ((len / FLOWCTL_BLK_SZ) + 1);
220}
221
222/* tsk_blocks(): translate a buffer size in bytes to number of
223 * advertisable blocks, taking into account the ratio truesize(len)/len
224 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
225 */
226static u16 tsk_adv_blocks(int len)
227{
228	return len / FLOWCTL_BLK_SZ / 4;
229}
230
231/* tsk_inc(): increment counter for sent or received data
232 * - If block based flow control is not supported by peer we
233 *   fall back to message based ditto, incrementing the counter
234 */
235static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
236{
237	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
238		return ((msglen / FLOWCTL_BLK_SZ) + 1);
239	return 1;
240}
241
242/* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
243 */
244static void tsk_set_nagle(struct tipc_sock *tsk)
245{
246	struct sock *sk = &tsk->sk;
247
248	tsk->maxnagle = 0;
249	if (sk->sk_type != SOCK_STREAM)
250		return;
251	if (tsk->nodelay)
252		return;
253	if (!(tsk->peer_caps & TIPC_NAGLE))
254		return;
255	/* Limit node local buffer size to avoid receive queue overflow */
256	if (tsk->max_pkt == MAX_MSG_SIZE)
257		tsk->maxnagle = 1500;
258	else
259		tsk->maxnagle = tsk->max_pkt;
260}
261
262/**
263 * tsk_advance_rx_queue - discard first buffer in socket receive queue
264 *
265 * Caller must hold socket lock
266 */
267static void tsk_advance_rx_queue(struct sock *sk)
268{
269	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
270	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
271}
272
273/* tipc_sk_respond() : send response message back to sender
274 */
275static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
276{
277	u32 selector;
278	u32 dnode;
279	u32 onode = tipc_own_addr(sock_net(sk));
280
281	if (!tipc_msg_reverse(onode, &skb, err))
282		return;
283
284	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
285	dnode = msg_destnode(buf_msg(skb));
286	selector = msg_origport(buf_msg(skb));
287	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
288}
289
290/**
291 * tsk_rej_rx_queue - reject all buffers in socket receive queue
292 *
293 * Caller must hold socket lock
294 */
295static void tsk_rej_rx_queue(struct sock *sk, int error)
296{
297	struct sk_buff *skb;
298
299	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
300		tipc_sk_respond(sk, skb, error);
301}
302
303static bool tipc_sk_connected(const struct sock *sk)
304{
305	return READ_ONCE(sk->sk_state) == TIPC_ESTABLISHED;
306}
307
308/* tipc_sk_type_connectionless - check if the socket is datagram socket
309 * @sk: socket
310 *
311 * Returns true if connection less, false otherwise
312 */
313static bool tipc_sk_type_connectionless(struct sock *sk)
314{
315	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
316}
317
318/* tsk_peer_msg - verify if message was sent by connected port's peer
319 *
320 * Handles cases where the node's network address has changed from
321 * the default of <0.0.0> to its configured setting.
322 */
323static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
324{
325	struct sock *sk = &tsk->sk;
326	u32 self = tipc_own_addr(sock_net(sk));
327	u32 peer_port = tsk_peer_port(tsk);
328	u32 orig_node, peer_node;
329
330	if (unlikely(!tipc_sk_connected(sk)))
331		return false;
332
333	if (unlikely(msg_origport(msg) != peer_port))
334		return false;
335
336	orig_node = msg_orignode(msg);
337	peer_node = tsk_peer_node(tsk);
338
339	if (likely(orig_node == peer_node))
340		return true;
341
342	if (!orig_node && peer_node == self)
343		return true;
344
345	if (!peer_node && orig_node == self)
346		return true;
347
348	return false;
349}
350
351/* tipc_set_sk_state - set the sk_state of the socket
352 * @sk: socket
353 *
354 * Caller must hold socket lock
355 *
356 * Returns 0 on success, errno otherwise
357 */
358static int tipc_set_sk_state(struct sock *sk, int state)
359{
360	int oldsk_state = sk->sk_state;
361	int res = -EINVAL;
362
363	switch (state) {
364	case TIPC_OPEN:
365		res = 0;
366		break;
367	case TIPC_LISTEN:
368	case TIPC_CONNECTING:
369		if (oldsk_state == TIPC_OPEN)
370			res = 0;
371		break;
372	case TIPC_ESTABLISHED:
373		if (oldsk_state == TIPC_CONNECTING ||
374		    oldsk_state == TIPC_OPEN)
375			res = 0;
376		break;
377	case TIPC_DISCONNECTING:
378		if (oldsk_state == TIPC_CONNECTING ||
379		    oldsk_state == TIPC_ESTABLISHED)
380			res = 0;
381		break;
382	}
383
384	if (!res)
385		sk->sk_state = state;
386
387	return res;
388}
389
390static int tipc_sk_sock_err(struct socket *sock, long *timeout)
391{
392	struct sock *sk = sock->sk;
393	int err = sock_error(sk);
394	int typ = sock->type;
395
396	if (err)
397		return err;
398	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
399		if (sk->sk_state == TIPC_DISCONNECTING)
400			return -EPIPE;
401		else if (!tipc_sk_connected(sk))
402			return -ENOTCONN;
403	}
404	if (!*timeout)
405		return -EAGAIN;
406	if (signal_pending(current))
407		return sock_intr_errno(*timeout);
408
409	return 0;
410}
411
412#define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
413({                                                                             \
414	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
415	struct sock *sk_;						       \
416	int rc_;							       \
417									       \
418	while ((rc_ = !(condition_))) {					       \
419		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
420		smp_rmb();                                                     \
421		sk_ = (sock_)->sk;					       \
422		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
423		if (rc_)						       \
424			break;						       \
425		add_wait_queue(sk_sleep(sk_), &wait_);                         \
426		release_sock(sk_);					       \
427		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
428		sched_annotate_sleep();				               \
429		lock_sock(sk_);						       \
430		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
431	}								       \
432	rc_;								       \
433})
434
435/**
436 * tipc_sk_create - create a TIPC socket
437 * @net: network namespace (must be default network)
438 * @sock: pre-allocated socket structure
439 * @protocol: protocol indicator (must be 0)
440 * @kern: caused by kernel or by userspace?
441 *
442 * This routine creates additional data structures used by the TIPC socket,
443 * initializes them, and links them together.
444 *
445 * Returns 0 on success, errno otherwise
446 */
447static int tipc_sk_create(struct net *net, struct socket *sock,
448			  int protocol, int kern)
449{
450	const struct proto_ops *ops;
451	struct sock *sk;
452	struct tipc_sock *tsk;
453	struct tipc_msg *msg;
454
455	/* Validate arguments */
456	if (unlikely(protocol != 0))
457		return -EPROTONOSUPPORT;
458
459	switch (sock->type) {
460	case SOCK_STREAM:
461		ops = &stream_ops;
462		break;
463	case SOCK_SEQPACKET:
464		ops = &packet_ops;
465		break;
466	case SOCK_DGRAM:
467	case SOCK_RDM:
468		ops = &msg_ops;
469		break;
470	default:
471		return -EPROTOTYPE;
472	}
473
474	/* Allocate socket's protocol area */
475	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
476	if (sk == NULL)
477		return -ENOMEM;
478
479	tsk = tipc_sk(sk);
480	tsk->max_pkt = MAX_PKT_DEFAULT;
481	tsk->maxnagle = 0;
482	tsk->nagle_start = NAGLE_START_INIT;
483	INIT_LIST_HEAD(&tsk->publications);
484	INIT_LIST_HEAD(&tsk->cong_links);
485	msg = &tsk->phdr;
486
487	/* Finish initializing socket data structures */
488	sock->ops = ops;
489	sock_init_data(sock, sk);
490	tipc_set_sk_state(sk, TIPC_OPEN);
491	if (tipc_sk_insert(tsk)) {
492		sk_free(sk);
493		pr_warn("Socket create failed; port number exhausted\n");
494		return -EINVAL;
495	}
496
497	/* Ensure tsk is visible before we read own_addr. */
498	smp_mb();
499
500	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
501		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
502
503	msg_set_origport(msg, tsk->portid);
504	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
505	sk->sk_shutdown = 0;
506	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
507	sk->sk_rcvbuf = READ_ONCE(sysctl_tipc_rmem[1]);
508	sk->sk_data_ready = tipc_data_ready;
509	sk->sk_write_space = tipc_write_space;
510	sk->sk_destruct = tipc_sock_destruct;
511	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
512	tsk->group_is_open = true;
513	atomic_set(&tsk->dupl_rcvcnt, 0);
514
515	/* Start out with safe limits until we receive an advertised window */
516	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
517	tsk->rcv_win = tsk->snd_win;
518
519	if (tipc_sk_type_connectionless(sk)) {
520		tsk_set_unreturnable(tsk, true);
521		if (sock->type == SOCK_DGRAM)
522			tsk_set_unreliable(tsk, true);
523	}
524	__skb_queue_head_init(&tsk->mc_method.deferredq);
525	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
526	return 0;
527}
528
529static void tipc_sk_callback(struct rcu_head *head)
530{
531	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
532
533	sock_put(&tsk->sk);
534}
535
536/* Caller should hold socket lock for the socket. */
537static void __tipc_shutdown(struct socket *sock, int error)
538{
539	struct sock *sk = sock->sk;
540	struct tipc_sock *tsk = tipc_sk(sk);
541	struct net *net = sock_net(sk);
542	long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
543	u32 dnode = tsk_peer_node(tsk);
544	struct sk_buff *skb;
545
546	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
547	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
548					    !tsk_conn_cong(tsk)));
549
550	/* Push out delayed messages if in Nagle mode */
551	tipc_sk_push_backlog(tsk, false);
552	/* Remove pending SYN */
553	__skb_queue_purge(&sk->sk_write_queue);
554
555	/* Remove partially received buffer if any */
556	skb = skb_peek(&sk->sk_receive_queue);
557	if (skb && TIPC_SKB_CB(skb)->bytes_read) {
558		__skb_unlink(skb, &sk->sk_receive_queue);
559		kfree_skb(skb);
560	}
561
562	/* Reject all unreceived messages if connectionless */
563	if (tipc_sk_type_connectionless(sk)) {
564		tsk_rej_rx_queue(sk, error);
565		return;
566	}
567
568	switch (sk->sk_state) {
569	case TIPC_CONNECTING:
570	case TIPC_ESTABLISHED:
571		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
572		tipc_node_remove_conn(net, dnode, tsk->portid);
573		/* Send a FIN+/- to its peer */
574		skb = __skb_dequeue(&sk->sk_receive_queue);
575		if (skb) {
576			__skb_queue_purge(&sk->sk_receive_queue);
577			tipc_sk_respond(sk, skb, error);
578			break;
579		}
580		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
581				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
582				      tsk_own_node(tsk), tsk_peer_port(tsk),
583				      tsk->portid, error);
584		if (skb)
585			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
586		break;
587	case TIPC_LISTEN:
588		/* Reject all SYN messages */
589		tsk_rej_rx_queue(sk, error);
590		break;
591	default:
592		__skb_queue_purge(&sk->sk_receive_queue);
593		break;
594	}
595}
596
597/**
598 * tipc_release - destroy a TIPC socket
599 * @sock: socket to destroy
600 *
601 * This routine cleans up any messages that are still queued on the socket.
602 * For DGRAM and RDM socket types, all queued messages are rejected.
603 * For SEQPACKET and STREAM socket types, the first message is rejected
604 * and any others are discarded.  (If the first message on a STREAM socket
605 * is partially-read, it is discarded and the next one is rejected instead.)
606 *
607 * NOTE: Rejected messages are not necessarily returned to the sender!  They
608 * are returned or discarded according to the "destination droppable" setting
609 * specified for the message by the sender.
610 *
611 * Returns 0 on success, errno otherwise
612 */
613static int tipc_release(struct socket *sock)
614{
615	struct sock *sk = sock->sk;
616	struct tipc_sock *tsk;
617
618	/*
619	 * Exit if socket isn't fully initialized (occurs when a failed accept()
620	 * releases a pre-allocated child socket that was never used)
621	 */
622	if (sk == NULL)
623		return 0;
624
625	tsk = tipc_sk(sk);
626	lock_sock(sk);
627
628	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
629	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
630	sk->sk_shutdown = SHUTDOWN_MASK;
631	tipc_sk_leave(tsk);
632	tipc_sk_withdraw(tsk, 0, NULL);
633	__skb_queue_purge(&tsk->mc_method.deferredq);
634	sk_stop_timer(sk, &sk->sk_timer);
635	tipc_sk_remove(tsk);
636
637	sock_orphan(sk);
638	/* Reject any messages that accumulated in backlog queue */
639	release_sock(sk);
640	tipc_dest_list_purge(&tsk->cong_links);
641	tsk->cong_link_cnt = 0;
642	call_rcu(&tsk->rcu, tipc_sk_callback);
643	sock->sk = NULL;
644
645	return 0;
646}
647
648/**
649 * tipc_bind - associate or disassocate TIPC name(s) with a socket
650 * @sock: socket structure
651 * @uaddr: socket address describing name(s) and desired operation
652 * @uaddr_len: size of socket address data structure
653 *
654 * Name and name sequence binding is indicated using a positive scope value;
655 * a negative scope value unbinds the specified name.  Specifying no name
656 * (i.e. a socket address length of 0) unbinds all names from the socket.
657 *
658 * Returns 0 on success, errno otherwise
659 *
660 * NOTE: This routine doesn't need to take the socket lock since it doesn't
661 *       access any non-constant socket information.
662 */
663static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
664		     int uaddr_len)
665{
666	struct sock *sk = sock->sk;
667	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
668	struct tipc_sock *tsk = tipc_sk(sk);
669	int res = -EINVAL;
670
671	lock_sock(sk);
672	if (unlikely(!uaddr_len)) {
673		res = tipc_sk_withdraw(tsk, 0, NULL);
674		goto exit;
675	}
676	if (tsk->group) {
677		res = -EACCES;
678		goto exit;
679	}
680	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
681		res = -EINVAL;
682		goto exit;
683	}
684	if (addr->family != AF_TIPC) {
685		res = -EAFNOSUPPORT;
686		goto exit;
687	}
688
689	if (addr->addrtype == TIPC_ADDR_NAME)
690		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
691	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
692		res = -EAFNOSUPPORT;
693		goto exit;
694	}
695
696	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
697	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
698	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
699		res = -EACCES;
700		goto exit;
701	}
702
703	res = (addr->scope >= 0) ?
704		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
705		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
706exit:
707	release_sock(sk);
708	return res;
709}
710
711/**
712 * tipc_getname - get port ID of socket or peer socket
713 * @sock: socket structure
714 * @uaddr: area for returned socket address
715 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
716 *
717 * Returns 0 on success, errno otherwise
718 *
719 * NOTE: This routine doesn't need to take the socket lock since it only
720 *       accesses socket information that is unchanging (or which changes in
721 *       a completely predictable manner).
722 */
723static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
724			int peer)
725{
726	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
727	struct sock *sk = sock->sk;
728	struct tipc_sock *tsk = tipc_sk(sk);
729
730	memset(addr, 0, sizeof(*addr));
731	if (peer) {
732		if ((!tipc_sk_connected(sk)) &&
733		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
734			return -ENOTCONN;
735		addr->addr.id.ref = tsk_peer_port(tsk);
736		addr->addr.id.node = tsk_peer_node(tsk);
737	} else {
738		addr->addr.id.ref = tsk->portid;
739		addr->addr.id.node = tipc_own_addr(sock_net(sk));
740	}
741
742	addr->addrtype = TIPC_ADDR_ID;
743	addr->family = AF_TIPC;
744	addr->scope = 0;
745	addr->addr.name.domain = 0;
746
747	return sizeof(*addr);
748}
749
750/**
751 * tipc_poll - read and possibly block on pollmask
752 * @file: file structure associated with the socket
753 * @sock: socket for which to calculate the poll bits
754 * @wait: ???
755 *
756 * Returns pollmask value
757 *
758 * COMMENTARY:
759 * It appears that the usual socket locking mechanisms are not useful here
760 * since the pollmask info is potentially out-of-date the moment this routine
761 * exits.  TCP and other protocols seem to rely on higher level poll routines
762 * to handle any preventable race conditions, so TIPC will do the same ...
763 *
764 * IMPORTANT: The fact that a read or write operation is indicated does NOT
765 * imply that the operation will succeed, merely that it should be performed
766 * and will not block.
767 */
768static __poll_t tipc_poll(struct file *file, struct socket *sock,
769			      poll_table *wait)
770{
771	struct sock *sk = sock->sk;
772	struct tipc_sock *tsk = tipc_sk(sk);
773	__poll_t revents = 0;
774
775	sock_poll_wait(file, sock, wait);
776	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
777
778	if (sk->sk_shutdown & RCV_SHUTDOWN)
779		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
780	if (sk->sk_shutdown == SHUTDOWN_MASK)
781		revents |= EPOLLHUP;
782
783	switch (sk->sk_state) {
784	case TIPC_ESTABLISHED:
785		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
786			revents |= EPOLLOUT;
787		fallthrough;
788	case TIPC_LISTEN:
789	case TIPC_CONNECTING:
790		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
791			revents |= EPOLLIN | EPOLLRDNORM;
792		break;
793	case TIPC_OPEN:
794		if (tsk->group_is_open && !tsk->cong_link_cnt)
795			revents |= EPOLLOUT;
796		if (!tipc_sk_type_connectionless(sk))
797			break;
798		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
799			break;
800		revents |= EPOLLIN | EPOLLRDNORM;
801		break;
802	case TIPC_DISCONNECTING:
803		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
804		break;
805	}
806	return revents;
807}
808
809/**
810 * tipc_sendmcast - send multicast message
811 * @sock: socket structure
812 * @seq: destination address
813 * @msg: message to send
814 * @dlen: length of data to send
815 * @timeout: timeout to wait for wakeup
816 *
817 * Called from function tipc_sendmsg(), which has done all sanity checks
818 * Returns the number of bytes sent on success, or errno
819 */
820static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
821			  struct msghdr *msg, size_t dlen, long timeout)
822{
823	struct sock *sk = sock->sk;
824	struct tipc_sock *tsk = tipc_sk(sk);
825	struct tipc_msg *hdr = &tsk->phdr;
826	struct net *net = sock_net(sk);
827	int mtu = tipc_bcast_get_mtu(net);
828	struct tipc_mc_method *method = &tsk->mc_method;
829	struct sk_buff_head pkts;
830	struct tipc_nlist dsts;
831	int rc;
832
833	if (tsk->group)
834		return -EACCES;
835
836	/* Block or return if any destination link is congested */
837	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
838	if (unlikely(rc))
839		return rc;
840
841	/* Lookup destination nodes */
842	tipc_nlist_init(&dsts, tipc_own_addr(net));
843	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
844				      seq->upper, &dsts);
845	if (!dsts.local && !dsts.remote)
846		return -EHOSTUNREACH;
847
848	/* Build message header */
849	msg_set_type(hdr, TIPC_MCAST_MSG);
850	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
851	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
852	msg_set_destport(hdr, 0);
853	msg_set_destnode(hdr, 0);
854	msg_set_nametype(hdr, seq->type);
855	msg_set_namelower(hdr, seq->lower);
856	msg_set_nameupper(hdr, seq->upper);
857
858	/* Build message as chain of buffers */
859	__skb_queue_head_init(&pkts);
860	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
861
862	/* Send message if build was successful */
863	if (unlikely(rc == dlen)) {
864		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
865					TIPC_DUMP_SK_SNDQ, " ");
866		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
867				     &tsk->cong_link_cnt);
868	}
869
870	tipc_nlist_purge(&dsts);
871
872	return rc ? rc : dlen;
873}
874
875/**
876 * tipc_send_group_msg - send a message to a member in the group
877 * @net: network namespace
878 * @m: message to send
879 * @mb: group member
880 * @dnode: destination node
881 * @dport: destination port
882 * @dlen: total length of message data
883 */
884static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
885			       struct msghdr *m, struct tipc_member *mb,
886			       u32 dnode, u32 dport, int dlen)
887{
888	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
889	struct tipc_mc_method *method = &tsk->mc_method;
890	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
891	struct tipc_msg *hdr = &tsk->phdr;
892	struct sk_buff_head pkts;
893	int mtu, rc;
894
895	/* Complete message header */
896	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
897	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
898	msg_set_destport(hdr, dport);
899	msg_set_destnode(hdr, dnode);
900	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
901
902	/* Build message as chain of buffers */
903	__skb_queue_head_init(&pkts);
904	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
905	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
906	if (unlikely(rc != dlen))
907		return rc;
908
909	/* Send message */
910	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
911	if (unlikely(rc == -ELINKCONG)) {
912		tipc_dest_push(&tsk->cong_links, dnode, 0);
913		tsk->cong_link_cnt++;
914	}
915
916	/* Update send window */
917	tipc_group_update_member(mb, blks);
918
919	/* A broadcast sent within next EXPIRE period must follow same path */
920	method->rcast = true;
921	method->mandatory = true;
922	return dlen;
923}
924
925/**
926 * tipc_send_group_unicast - send message to a member in the group
927 * @sock: socket structure
928 * @m: message to send
929 * @dlen: total length of message data
930 * @timeout: timeout to wait for wakeup
931 *
932 * Called from function tipc_sendmsg(), which has done all sanity checks
933 * Returns the number of bytes sent on success, or errno
934 */
935static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
936				   int dlen, long timeout)
937{
938	struct sock *sk = sock->sk;
939	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
940	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
941	struct tipc_sock *tsk = tipc_sk(sk);
942	struct net *net = sock_net(sk);
943	struct tipc_member *mb = NULL;
944	u32 node, port;
945	int rc;
946
947	node = dest->addr.id.node;
948	port = dest->addr.id.ref;
949	if (!port && !node)
950		return -EHOSTUNREACH;
951
952	/* Block or return if destination link or member is congested */
953	rc = tipc_wait_for_cond(sock, &timeout,
954				!tipc_dest_find(&tsk->cong_links, node, 0) &&
955				tsk->group &&
956				!tipc_group_cong(tsk->group, node, port, blks,
957						 &mb));
958	if (unlikely(rc))
959		return rc;
960
961	if (unlikely(!mb))
962		return -EHOSTUNREACH;
963
964	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
965
966	return rc ? rc : dlen;
967}
968
969/**
970 * tipc_send_group_anycast - send message to any member with given identity
971 * @sock: socket structure
972 * @m: message to send
973 * @dlen: total length of message data
974 * @timeout: timeout to wait for wakeup
975 *
976 * Called from function tipc_sendmsg(), which has done all sanity checks
977 * Returns the number of bytes sent on success, or errno
978 */
979static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
980				   int dlen, long timeout)
981{
982	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
983	struct sock *sk = sock->sk;
984	struct tipc_sock *tsk = tipc_sk(sk);
985	struct list_head *cong_links = &tsk->cong_links;
986	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
987	struct tipc_msg *hdr = &tsk->phdr;
988	struct tipc_member *first = NULL;
989	struct tipc_member *mbr = NULL;
990	struct net *net = sock_net(sk);
991	u32 node, port, exclude;
992	struct list_head dsts;
993	u32 type, inst, scope;
994	int lookups = 0;
995	int dstcnt, rc;
996	bool cong;
997
998	INIT_LIST_HEAD(&dsts);
999
1000	type = msg_nametype(hdr);
1001	inst = dest->addr.name.name.instance;
1002	scope = msg_lookup_scope(hdr);
1003
1004	while (++lookups < 4) {
1005		exclude = tipc_group_exclude(tsk->group);
1006
1007		first = NULL;
1008
1009		/* Look for a non-congested destination member, if any */
1010		while (1) {
1011			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1012						 &dstcnt, exclude, false))
1013				return -EHOSTUNREACH;
1014			tipc_dest_pop(&dsts, &node, &port);
1015			cong = tipc_group_cong(tsk->group, node, port, blks,
1016					       &mbr);
1017			if (!cong)
1018				break;
1019			if (mbr == first)
1020				break;
1021			if (!first)
1022				first = mbr;
1023		}
1024
1025		/* Start over if destination was not in member list */
1026		if (unlikely(!mbr))
1027			continue;
1028
1029		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1030			break;
1031
1032		/* Block or return if destination link or member is congested */
1033		rc = tipc_wait_for_cond(sock, &timeout,
1034					!tipc_dest_find(cong_links, node, 0) &&
1035					tsk->group &&
1036					!tipc_group_cong(tsk->group, node, port,
1037							 blks, &mbr));
1038		if (unlikely(rc))
1039			return rc;
1040
1041		/* Send, unless destination disappeared while waiting */
1042		if (likely(mbr))
1043			break;
1044	}
1045
1046	if (unlikely(lookups >= 4))
1047		return -EHOSTUNREACH;
1048
1049	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1050
1051	return rc ? rc : dlen;
1052}
1053
1054/**
1055 * tipc_send_group_bcast - send message to all members in communication group
1056 * @sock: socket structure
1057 * @m: message to send
1058 * @dlen: total length of message data
1059 * @timeout: timeout to wait for wakeup
1060 *
1061 * Called from function tipc_sendmsg(), which has done all sanity checks
1062 * Returns the number of bytes sent on success, or errno
1063 */
1064static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1065				 int dlen, long timeout)
1066{
1067	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1068	struct sock *sk = sock->sk;
1069	struct net *net = sock_net(sk);
1070	struct tipc_sock *tsk = tipc_sk(sk);
1071	struct tipc_nlist *dsts;
1072	struct tipc_mc_method *method = &tsk->mc_method;
1073	bool ack = method->mandatory && method->rcast;
1074	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1075	struct tipc_msg *hdr = &tsk->phdr;
1076	int mtu = tipc_bcast_get_mtu(net);
1077	struct sk_buff_head pkts;
1078	int rc = -EHOSTUNREACH;
1079
1080	/* Block or return if any destination link or member is congested */
1081	rc = tipc_wait_for_cond(sock, &timeout,
1082				!tsk->cong_link_cnt && tsk->group &&
1083				!tipc_group_bc_cong(tsk->group, blks));
1084	if (unlikely(rc))
1085		return rc;
1086
1087	dsts = tipc_group_dests(tsk->group);
1088	if (!dsts->local && !dsts->remote)
1089		return -EHOSTUNREACH;
1090
1091	/* Complete message header */
1092	if (dest) {
1093		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1094		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1095	} else {
1096		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1097		msg_set_nameinst(hdr, 0);
1098	}
1099	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1100	msg_set_destport(hdr, 0);
1101	msg_set_destnode(hdr, 0);
1102	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1103
1104	/* Avoid getting stuck with repeated forced replicasts */
1105	msg_set_grp_bc_ack_req(hdr, ack);
1106
1107	/* Build message as chain of buffers */
1108	__skb_queue_head_init(&pkts);
1109	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1110	if (unlikely(rc != dlen))
1111		return rc;
1112
1113	/* Send message */
1114	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1115	if (unlikely(rc))
1116		return rc;
1117
1118	/* Update broadcast sequence number and send windows */
1119	tipc_group_update_bc_members(tsk->group, blks, ack);
1120
1121	/* Broadcast link is now free to choose method for next broadcast */
1122	method->mandatory = false;
1123	method->expires = jiffies;
1124
1125	return dlen;
1126}
1127
1128/**
1129 * tipc_send_group_mcast - send message to all members with given identity
1130 * @sock: socket structure
1131 * @m: message to send
1132 * @dlen: total length of message data
1133 * @timeout: timeout to wait for wakeup
1134 *
1135 * Called from function tipc_sendmsg(), which has done all sanity checks
1136 * Returns the number of bytes sent on success, or errno
1137 */
1138static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1139				 int dlen, long timeout)
1140{
1141	struct sock *sk = sock->sk;
1142	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1143	struct tipc_sock *tsk = tipc_sk(sk);
1144	struct tipc_group *grp = tsk->group;
1145	struct tipc_msg *hdr = &tsk->phdr;
1146	struct net *net = sock_net(sk);
1147	u32 type, inst, scope, exclude;
1148	struct list_head dsts;
1149	u32 dstcnt;
1150
1151	INIT_LIST_HEAD(&dsts);
1152
1153	type = msg_nametype(hdr);
1154	inst = dest->addr.name.name.instance;
1155	scope = msg_lookup_scope(hdr);
1156	exclude = tipc_group_exclude(grp);
1157
1158	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1159				 &dstcnt, exclude, true))
1160		return -EHOSTUNREACH;
1161
1162	if (dstcnt == 1) {
1163		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1164		return tipc_send_group_unicast(sock, m, dlen, timeout);
1165	}
1166
1167	tipc_dest_list_purge(&dsts);
1168	return tipc_send_group_bcast(sock, m, dlen, timeout);
1169}
1170
1171/**
1172 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1173 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1174 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1175 *
1176 * Multi-threaded: parallel calls with reference to same queues may occur
1177 */
1178void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1179		       struct sk_buff_head *inputq)
1180{
1181	u32 self = tipc_own_addr(net);
1182	u32 type, lower, upper, scope;
1183	struct sk_buff *skb, *_skb;
1184	u32 portid, onode;
1185	struct sk_buff_head tmpq;
1186	struct list_head dports;
1187	struct tipc_msg *hdr;
1188	int user, mtyp, hlen;
1189	bool exact;
1190
1191	__skb_queue_head_init(&tmpq);
1192	INIT_LIST_HEAD(&dports);
1193
1194	skb = tipc_skb_peek(arrvq, &inputq->lock);
1195	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1196		hdr = buf_msg(skb);
1197		user = msg_user(hdr);
1198		mtyp = msg_type(hdr);
1199		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1200		onode = msg_orignode(hdr);
1201		type = msg_nametype(hdr);
1202
1203		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1204			spin_lock_bh(&inputq->lock);
1205			if (skb_peek(arrvq) == skb) {
1206				__skb_dequeue(arrvq);
1207				__skb_queue_tail(inputq, skb);
1208			}
1209			kfree_skb(skb);
1210			spin_unlock_bh(&inputq->lock);
1211			continue;
1212		}
1213
1214		/* Group messages require exact scope match */
1215		if (msg_in_group(hdr)) {
1216			lower = 0;
1217			upper = ~0;
1218			scope = msg_lookup_scope(hdr);
1219			exact = true;
1220		} else {
1221			/* TIPC_NODE_SCOPE means "any scope" in this context */
1222			if (onode == self)
1223				scope = TIPC_NODE_SCOPE;
1224			else
1225				scope = TIPC_CLUSTER_SCOPE;
1226			exact = false;
1227			lower = msg_namelower(hdr);
1228			upper = msg_nameupper(hdr);
1229		}
1230
1231		/* Create destination port list: */
1232		tipc_nametbl_mc_lookup(net, type, lower, upper,
1233				       scope, exact, &dports);
1234
1235		/* Clone message per destination */
1236		while (tipc_dest_pop(&dports, NULL, &portid)) {
1237			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1238			if (_skb) {
1239				msg_set_destport(buf_msg(_skb), portid);
1240				__skb_queue_tail(&tmpq, _skb);
1241				continue;
1242			}
1243			pr_warn("Failed to clone mcast rcv buffer\n");
1244		}
1245		/* Append to inputq if not already done by other thread */
1246		spin_lock_bh(&inputq->lock);
1247		if (skb_peek(arrvq) == skb) {
1248			skb_queue_splice_tail_init(&tmpq, inputq);
1249			/* Decrease the skb's refcnt as increasing in the
1250			 * function tipc_skb_peek
1251			 */
1252			kfree_skb(__skb_dequeue(arrvq));
1253		}
1254		spin_unlock_bh(&inputq->lock);
1255		__skb_queue_purge(&tmpq);
1256		kfree_skb(skb);
1257	}
1258	tipc_sk_rcv(net, inputq);
1259}
1260
1261/* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1262 *                         when socket is in Nagle mode
1263 */
1264static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1265{
1266	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1267	struct sk_buff *skb = skb_peek_tail(txq);
1268	struct net *net = sock_net(&tsk->sk);
1269	u32 dnode = tsk_peer_node(tsk);
1270	int rc;
1271
1272	if (nagle_ack) {
1273		tsk->pkt_cnt += skb_queue_len(txq);
1274		if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1275			tsk->oneway = 0;
1276			if (tsk->nagle_start < NAGLE_START_MAX)
1277				tsk->nagle_start *= 2;
1278			tsk->expect_ack = false;
1279			pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1280				 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1281				 tsk->nagle_start);
1282		} else {
1283			tsk->nagle_start = NAGLE_START_INIT;
1284			if (skb) {
1285				msg_set_ack_required(buf_msg(skb));
1286				tsk->expect_ack = true;
1287			} else {
1288				tsk->expect_ack = false;
1289			}
1290		}
1291		tsk->msg_acc = 0;
1292		tsk->pkt_cnt = 0;
1293	}
1294
1295	if (!skb || tsk->cong_link_cnt)
1296		return;
1297
1298	/* Do not send SYN again after congestion */
1299	if (msg_is_syn(buf_msg(skb)))
1300		return;
1301
1302	if (tsk->msg_acc)
1303		tsk->pkt_cnt += skb_queue_len(txq);
1304	tsk->snt_unacked += tsk->snd_backlog;
1305	tsk->snd_backlog = 0;
1306	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1307	if (rc == -ELINKCONG)
1308		tsk->cong_link_cnt = 1;
1309}
1310
1311/**
1312 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1313 * @tsk: receiving socket
1314 * @skb: pointer to message buffer.
1315 */
1316static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1317				   struct sk_buff_head *inputq,
1318				   struct sk_buff_head *xmitq)
1319{
1320	struct tipc_msg *hdr = buf_msg(skb);
1321	u32 onode = tsk_own_node(tsk);
1322	struct sock *sk = &tsk->sk;
1323	int mtyp = msg_type(hdr);
1324	bool was_cong;
1325
1326	/* Ignore if connection cannot be validated: */
1327	if (!tsk_peer_msg(tsk, hdr)) {
1328		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1329		goto exit;
1330	}
1331
1332	if (unlikely(msg_errcode(hdr))) {
1333		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1334		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1335				      tsk_peer_port(tsk));
1336		sk->sk_state_change(sk);
1337
1338		/* State change is ignored if socket already awake,
1339		 * - convert msg to abort msg and add to inqueue
1340		 */
1341		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1342		msg_set_type(hdr, TIPC_CONN_MSG);
1343		msg_set_size(hdr, BASIC_H_SIZE);
1344		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1345		__skb_queue_tail(inputq, skb);
1346		return;
1347	}
1348
1349	tsk->probe_unacked = false;
1350
1351	if (mtyp == CONN_PROBE) {
1352		msg_set_type(hdr, CONN_PROBE_REPLY);
1353		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1354			__skb_queue_tail(xmitq, skb);
1355		return;
1356	} else if (mtyp == CONN_ACK) {
1357		was_cong = tsk_conn_cong(tsk);
1358		tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1359		tsk->snt_unacked -= msg_conn_ack(hdr);
1360		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1361			tsk->snd_win = msg_adv_win(hdr);
1362		if (was_cong && !tsk_conn_cong(tsk))
1363			sk->sk_write_space(sk);
1364	} else if (mtyp != CONN_PROBE_REPLY) {
1365		pr_warn("Received unknown CONN_PROTO msg\n");
1366	}
1367exit:
1368	kfree_skb(skb);
1369}
1370
1371/**
1372 * tipc_sendmsg - send message in connectionless manner
1373 * @sock: socket structure
1374 * @m: message to send
1375 * @dsz: amount of user data to be sent
1376 *
1377 * Message must have an destination specified explicitly.
1378 * Used for SOCK_RDM and SOCK_DGRAM messages,
1379 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1380 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1381 *
1382 * Returns the number of bytes sent on success, or errno otherwise
1383 */
1384static int tipc_sendmsg(struct socket *sock,
1385			struct msghdr *m, size_t dsz)
1386{
1387	struct sock *sk = sock->sk;
1388	int ret;
1389
1390	lock_sock(sk);
1391	ret = __tipc_sendmsg(sock, m, dsz);
1392	release_sock(sk);
1393
1394	return ret;
1395}
1396
1397static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1398{
1399	struct sock *sk = sock->sk;
1400	struct net *net = sock_net(sk);
1401	struct tipc_sock *tsk = tipc_sk(sk);
1402	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1403	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1404	struct list_head *clinks = &tsk->cong_links;
1405	bool syn = !tipc_sk_type_connectionless(sk);
1406	struct tipc_group *grp = tsk->group;
1407	struct tipc_msg *hdr = &tsk->phdr;
1408	struct tipc_name_seq *seq;
1409	struct sk_buff_head pkts;
1410	u32 dport = 0, dnode = 0;
1411	u32 type = 0, inst = 0;
1412	int mtu, rc;
1413
1414	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1415		return -EMSGSIZE;
1416
1417	if (likely(dest)) {
1418		if (unlikely(m->msg_namelen < sizeof(*dest)))
1419			return -EINVAL;
1420		if (unlikely(dest->family != AF_TIPC))
1421			return -EINVAL;
1422	}
1423
1424	if (grp) {
1425		if (!dest)
1426			return tipc_send_group_bcast(sock, m, dlen, timeout);
1427		if (dest->addrtype == TIPC_ADDR_NAME)
1428			return tipc_send_group_anycast(sock, m, dlen, timeout);
1429		if (dest->addrtype == TIPC_ADDR_ID)
1430			return tipc_send_group_unicast(sock, m, dlen, timeout);
1431		if (dest->addrtype == TIPC_ADDR_MCAST)
1432			return tipc_send_group_mcast(sock, m, dlen, timeout);
1433		return -EINVAL;
1434	}
1435
1436	if (unlikely(!dest)) {
1437		dest = &tsk->peer;
1438		if (!syn && dest->family != AF_TIPC)
1439			return -EDESTADDRREQ;
1440	}
1441
1442	if (unlikely(syn)) {
1443		if (sk->sk_state == TIPC_LISTEN)
1444			return -EPIPE;
1445		if (sk->sk_state != TIPC_OPEN)
1446			return -EISCONN;
1447		if (tsk->published)
1448			return -EOPNOTSUPP;
1449		if (dest->addrtype == TIPC_ADDR_NAME) {
1450			tsk->conn_type = dest->addr.name.name.type;
1451			tsk->conn_instance = dest->addr.name.name.instance;
1452		}
1453		msg_set_syn(hdr, 1);
1454	}
1455
1456	seq = &dest->addr.nameseq;
1457	if (dest->addrtype == TIPC_ADDR_MCAST)
1458		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1459
1460	if (dest->addrtype == TIPC_ADDR_NAME) {
1461		type = dest->addr.name.name.type;
1462		inst = dest->addr.name.name.instance;
1463		dnode = dest->addr.name.domain;
1464		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1465		if (unlikely(!dport && !dnode))
1466			return -EHOSTUNREACH;
1467	} else if (dest->addrtype == TIPC_ADDR_ID) {
1468		dnode = dest->addr.id.node;
1469	} else {
1470		return -EINVAL;
1471	}
1472
1473	/* Block or return if destination link is congested */
1474	rc = tipc_wait_for_cond(sock, &timeout,
1475				!tipc_dest_find(clinks, dnode, 0));
1476	if (unlikely(rc))
1477		return rc;
1478
1479	if (dest->addrtype == TIPC_ADDR_NAME) {
1480		msg_set_type(hdr, TIPC_NAMED_MSG);
1481		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1482		msg_set_nametype(hdr, type);
1483		msg_set_nameinst(hdr, inst);
1484		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1485		msg_set_destnode(hdr, dnode);
1486		msg_set_destport(hdr, dport);
1487	} else { /* TIPC_ADDR_ID */
1488		msg_set_type(hdr, TIPC_DIRECT_MSG);
1489		msg_set_lookup_scope(hdr, 0);
1490		msg_set_destnode(hdr, dnode);
1491		msg_set_destport(hdr, dest->addr.id.ref);
1492		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1493	}
1494
1495	__skb_queue_head_init(&pkts);
1496	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, true);
1497	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1498	if (unlikely(rc != dlen))
1499		return rc;
1500	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1501		__skb_queue_purge(&pkts);
1502		return -ENOMEM;
1503	}
1504
1505	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1506	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1507	if (unlikely(rc == -ELINKCONG)) {
1508		tipc_dest_push(clinks, dnode, 0);
1509		tsk->cong_link_cnt++;
1510		rc = 0;
1511	}
1512
1513	if (unlikely(syn && !rc)) {
1514		tipc_set_sk_state(sk, TIPC_CONNECTING);
1515		if (dlen && timeout) {
1516			timeout = msecs_to_jiffies(timeout);
1517			tipc_wait_for_connect(sock, &timeout);
1518		}
1519	}
1520
1521	return rc ? rc : dlen;
1522}
1523
1524/**
1525 * tipc_sendstream - send stream-oriented data
1526 * @sock: socket structure
1527 * @m: data to send
1528 * @dsz: total length of data to be transmitted
1529 *
1530 * Used for SOCK_STREAM data.
1531 *
1532 * Returns the number of bytes sent on success (or partial success),
1533 * or errno if no data sent
1534 */
1535static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1536{
1537	struct sock *sk = sock->sk;
1538	int ret;
1539
1540	lock_sock(sk);
1541	ret = __tipc_sendstream(sock, m, dsz);
1542	release_sock(sk);
1543
1544	return ret;
1545}
1546
1547static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1548{
1549	struct sock *sk = sock->sk;
1550	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1551	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1552	struct sk_buff_head *txq = &sk->sk_write_queue;
1553	struct tipc_sock *tsk = tipc_sk(sk);
1554	struct tipc_msg *hdr = &tsk->phdr;
1555	struct net *net = sock_net(sk);
1556	struct sk_buff *skb;
1557	u32 dnode = tsk_peer_node(tsk);
1558	int maxnagle = tsk->maxnagle;
1559	int maxpkt = tsk->max_pkt;
1560	int send, sent = 0;
1561	int blocks, rc = 0;
1562
1563	if (unlikely(dlen > INT_MAX))
1564		return -EMSGSIZE;
1565
1566	/* Handle implicit connection setup */
1567	if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1568		rc = __tipc_sendmsg(sock, m, dlen);
1569		if (dlen && dlen == rc) {
1570			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1571			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1572		}
1573		return rc;
1574	}
1575
1576	do {
1577		rc = tipc_wait_for_cond(sock, &timeout,
1578					(!tsk->cong_link_cnt &&
1579					 !tsk_conn_cong(tsk) &&
1580					 tipc_sk_connected(sk)));
1581		if (unlikely(rc))
1582			break;
1583		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1584		blocks = tsk->snd_backlog;
1585		if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1586		    send <= maxnagle) {
1587			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1588			if (unlikely(rc < 0))
1589				break;
1590			blocks += rc;
1591			tsk->msg_acc++;
1592			if (blocks <= 64 && tsk->expect_ack) {
1593				tsk->snd_backlog = blocks;
1594				sent += send;
1595				break;
1596			} else if (blocks > 64) {
1597				tsk->pkt_cnt += skb_queue_len(txq);
1598			} else {
1599				skb = skb_peek_tail(txq);
1600				if (skb) {
1601					msg_set_ack_required(buf_msg(skb));
1602					tsk->expect_ack = true;
1603				} else {
1604					tsk->expect_ack = false;
1605				}
1606				tsk->msg_acc = 0;
1607				tsk->pkt_cnt = 0;
1608			}
1609		} else {
1610			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1611			if (unlikely(rc != send))
1612				break;
1613			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1614		}
1615		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1616					 TIPC_DUMP_SK_SNDQ, " ");
1617		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1618		if (unlikely(rc == -ELINKCONG)) {
1619			tsk->cong_link_cnt = 1;
1620			rc = 0;
1621		}
1622		if (likely(!rc)) {
1623			tsk->snt_unacked += blocks;
1624			tsk->snd_backlog = 0;
1625			sent += send;
1626		}
1627	} while (sent < dlen && !rc);
1628
1629	return sent ? sent : rc;
1630}
1631
1632/**
1633 * tipc_send_packet - send a connection-oriented message
1634 * @sock: socket structure
1635 * @m: message to send
1636 * @dsz: length of data to be transmitted
1637 *
1638 * Used for SOCK_SEQPACKET messages.
1639 *
1640 * Returns the number of bytes sent on success, or errno otherwise
1641 */
1642static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1643{
1644	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1645		return -EMSGSIZE;
1646
1647	return tipc_sendstream(sock, m, dsz);
1648}
1649
1650/* tipc_sk_finish_conn - complete the setup of a connection
1651 */
1652static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1653				u32 peer_node)
1654{
1655	struct sock *sk = &tsk->sk;
1656	struct net *net = sock_net(sk);
1657	struct tipc_msg *msg = &tsk->phdr;
1658
1659	msg_set_syn(msg, 0);
1660	msg_set_destnode(msg, peer_node);
1661	msg_set_destport(msg, peer_port);
1662	msg_set_type(msg, TIPC_CONN_MSG);
1663	msg_set_lookup_scope(msg, 0);
1664	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1665
1666	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1667	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1668	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1669	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1670	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1671	tsk_set_nagle(tsk);
1672	__skb_queue_purge(&sk->sk_write_queue);
1673	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1674		return;
1675
1676	/* Fall back to message based flow control */
1677	tsk->rcv_win = FLOWCTL_MSG_WIN;
1678	tsk->snd_win = FLOWCTL_MSG_WIN;
1679}
1680
1681/**
1682 * tipc_sk_set_orig_addr - capture sender's address for received message
1683 * @m: descriptor for message info
1684 * @skb: received message
1685 *
1686 * Note: Address is not captured if not requested by receiver.
1687 */
1688static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1689{
1690	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1691	struct tipc_msg *hdr = buf_msg(skb);
1692
1693	if (!srcaddr)
1694		return;
1695
1696	srcaddr->sock.family = AF_TIPC;
1697	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1698	srcaddr->sock.scope = 0;
1699	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1700	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1701	srcaddr->sock.addr.name.domain = 0;
1702	m->msg_namelen = sizeof(struct sockaddr_tipc);
1703
1704	if (!msg_in_group(hdr))
1705		return;
1706
1707	/* Group message users may also want to know sending member's id */
1708	srcaddr->member.family = AF_TIPC;
1709	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1710	srcaddr->member.scope = 0;
1711	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1712	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1713	srcaddr->member.addr.name.domain = 0;
1714	m->msg_namelen = sizeof(*srcaddr);
1715}
1716
1717/**
1718 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1719 * @m: descriptor for message info
1720 * @skb: received message buffer
1721 * @tsk: TIPC port associated with message
1722 *
1723 * Note: Ancillary data is not captured if not requested by receiver.
1724 *
1725 * Returns 0 if successful, otherwise errno
1726 */
1727static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1728				 struct tipc_sock *tsk)
1729{
1730	struct tipc_msg *msg;
1731	u32 anc_data[3];
1732	u32 err;
1733	u32 dest_type;
1734	int has_name;
1735	int res;
1736
1737	if (likely(m->msg_controllen == 0))
1738		return 0;
1739	msg = buf_msg(skb);
1740
1741	/* Optionally capture errored message object(s) */
1742	err = msg ? msg_errcode(msg) : 0;
1743	if (unlikely(err)) {
1744		anc_data[0] = err;
1745		anc_data[1] = msg_data_sz(msg);
1746		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1747		if (res)
1748			return res;
1749		if (anc_data[1]) {
1750			if (skb_linearize(skb))
1751				return -ENOMEM;
1752			msg = buf_msg(skb);
1753			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1754				       msg_data(msg));
1755			if (res)
1756				return res;
1757		}
1758	}
1759
1760	/* Optionally capture message destination object */
1761	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1762	switch (dest_type) {
1763	case TIPC_NAMED_MSG:
1764		has_name = 1;
1765		anc_data[0] = msg_nametype(msg);
1766		anc_data[1] = msg_namelower(msg);
1767		anc_data[2] = msg_namelower(msg);
1768		break;
1769	case TIPC_MCAST_MSG:
1770		has_name = 1;
1771		anc_data[0] = msg_nametype(msg);
1772		anc_data[1] = msg_namelower(msg);
1773		anc_data[2] = msg_nameupper(msg);
1774		break;
1775	case TIPC_CONN_MSG:
1776		has_name = (tsk->conn_type != 0);
1777		anc_data[0] = tsk->conn_type;
1778		anc_data[1] = tsk->conn_instance;
1779		anc_data[2] = tsk->conn_instance;
1780		break;
1781	default:
1782		has_name = 0;
1783	}
1784	if (has_name) {
1785		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1786		if (res)
1787			return res;
1788	}
1789
1790	return 0;
1791}
1792
1793static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1794{
1795	struct sock *sk = &tsk->sk;
1796	struct sk_buff *skb = NULL;
1797	struct tipc_msg *msg;
1798	u32 peer_port = tsk_peer_port(tsk);
1799	u32 dnode = tsk_peer_node(tsk);
1800
1801	if (!tipc_sk_connected(sk))
1802		return NULL;
1803	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1804			      dnode, tsk_own_node(tsk), peer_port,
1805			      tsk->portid, TIPC_OK);
1806	if (!skb)
1807		return NULL;
1808	msg = buf_msg(skb);
1809	msg_set_conn_ack(msg, tsk->rcv_unacked);
1810	tsk->rcv_unacked = 0;
1811
1812	/* Adjust to and advertize the correct window limit */
1813	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1814		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1815		msg_set_adv_win(msg, tsk->rcv_win);
1816	}
1817	return skb;
1818}
1819
1820static void tipc_sk_send_ack(struct tipc_sock *tsk)
1821{
1822	struct sk_buff *skb;
1823
1824	skb = tipc_sk_build_ack(tsk);
1825	if (!skb)
1826		return;
1827
1828	tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1829			   msg_link_selector(buf_msg(skb)));
1830}
1831
1832static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1833{
1834	struct sock *sk = sock->sk;
1835	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1836	long timeo = *timeop;
1837	int err = sock_error(sk);
1838
1839	if (err)
1840		return err;
1841
1842	for (;;) {
1843		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1844			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1845				err = -ENOTCONN;
1846				break;
1847			}
1848			add_wait_queue(sk_sleep(sk), &wait);
1849			release_sock(sk);
1850			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1851			sched_annotate_sleep();
1852			lock_sock(sk);
1853			remove_wait_queue(sk_sleep(sk), &wait);
1854		}
1855		err = 0;
1856		if (!skb_queue_empty(&sk->sk_receive_queue))
1857			break;
1858		err = -EAGAIN;
1859		if (!timeo)
1860			break;
1861		err = sock_intr_errno(timeo);
1862		if (signal_pending(current))
1863			break;
1864
1865		err = sock_error(sk);
1866		if (err)
1867			break;
1868	}
1869	*timeop = timeo;
1870	return err;
1871}
1872
1873/**
1874 * tipc_recvmsg - receive packet-oriented message
1875 * @m: descriptor for message info
1876 * @buflen: length of user buffer area
1877 * @flags: receive flags
1878 *
1879 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1880 * If the complete message doesn't fit in user area, truncate it.
1881 *
1882 * Returns size of returned message data, errno otherwise
1883 */
1884static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1885			size_t buflen,	int flags)
1886{
1887	struct sock *sk = sock->sk;
1888	bool connected = !tipc_sk_type_connectionless(sk);
1889	struct tipc_sock *tsk = tipc_sk(sk);
1890	int rc, err, hlen, dlen, copy;
1891	struct tipc_skb_cb *skb_cb;
1892	struct sk_buff_head xmitq;
1893	struct tipc_msg *hdr;
1894	struct sk_buff *skb;
1895	bool grp_evt;
1896	long timeout;
1897
1898	/* Catch invalid receive requests */
1899	if (unlikely(!buflen))
1900		return -EINVAL;
1901
1902	lock_sock(sk);
1903	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1904		rc = -ENOTCONN;
1905		goto exit;
1906	}
1907	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1908
1909	/* Step rcv queue to first msg with data or error; wait if necessary */
1910	do {
1911		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1912		if (unlikely(rc))
1913			goto exit;
1914		skb = skb_peek(&sk->sk_receive_queue);
1915		skb_cb = TIPC_SKB_CB(skb);
1916		hdr = buf_msg(skb);
1917		dlen = msg_data_sz(hdr);
1918		hlen = msg_hdr_sz(hdr);
1919		err = msg_errcode(hdr);
1920		grp_evt = msg_is_grp_evt(hdr);
1921		if (likely(dlen || err))
1922			break;
1923		tsk_advance_rx_queue(sk);
1924	} while (1);
1925
1926	/* Collect msg meta data, including error code and rejected data */
1927	tipc_sk_set_orig_addr(m, skb);
1928	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1929	if (unlikely(rc))
1930		goto exit;
1931	hdr = buf_msg(skb);
1932
1933	/* Capture data if non-error msg, otherwise just set return value */
1934	if (likely(!err)) {
1935		int offset = skb_cb->bytes_read;
1936
1937		copy = min_t(int, dlen - offset, buflen);
1938		rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1939		if (unlikely(rc))
1940			goto exit;
1941		if (unlikely(offset + copy < dlen)) {
1942			if (flags & MSG_EOR) {
1943				if (!(flags & MSG_PEEK))
1944					skb_cb->bytes_read = offset + copy;
1945			} else {
1946				m->msg_flags |= MSG_TRUNC;
1947				skb_cb->bytes_read = 0;
1948			}
1949		} else {
1950			if (flags & MSG_EOR)
1951				m->msg_flags |= MSG_EOR;
1952			skb_cb->bytes_read = 0;
1953		}
1954	} else {
1955		copy = 0;
1956		rc = 0;
1957		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1958			rc = -ECONNRESET;
1959			goto exit;
1960		}
1961	}
1962
1963	/* Mark message as group event if applicable */
1964	if (unlikely(grp_evt)) {
1965		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1966			m->msg_flags |= MSG_EOR;
1967		m->msg_flags |= MSG_OOB;
1968		copy = 0;
1969	}
1970
1971	/* Caption of data or error code/rejected data was successful */
1972	if (unlikely(flags & MSG_PEEK))
1973		goto exit;
1974
1975	/* Send group flow control advertisement when applicable */
1976	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1977		__skb_queue_head_init(&xmitq);
1978		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1979					  msg_orignode(hdr), msg_origport(hdr),
1980					  &xmitq);
1981		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1982	}
1983
1984	if (skb_cb->bytes_read)
1985		goto exit;
1986
1987	tsk_advance_rx_queue(sk);
1988
1989	if (likely(!connected))
1990		goto exit;
1991
1992	/* Send connection flow control advertisement when applicable */
1993	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1994	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1995		tipc_sk_send_ack(tsk);
1996exit:
1997	release_sock(sk);
1998	return rc ? rc : copy;
1999}
2000
2001/**
2002 * tipc_recvstream - receive stream-oriented data
2003 * @m: descriptor for message info
2004 * @buflen: total size of user buffer area
2005 * @flags: receive flags
2006 *
2007 * Used for SOCK_STREAM messages only.  If not enough data is available
2008 * will optionally wait for more; never truncates data.
2009 *
2010 * Returns size of returned message data, errno otherwise
2011 */
2012static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2013			   size_t buflen, int flags)
2014{
2015	struct sock *sk = sock->sk;
2016	struct tipc_sock *tsk = tipc_sk(sk);
2017	struct sk_buff *skb;
2018	struct tipc_msg *hdr;
2019	struct tipc_skb_cb *skb_cb;
2020	bool peek = flags & MSG_PEEK;
2021	int offset, required, copy, copied = 0;
2022	int hlen, dlen, err, rc;
2023	long timeout;
2024
2025	/* Catch invalid receive attempts */
2026	if (unlikely(!buflen))
2027		return -EINVAL;
2028
2029	lock_sock(sk);
2030
2031	if (unlikely(sk->sk_state == TIPC_OPEN)) {
2032		rc = -ENOTCONN;
2033		goto exit;
2034	}
2035	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2036	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2037
2038	do {
2039		/* Look at first msg in receive queue; wait if necessary */
2040		rc = tipc_wait_for_rcvmsg(sock, &timeout);
2041		if (unlikely(rc))
2042			break;
2043		skb = skb_peek(&sk->sk_receive_queue);
2044		skb_cb = TIPC_SKB_CB(skb);
2045		hdr = buf_msg(skb);
2046		dlen = msg_data_sz(hdr);
2047		hlen = msg_hdr_sz(hdr);
2048		err = msg_errcode(hdr);
2049
2050		/* Discard any empty non-errored (SYN-) message */
2051		if (unlikely(!dlen && !err)) {
2052			tsk_advance_rx_queue(sk);
2053			continue;
2054		}
2055
2056		/* Collect msg meta data, incl. error code and rejected data */
2057		if (!copied) {
2058			tipc_sk_set_orig_addr(m, skb);
2059			rc = tipc_sk_anc_data_recv(m, skb, tsk);
2060			if (rc)
2061				break;
2062			hdr = buf_msg(skb);
2063		}
2064
2065		/* Copy data if msg ok, otherwise return error/partial data */
2066		if (likely(!err)) {
2067			offset = skb_cb->bytes_read;
2068			copy = min_t(int, dlen - offset, buflen - copied);
2069			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2070			if (unlikely(rc))
2071				break;
2072			copied += copy;
2073			offset += copy;
2074			if (unlikely(offset < dlen)) {
2075				if (!peek)
2076					skb_cb->bytes_read = offset;
2077				break;
2078			}
2079		} else {
2080			rc = 0;
2081			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2082				rc = -ECONNRESET;
2083			if (copied || rc)
2084				break;
2085		}
2086
2087		if (unlikely(peek))
2088			break;
2089
2090		tsk_advance_rx_queue(sk);
2091
2092		/* Send connection flow control advertisement when applicable */
2093		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2094		if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2095			tipc_sk_send_ack(tsk);
2096
2097		/* Exit if all requested data or FIN/error received */
2098		if (copied == buflen || err)
2099			break;
2100
2101	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2102exit:
2103	release_sock(sk);
2104	return copied ? copied : rc;
2105}
2106
2107/**
2108 * tipc_write_space - wake up thread if port congestion is released
2109 * @sk: socket
2110 */
2111static void tipc_write_space(struct sock *sk)
2112{
2113	struct socket_wq *wq;
2114
2115	rcu_read_lock();
2116	wq = rcu_dereference(sk->sk_wq);
2117	if (skwq_has_sleeper(wq))
2118		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2119						EPOLLWRNORM | EPOLLWRBAND);
2120	rcu_read_unlock();
2121}
2122
2123/**
2124 * tipc_data_ready - wake up threads to indicate messages have been received
2125 * @sk: socket
2126 */
2127static void tipc_data_ready(struct sock *sk)
2128{
2129	struct socket_wq *wq;
2130
2131	rcu_read_lock();
2132	wq = rcu_dereference(sk->sk_wq);
2133	if (skwq_has_sleeper(wq))
2134		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2135						EPOLLRDNORM | EPOLLRDBAND);
2136	rcu_read_unlock();
2137}
2138
2139static void tipc_sock_destruct(struct sock *sk)
2140{
2141	__skb_queue_purge(&sk->sk_receive_queue);
2142}
2143
2144static void tipc_sk_proto_rcv(struct sock *sk,
2145			      struct sk_buff_head *inputq,
2146			      struct sk_buff_head *xmitq)
2147{
2148	struct sk_buff *skb = __skb_dequeue(inputq);
2149	struct tipc_sock *tsk = tipc_sk(sk);
2150	struct tipc_msg *hdr = buf_msg(skb);
2151	struct tipc_group *grp = tsk->group;
2152	bool wakeup = false;
2153
2154	switch (msg_user(hdr)) {
2155	case CONN_MANAGER:
2156		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2157		return;
2158	case SOCK_WAKEUP:
2159		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2160		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2161		smp_wmb();
2162		tsk->cong_link_cnt--;
2163		wakeup = true;
2164		tipc_sk_push_backlog(tsk, false);
2165		break;
2166	case GROUP_PROTOCOL:
2167		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2168		break;
2169	case TOP_SRV:
2170		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2171				      hdr, inputq, xmitq);
2172		break;
2173	default:
2174		break;
2175	}
2176
2177	if (wakeup)
2178		sk->sk_write_space(sk);
2179
2180	kfree_skb(skb);
2181}
2182
2183/**
2184 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2185 * @tsk: TIPC socket
2186 * @skb: pointer to message buffer.
2187 * @xmitq: for Nagle ACK if any
2188 * Returns true if message should be added to receive queue, false otherwise
2189 */
2190static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2191				   struct sk_buff_head *xmitq)
2192{
2193	struct sock *sk = &tsk->sk;
2194	struct net *net = sock_net(sk);
2195	struct tipc_msg *hdr = buf_msg(skb);
2196	bool con_msg = msg_connected(hdr);
2197	u32 pport = tsk_peer_port(tsk);
2198	u32 pnode = tsk_peer_node(tsk);
2199	u32 oport = msg_origport(hdr);
2200	u32 onode = msg_orignode(hdr);
2201	int err = msg_errcode(hdr);
2202	unsigned long delay;
2203
2204	if (unlikely(msg_mcast(hdr)))
2205		return false;
2206	tsk->oneway = 0;
2207
2208	switch (sk->sk_state) {
2209	case TIPC_CONNECTING:
2210		/* Setup ACK */
2211		if (likely(con_msg)) {
2212			if (err)
2213				break;
2214			tipc_sk_finish_conn(tsk, oport, onode);
2215			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2216			/* ACK+ message with data is added to receive queue */
2217			if (msg_data_sz(hdr))
2218				return true;
2219			/* Empty ACK-, - wake up sleeping connect() and drop */
2220			sk->sk_state_change(sk);
2221			msg_set_dest_droppable(hdr, 1);
2222			return false;
2223		}
2224		/* Ignore connectionless message if not from listening socket */
2225		if (oport != pport || onode != pnode)
2226			return false;
2227
2228		/* Rejected SYN */
2229		if (err != TIPC_ERR_OVERLOAD)
2230			break;
2231
2232		/* Prepare for new setup attempt if we have a SYN clone */
2233		if (skb_queue_empty(&sk->sk_write_queue))
2234			break;
2235		get_random_bytes(&delay, 2);
2236		delay %= (tsk->conn_timeout / 4);
2237		delay = msecs_to_jiffies(delay + 100);
2238		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2239		return false;
2240	case TIPC_OPEN:
2241	case TIPC_DISCONNECTING:
2242		return false;
2243	case TIPC_LISTEN:
2244		/* Accept only SYN message */
2245		if (!msg_is_syn(hdr) &&
2246		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2247			return false;
2248		if (!con_msg && !err)
2249			return true;
2250		return false;
2251	case TIPC_ESTABLISHED:
2252		if (!skb_queue_empty(&sk->sk_write_queue))
2253			tipc_sk_push_backlog(tsk, false);
2254		/* Accept only connection-based messages sent by peer */
2255		if (likely(con_msg && !err && pport == oport &&
2256			   pnode == onode)) {
2257			if (msg_ack_required(hdr)) {
2258				struct sk_buff *skb;
2259
2260				skb = tipc_sk_build_ack(tsk);
2261				if (skb) {
2262					msg_set_nagle_ack(buf_msg(skb));
2263					__skb_queue_tail(xmitq, skb);
2264				}
2265			}
2266			return true;
2267		}
2268		if (!tsk_peer_msg(tsk, hdr))
2269			return false;
2270		if (!err)
2271			return true;
2272		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2273		tipc_node_remove_conn(net, pnode, tsk->portid);
2274		sk->sk_state_change(sk);
2275		return true;
2276	default:
2277		pr_err("Unknown sk_state %u\n", sk->sk_state);
2278	}
2279	/* Abort connection setup attempt */
2280	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2281	sk->sk_err = ECONNREFUSED;
2282	sk->sk_state_change(sk);
2283	return true;
2284}
2285
2286/**
2287 * rcvbuf_limit - get proper overload limit of socket receive queue
2288 * @sk: socket
2289 * @skb: message
2290 *
2291 * For connection oriented messages, irrespective of importance,
2292 * default queue limit is 2 MB.
2293 *
2294 * For connectionless messages, queue limits are based on message
2295 * importance as follows:
2296 *
2297 * TIPC_LOW_IMPORTANCE       (2 MB)
2298 * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2299 * TIPC_HIGH_IMPORTANCE      (8 MB)
2300 * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2301 *
2302 * Returns overload limit according to corresponding message importance
2303 */
2304static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2305{
2306	struct tipc_sock *tsk = tipc_sk(sk);
2307	struct tipc_msg *hdr = buf_msg(skb);
2308
2309	if (unlikely(msg_in_group(hdr)))
2310		return READ_ONCE(sk->sk_rcvbuf);
2311
2312	if (unlikely(!msg_connected(hdr)))
2313		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2314
2315	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2316		return READ_ONCE(sk->sk_rcvbuf);
2317
2318	return FLOWCTL_MSG_LIM;
2319}
2320
2321/**
2322 * tipc_sk_filter_rcv - validate incoming message
2323 * @sk: socket
2324 * @skb: pointer to message.
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 */
2332static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2333			       struct sk_buff_head *xmitq)
2334{
2335	bool sk_conn = !tipc_sk_type_connectionless(sk);
2336	struct tipc_sock *tsk = tipc_sk(sk);
2337	struct tipc_group *grp = tsk->group;
2338	struct tipc_msg *hdr = buf_msg(skb);
2339	struct net *net = sock_net(sk);
2340	struct sk_buff_head inputq;
2341	int mtyp = msg_type(hdr);
2342	int limit, err = TIPC_OK;
2343
2344	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2345	TIPC_SKB_CB(skb)->bytes_read = 0;
2346	__skb_queue_head_init(&inputq);
2347	__skb_queue_tail(&inputq, skb);
2348
2349	if (unlikely(!msg_isdata(hdr)))
2350		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2351
2352	if (unlikely(grp))
2353		tipc_group_filter_msg(grp, &inputq, xmitq);
2354
2355	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2356		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2357
2358	/* Validate and add to receive buffer if there is space */
2359	while ((skb = __skb_dequeue(&inputq))) {
2360		hdr = buf_msg(skb);
2361		limit = rcvbuf_limit(sk, skb);
2362		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2363		    (!sk_conn && msg_connected(hdr)) ||
2364		    (!grp && msg_in_group(hdr)))
2365			err = TIPC_ERR_NO_PORT;
2366		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2367			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2368					   "err_overload2!");
2369			atomic_inc(&sk->sk_drops);
2370			err = TIPC_ERR_OVERLOAD;
2371		}
2372
2373		if (unlikely(err)) {
2374			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2375				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2376						      "@filter_rcv!");
2377				__skb_queue_tail(xmitq, skb);
2378			}
2379			err = TIPC_OK;
2380			continue;
2381		}
2382		__skb_queue_tail(&sk->sk_receive_queue, skb);
2383		skb_set_owner_r(skb, sk);
2384		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2385					 "rcvq >90% allocated!");
2386		sk->sk_data_ready(sk);
2387	}
2388}
2389
2390/**
2391 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2392 * @sk: socket
2393 * @skb: message
2394 *
2395 * Caller must hold socket lock
2396 */
2397static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2398{
2399	unsigned int before = sk_rmem_alloc_get(sk);
2400	struct sk_buff_head xmitq;
2401	unsigned int added;
2402
2403	__skb_queue_head_init(&xmitq);
2404
2405	tipc_sk_filter_rcv(sk, skb, &xmitq);
2406	added = sk_rmem_alloc_get(sk) - before;
2407	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2408
2409	/* Send pending response/rejected messages, if any */
2410	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2411	return 0;
2412}
2413
2414/**
2415 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2416 *                   inputq and try adding them to socket or backlog queue
2417 * @inputq: list of incoming buffers with potentially different destinations
2418 * @sk: socket where the buffers should be enqueued
2419 * @dport: port number for the socket
2420 *
2421 * Caller must hold socket lock
2422 */
2423static 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 + usecs_to_jiffies(20000);
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 * @inputq: buffer list containing the buffers
2473 * Consumes all buffers in list until inputq is empty
2474 * Note: may be called in multiple threads referring to the same queue
2475 */
2476void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2477{
2478	struct sk_buff_head xmitq;
2479	u32 dnode, dport = 0;
2480	int err;
2481	struct tipc_sock *tsk;
2482	struct sock *sk;
2483	struct sk_buff *skb;
2484
2485	__skb_queue_head_init(&xmitq);
2486	while (skb_queue_len(inputq)) {
2487		dport = tipc_skb_peek_port(inputq, dport);
2488		tsk = tipc_sk_lookup(net, dport);
2489
2490		if (likely(tsk)) {
2491			sk = &tsk->sk;
2492			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2493				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2494				spin_unlock_bh(&sk->sk_lock.slock);
2495			}
2496			/* Send pending response/rejected messages, if any */
2497			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2498			sock_put(sk);
2499			continue;
2500		}
2501		/* No destination socket => dequeue skb if still there */
2502		skb = tipc_skb_dequeue(inputq, dport);
2503		if (!skb)
2504			return;
2505
2506		/* Try secondary lookup if unresolved named message */
2507		err = TIPC_ERR_NO_PORT;
2508		if (tipc_msg_lookup_dest(net, skb, &err))
2509			goto xmit;
2510
2511		/* Prepare for message rejection */
2512		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2513			continue;
2514
2515		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2516xmit:
2517		dnode = msg_destnode(buf_msg(skb));
2518		tipc_node_xmit_skb(net, skb, dnode, dport);
2519	}
2520}
2521
2522static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2523{
2524	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2525	struct sock *sk = sock->sk;
2526	int done;
2527
2528	do {
2529		int err = sock_error(sk);
2530		if (err)
2531			return err;
2532		if (!*timeo_p)
2533			return -ETIMEDOUT;
2534		if (signal_pending(current))
2535			return sock_intr_errno(*timeo_p);
2536		if (sk->sk_state == TIPC_DISCONNECTING)
2537			break;
2538
2539		add_wait_queue(sk_sleep(sk), &wait);
2540		done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2541				     &wait);
2542		remove_wait_queue(sk_sleep(sk), &wait);
2543	} while (!done);
2544	return 0;
2545}
2546
2547static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2548{
2549	if (addr->family != AF_TIPC)
2550		return false;
2551	if (addr->addrtype == TIPC_SERVICE_RANGE)
2552		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2553	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2554		addr->addrtype == TIPC_SOCKET_ADDR);
2555}
2556
2557/**
2558 * tipc_connect - establish a connection to another TIPC port
2559 * @sock: socket structure
2560 * @dest: socket address for destination port
2561 * @destlen: size of socket address data structure
2562 * @flags: file-related flags associated with socket
2563 *
2564 * Returns 0 on success, errno otherwise
2565 */
2566static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2567			int destlen, int flags)
2568{
2569	struct sock *sk = sock->sk;
2570	struct tipc_sock *tsk = tipc_sk(sk);
2571	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2572	struct msghdr m = {NULL,};
2573	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2574	int previous;
2575	int res = 0;
2576
2577	if (destlen != sizeof(struct sockaddr_tipc))
2578		return -EINVAL;
2579
2580	lock_sock(sk);
2581
2582	if (tsk->group) {
2583		res = -EINVAL;
2584		goto exit;
2585	}
2586
2587	if (dst->family == AF_UNSPEC) {
2588		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2589		if (!tipc_sk_type_connectionless(sk))
2590			res = -EINVAL;
2591		goto exit;
2592	}
2593	if (!tipc_sockaddr_is_sane(dst)) {
2594		res = -EINVAL;
2595		goto exit;
2596	}
2597	/* DGRAM/RDM connect(), just save the destaddr */
2598	if (tipc_sk_type_connectionless(sk)) {
2599		memcpy(&tsk->peer, dest, destlen);
2600		goto exit;
2601	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2602		res = -EINVAL;
2603		goto exit;
2604	}
2605
2606	previous = sk->sk_state;
2607
2608	switch (sk->sk_state) {
2609	case TIPC_OPEN:
2610		/* Send a 'SYN-' to destination */
2611		m.msg_name = dest;
2612		m.msg_namelen = destlen;
2613
2614		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2615		 * indicate send_msg() is never blocked.
2616		 */
2617		if (!timeout)
2618			m.msg_flags = MSG_DONTWAIT;
2619
2620		res = __tipc_sendmsg(sock, &m, 0);
2621		if ((res < 0) && (res != -EWOULDBLOCK))
2622			goto exit;
2623
2624		/* Just entered TIPC_CONNECTING state; the only
2625		 * difference is that return value in non-blocking
2626		 * case is EINPROGRESS, rather than EALREADY.
2627		 */
2628		res = -EINPROGRESS;
2629		fallthrough;
2630	case TIPC_CONNECTING:
2631		if (!timeout) {
2632			if (previous == TIPC_CONNECTING)
2633				res = -EALREADY;
2634			goto exit;
2635		}
2636		timeout = msecs_to_jiffies(timeout);
2637		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2638		res = tipc_wait_for_connect(sock, &timeout);
2639		break;
2640	case TIPC_ESTABLISHED:
2641		res = -EISCONN;
2642		break;
2643	default:
2644		res = -EINVAL;
2645	}
2646
2647exit:
2648	release_sock(sk);
2649	return res;
2650}
2651
2652/**
2653 * tipc_listen - allow socket to listen for incoming connections
2654 * @sock: socket structure
2655 * @len: (unused)
2656 *
2657 * Returns 0 on success, errno otherwise
2658 */
2659static int tipc_listen(struct socket *sock, int len)
2660{
2661	struct sock *sk = sock->sk;
2662	int res;
2663
2664	lock_sock(sk);
2665	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2666	release_sock(sk);
2667
2668	return res;
2669}
2670
2671static int tipc_wait_for_accept(struct socket *sock, long timeo)
2672{
2673	struct sock *sk = sock->sk;
2674	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2675	int err;
2676
2677	/* True wake-one mechanism for incoming connections: only
2678	 * one process gets woken up, not the 'whole herd'.
2679	 * Since we do not 'race & poll' for established sockets
2680	 * anymore, the common case will execute the loop only once.
2681	*/
2682	for (;;) {
2683		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2684			add_wait_queue(sk_sleep(sk), &wait);
2685			release_sock(sk);
2686			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2687			lock_sock(sk);
2688			remove_wait_queue(sk_sleep(sk), &wait);
2689		}
2690		err = 0;
2691		if (!skb_queue_empty(&sk->sk_receive_queue))
2692			break;
2693		err = -EAGAIN;
2694		if (!timeo)
2695			break;
2696		err = sock_intr_errno(timeo);
2697		if (signal_pending(current))
2698			break;
2699	}
2700	return err;
2701}
2702
2703/**
2704 * tipc_accept - wait for connection request
2705 * @sock: listening socket
2706 * @new_sock: new socket that is to be connected
2707 * @flags: file-related flags associated with socket
2708 *
2709 * Returns 0 on success, errno otherwise
2710 */
2711static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2712		       bool kern)
2713{
2714	struct sock *new_sk, *sk = sock->sk;
2715	struct tipc_sock *new_tsock;
2716	struct msghdr m = {NULL,};
2717	struct tipc_msg *msg;
2718	struct sk_buff *buf;
2719	long timeo;
2720	int res;
2721
2722	lock_sock(sk);
2723
2724	if (sk->sk_state != TIPC_LISTEN) {
2725		res = -EINVAL;
2726		goto exit;
2727	}
2728	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2729	res = tipc_wait_for_accept(sock, timeo);
2730	if (res)
2731		goto exit;
2732
2733	buf = skb_peek(&sk->sk_receive_queue);
2734
2735	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2736	if (res)
2737		goto exit;
2738	security_sk_clone(sock->sk, new_sock->sk);
2739
2740	new_sk = new_sock->sk;
2741	new_tsock = tipc_sk(new_sk);
2742	msg = buf_msg(buf);
2743
2744	/* we lock on new_sk; but lockdep sees the lock on sk */
2745	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2746
2747	/*
2748	 * Reject any stray messages received by new socket
2749	 * before the socket lock was taken (very, very unlikely)
2750	 */
2751	tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2752
2753	/* Connect new socket to it's peer */
2754	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2755
2756	tsk_set_importance(new_sk, msg_importance(msg));
2757	if (msg_named(msg)) {
2758		new_tsock->conn_type = msg_nametype(msg);
2759		new_tsock->conn_instance = msg_nameinst(msg);
2760	}
2761
2762	/*
2763	 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2764	 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2765	 */
2766	if (!msg_data_sz(msg)) {
2767		tsk_advance_rx_queue(sk);
2768	} else {
2769		__skb_dequeue(&sk->sk_receive_queue);
2770		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2771		skb_set_owner_r(buf, new_sk);
2772	}
2773	__tipc_sendstream(new_sock, &m, 0);
2774	release_sock(new_sk);
2775exit:
2776	release_sock(sk);
2777	return res;
2778}
2779
2780/**
2781 * tipc_shutdown - shutdown socket connection
2782 * @sock: socket structure
2783 * @how: direction to close (must be SHUT_RDWR)
2784 *
2785 * Terminates connection (if necessary), then purges socket's receive queue.
2786 *
2787 * Returns 0 on success, errno otherwise
2788 */
2789static int tipc_shutdown(struct socket *sock, int how)
2790{
2791	struct sock *sk = sock->sk;
2792	int res;
2793
2794	if (how != SHUT_RDWR)
2795		return -EINVAL;
2796
2797	lock_sock(sk);
2798
2799	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2800	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2801	sk->sk_shutdown = SHUTDOWN_MASK;
2802
2803	if (sk->sk_state == TIPC_DISCONNECTING) {
2804		/* Discard any unreceived messages */
2805		__skb_queue_purge(&sk->sk_receive_queue);
2806
2807		res = 0;
2808	} else {
2809		res = -ENOTCONN;
2810	}
2811	/* Wake up anyone sleeping in poll. */
2812	sk->sk_state_change(sk);
2813
2814	release_sock(sk);
2815	return res;
2816}
2817
2818static void tipc_sk_check_probing_state(struct sock *sk,
2819					struct sk_buff_head *list)
2820{
2821	struct tipc_sock *tsk = tipc_sk(sk);
2822	u32 pnode = tsk_peer_node(tsk);
2823	u32 pport = tsk_peer_port(tsk);
2824	u32 self = tsk_own_node(tsk);
2825	u32 oport = tsk->portid;
2826	struct sk_buff *skb;
2827
2828	if (tsk->probe_unacked) {
2829		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2830		sk->sk_err = ECONNABORTED;
2831		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2832		sk->sk_state_change(sk);
2833		return;
2834	}
2835	/* Prepare new probe */
2836	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2837			      pnode, self, pport, oport, TIPC_OK);
2838	if (skb)
2839		__skb_queue_tail(list, skb);
2840	tsk->probe_unacked = true;
2841	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2842}
2843
2844static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2845{
2846	struct tipc_sock *tsk = tipc_sk(sk);
2847
2848	/* Try again later if dest link is congested */
2849	if (tsk->cong_link_cnt) {
2850		sk_reset_timer(sk, &sk->sk_timer,
2851			       jiffies + msecs_to_jiffies(100));
2852		return;
2853	}
2854	/* Prepare SYN for retransmit */
2855	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2856}
2857
2858static void tipc_sk_timeout(struct timer_list *t)
2859{
2860	struct sock *sk = from_timer(sk, t, sk_timer);
2861	struct tipc_sock *tsk = tipc_sk(sk);
2862	u32 pnode = tsk_peer_node(tsk);
2863	struct sk_buff_head list;
2864	int rc = 0;
2865
2866	__skb_queue_head_init(&list);
2867	bh_lock_sock(sk);
2868
2869	/* Try again later if socket is busy */
2870	if (sock_owned_by_user(sk)) {
2871		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2872		bh_unlock_sock(sk);
2873		sock_put(sk);
2874		return;
2875	}
2876
2877	if (sk->sk_state == TIPC_ESTABLISHED)
2878		tipc_sk_check_probing_state(sk, &list);
2879	else if (sk->sk_state == TIPC_CONNECTING)
2880		tipc_sk_retry_connect(sk, &list);
2881
2882	bh_unlock_sock(sk);
2883
2884	if (!skb_queue_empty(&list))
2885		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2886
2887	/* SYN messages may cause link congestion */
2888	if (rc == -ELINKCONG) {
2889		tipc_dest_push(&tsk->cong_links, pnode, 0);
2890		tsk->cong_link_cnt = 1;
2891	}
2892	sock_put(sk);
2893}
2894
2895static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2896			   struct tipc_name_seq const *seq)
2897{
2898	struct sock *sk = &tsk->sk;
2899	struct net *net = sock_net(sk);
2900	struct publication *publ;
2901	u32 key;
2902
2903	if (scope != TIPC_NODE_SCOPE)
2904		scope = TIPC_CLUSTER_SCOPE;
2905
2906	if (tipc_sk_connected(sk))
2907		return -EINVAL;
2908	key = tsk->portid + tsk->pub_count + 1;
2909	if (key == tsk->portid)
2910		return -EADDRINUSE;
2911
2912	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2913				    scope, tsk->portid, key);
2914	if (unlikely(!publ))
2915		return -EINVAL;
2916
2917	list_add(&publ->binding_sock, &tsk->publications);
2918	tsk->pub_count++;
2919	tsk->published = 1;
2920	return 0;
2921}
2922
2923static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2924			    struct tipc_name_seq const *seq)
2925{
2926	struct net *net = sock_net(&tsk->sk);
2927	struct publication *publ;
2928	struct publication *safe;
2929	int rc = -EINVAL;
2930
2931	if (scope != TIPC_NODE_SCOPE)
2932		scope = TIPC_CLUSTER_SCOPE;
2933
2934	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2935		if (seq) {
2936			if (publ->scope != scope)
2937				continue;
2938			if (publ->type != seq->type)
2939				continue;
2940			if (publ->lower != seq->lower)
2941				continue;
2942			if (publ->upper != seq->upper)
2943				break;
2944			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2945					      publ->upper, publ->key);
2946			rc = 0;
2947			break;
2948		}
2949		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2950				      publ->upper, publ->key);
2951		rc = 0;
2952	}
2953	if (list_empty(&tsk->publications))
2954		tsk->published = 0;
2955	return rc;
2956}
2957
2958/* tipc_sk_reinit: set non-zero address in all existing sockets
2959 *                 when we go from standalone to network mode.
2960 */
2961void tipc_sk_reinit(struct net *net)
2962{
2963	struct tipc_net *tn = net_generic(net, tipc_net_id);
2964	struct rhashtable_iter iter;
2965	struct tipc_sock *tsk;
2966	struct tipc_msg *msg;
2967
2968	rhashtable_walk_enter(&tn->sk_rht, &iter);
2969
2970	do {
2971		rhashtable_walk_start(&iter);
2972
2973		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2974			sock_hold(&tsk->sk);
2975			rhashtable_walk_stop(&iter);
2976			lock_sock(&tsk->sk);
2977			msg = &tsk->phdr;
2978			msg_set_prevnode(msg, tipc_own_addr(net));
2979			msg_set_orignode(msg, tipc_own_addr(net));
2980			release_sock(&tsk->sk);
2981			rhashtable_walk_start(&iter);
2982			sock_put(&tsk->sk);
2983		}
2984
2985		rhashtable_walk_stop(&iter);
2986	} while (tsk == ERR_PTR(-EAGAIN));
2987
2988	rhashtable_walk_exit(&iter);
2989}
2990
2991static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2992{
2993	struct tipc_net *tn = net_generic(net, tipc_net_id);
2994	struct tipc_sock *tsk;
2995
2996	rcu_read_lock();
2997	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2998	if (tsk)
2999		sock_hold(&tsk->sk);
3000	rcu_read_unlock();
3001
3002	return tsk;
3003}
3004
3005static int tipc_sk_insert(struct tipc_sock *tsk)
3006{
3007	struct sock *sk = &tsk->sk;
3008	struct net *net = sock_net(sk);
3009	struct tipc_net *tn = net_generic(net, tipc_net_id);
3010	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3011	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
3012
3013	while (remaining--) {
3014		portid++;
3015		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3016			portid = TIPC_MIN_PORT;
3017		tsk->portid = portid;
3018		sock_hold(&tsk->sk);
3019		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3020						   tsk_rht_params))
3021			return 0;
3022		sock_put(&tsk->sk);
3023	}
3024
3025	return -1;
3026}
3027
3028static void tipc_sk_remove(struct tipc_sock *tsk)
3029{
3030	struct sock *sk = &tsk->sk;
3031	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3032
3033	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3034		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3035		__sock_put(sk);
3036	}
3037}
3038
3039static const struct rhashtable_params tsk_rht_params = {
3040	.nelem_hint = 192,
3041	.head_offset = offsetof(struct tipc_sock, node),
3042	.key_offset = offsetof(struct tipc_sock, portid),
3043	.key_len = sizeof(u32), /* portid */
3044	.max_size = 1048576,
3045	.min_size = 256,
3046	.automatic_shrinking = true,
3047};
3048
3049int tipc_sk_rht_init(struct net *net)
3050{
3051	struct tipc_net *tn = net_generic(net, tipc_net_id);
3052
3053	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3054}
3055
3056void tipc_sk_rht_destroy(struct net *net)
3057{
3058	struct tipc_net *tn = net_generic(net, tipc_net_id);
3059
3060	/* Wait for socket readers to complete */
3061	synchronize_net();
3062
3063	rhashtable_destroy(&tn->sk_rht);
3064}
3065
3066static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3067{
3068	struct net *net = sock_net(&tsk->sk);
3069	struct tipc_group *grp = tsk->group;
3070	struct tipc_msg *hdr = &tsk->phdr;
3071	struct tipc_name_seq seq;
3072	int rc;
3073
3074	if (mreq->type < TIPC_RESERVED_TYPES)
3075		return -EACCES;
3076	if (mreq->scope > TIPC_NODE_SCOPE)
3077		return -EINVAL;
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	seq.type = mreq->type;
3088	seq.lower = mreq->instance;
3089	seq.upper = seq.lower;
3090	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
3091	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
3092	if (rc) {
3093		tipc_group_delete(net, grp);
3094		tsk->group = NULL;
3095		return rc;
3096	}
3097	/* Eliminate any risk that a broadcast overtakes sent JOINs */
3098	tsk->mc_method.rcast = true;
3099	tsk->mc_method.mandatory = true;
3100	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3101	return rc;
3102}
3103
3104static int tipc_sk_leave(struct tipc_sock *tsk)
3105{
3106	struct net *net = sock_net(&tsk->sk);
3107	struct tipc_group *grp = tsk->group;
3108	struct tipc_name_seq seq;
3109	int scope;
3110
3111	if (!grp)
3112		return -EINVAL;
3113	tipc_group_self(grp, &seq, &scope);
3114	tipc_group_delete(net, grp);
3115	tsk->group = NULL;
3116	tipc_sk_withdraw(tsk, scope, &seq);
3117	return 0;
3118}
3119
3120/**
3121 * tipc_setsockopt - set socket option
3122 * @sock: socket structure
3123 * @lvl: option level
3124 * @opt: option identifier
3125 * @ov: pointer to new option value
3126 * @ol: length of option value
3127 *
3128 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3129 * (to ease compatibility).
3130 *
3131 * Returns 0 on success, errno otherwise
3132 */
3133static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3134			   sockptr_t ov, unsigned int ol)
3135{
3136	struct sock *sk = sock->sk;
3137	struct tipc_sock *tsk = tipc_sk(sk);
3138	struct tipc_group_req mreq;
3139	u32 value = 0;
3140	int res = 0;
3141
3142	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3143		return 0;
3144	if (lvl != SOL_TIPC)
3145		return -ENOPROTOOPT;
3146
3147	switch (opt) {
3148	case TIPC_IMPORTANCE:
3149	case TIPC_SRC_DROPPABLE:
3150	case TIPC_DEST_DROPPABLE:
3151	case TIPC_CONN_TIMEOUT:
3152	case TIPC_NODELAY:
3153		if (ol < sizeof(value))
3154			return -EINVAL;
3155		if (copy_from_sockptr(&value, ov, sizeof(u32)))
3156			return -EFAULT;
3157		break;
3158	case TIPC_GROUP_JOIN:
3159		if (ol < sizeof(mreq))
3160			return -EINVAL;
3161		if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3162			return -EFAULT;
3163		break;
3164	default:
3165		if (!sockptr_is_null(ov) || ol)
3166			return -EINVAL;
3167	}
3168
3169	lock_sock(sk);
3170
3171	switch (opt) {
3172	case TIPC_IMPORTANCE:
3173		res = tsk_set_importance(sk, value);
3174		break;
3175	case TIPC_SRC_DROPPABLE:
3176		if (sock->type != SOCK_STREAM)
3177			tsk_set_unreliable(tsk, value);
3178		else
3179			res = -ENOPROTOOPT;
3180		break;
3181	case TIPC_DEST_DROPPABLE:
3182		tsk_set_unreturnable(tsk, value);
3183		break;
3184	case TIPC_CONN_TIMEOUT:
3185		tipc_sk(sk)->conn_timeout = value;
3186		break;
3187	case TIPC_MCAST_BROADCAST:
3188		tsk->mc_method.rcast = false;
3189		tsk->mc_method.mandatory = true;
3190		break;
3191	case TIPC_MCAST_REPLICAST:
3192		tsk->mc_method.rcast = true;
3193		tsk->mc_method.mandatory = true;
3194		break;
3195	case TIPC_GROUP_JOIN:
3196		res = tipc_sk_join(tsk, &mreq);
3197		break;
3198	case TIPC_GROUP_LEAVE:
3199		res = tipc_sk_leave(tsk);
3200		break;
3201	case TIPC_NODELAY:
3202		tsk->nodelay = !!value;
3203		tsk_set_nagle(tsk);
3204		break;
3205	default:
3206		res = -EINVAL;
3207	}
3208
3209	release_sock(sk);
3210
3211	return res;
3212}
3213
3214/**
3215 * tipc_getsockopt - get socket option
3216 * @sock: socket structure
3217 * @lvl: option level
3218 * @opt: option identifier
3219 * @ov: receptacle for option value
3220 * @ol: receptacle for length of option value
3221 *
3222 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3223 * (to ease compatibility).
3224 *
3225 * Returns 0 on success, errno otherwise
3226 */
3227static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3228			   char __user *ov, int __user *ol)
3229{
3230	struct sock *sk = sock->sk;
3231	struct tipc_sock *tsk = tipc_sk(sk);
3232	struct tipc_name_seq seq;
3233	int len, scope;
3234	u32 value;
3235	int res;
3236
3237	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3238		return put_user(0, ol);
3239	if (lvl != SOL_TIPC)
3240		return -ENOPROTOOPT;
3241	res = get_user(len, ol);
3242	if (res)
3243		return res;
3244
3245	lock_sock(sk);
3246
3247	switch (opt) {
3248	case TIPC_IMPORTANCE:
3249		value = tsk_importance(tsk);
3250		break;
3251	case TIPC_SRC_DROPPABLE:
3252		value = tsk_unreliable(tsk);
3253		break;
3254	case TIPC_DEST_DROPPABLE:
3255		value = tsk_unreturnable(tsk);
3256		break;
3257	case TIPC_CONN_TIMEOUT:
3258		value = tsk->conn_timeout;
3259		/* no need to set "res", since already 0 at this point */
3260		break;
3261	case TIPC_NODE_RECVQ_DEPTH:
3262		value = 0; /* was tipc_queue_size, now obsolete */
3263		break;
3264	case TIPC_SOCK_RECVQ_DEPTH:
3265		value = skb_queue_len(&sk->sk_receive_queue);
3266		break;
3267	case TIPC_SOCK_RECVQ_USED:
3268		value = sk_rmem_alloc_get(sk);
3269		break;
3270	case TIPC_GROUP_JOIN:
3271		seq.type = 0;
3272		if (tsk->group)
3273			tipc_group_self(tsk->group, &seq, &scope);
3274		value = seq.type;
3275		break;
3276	default:
3277		res = -EINVAL;
3278	}
3279
3280	release_sock(sk);
3281
3282	if (res)
3283		return res;	/* "get" failed */
3284
3285	if (len < sizeof(value))
3286		return -EINVAL;
3287
3288	if (copy_to_user(ov, &value, sizeof(value)))
3289		return -EFAULT;
3290
3291	return put_user(sizeof(value), ol);
3292}
3293
3294static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3295{
3296	struct net *net = sock_net(sock->sk);
3297	struct tipc_sioc_nodeid_req nr = {0};
3298	struct tipc_sioc_ln_req lnr;
3299	void __user *argp = (void __user *)arg;
3300
3301	switch (cmd) {
3302	case SIOCGETLINKNAME:
3303		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3304			return -EFAULT;
3305		if (!tipc_node_get_linkname(net,
3306					    lnr.bearer_id & 0xffff, lnr.peer,
3307					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3308			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3309				return -EFAULT;
3310			return 0;
3311		}
3312		return -EADDRNOTAVAIL;
3313	case SIOCGETNODEID:
3314		if (copy_from_user(&nr, argp, sizeof(nr)))
3315			return -EFAULT;
3316		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3317			return -EADDRNOTAVAIL;
3318		if (copy_to_user(argp, &nr, sizeof(nr)))
3319			return -EFAULT;
3320		return 0;
3321	default:
3322		return -ENOIOCTLCMD;
3323	}
3324}
3325
3326static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3327{
3328	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3329	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3330	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3331
3332	tsk1->peer.family = AF_TIPC;
3333	tsk1->peer.addrtype = TIPC_ADDR_ID;
3334	tsk1->peer.scope = TIPC_NODE_SCOPE;
3335	tsk1->peer.addr.id.ref = tsk2->portid;
3336	tsk1->peer.addr.id.node = onode;
3337	tsk2->peer.family = AF_TIPC;
3338	tsk2->peer.addrtype = TIPC_ADDR_ID;
3339	tsk2->peer.scope = TIPC_NODE_SCOPE;
3340	tsk2->peer.addr.id.ref = tsk1->portid;
3341	tsk2->peer.addr.id.node = onode;
3342
3343	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3344	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3345	return 0;
3346}
3347
3348/* Protocol switches for the various types of TIPC sockets */
3349
3350static const struct proto_ops msg_ops = {
3351	.owner		= THIS_MODULE,
3352	.family		= AF_TIPC,
3353	.release	= tipc_release,
3354	.bind		= tipc_bind,
3355	.connect	= tipc_connect,
3356	.socketpair	= tipc_socketpair,
3357	.accept		= sock_no_accept,
3358	.getname	= tipc_getname,
3359	.poll		= tipc_poll,
3360	.ioctl		= tipc_ioctl,
3361	.listen		= sock_no_listen,
3362	.shutdown	= tipc_shutdown,
3363	.setsockopt	= tipc_setsockopt,
3364	.getsockopt	= tipc_getsockopt,
3365	.sendmsg	= tipc_sendmsg,
3366	.recvmsg	= tipc_recvmsg,
3367	.mmap		= sock_no_mmap,
3368	.sendpage	= sock_no_sendpage
3369};
3370
3371static const struct proto_ops packet_ops = {
3372	.owner		= THIS_MODULE,
3373	.family		= AF_TIPC,
3374	.release	= tipc_release,
3375	.bind		= tipc_bind,
3376	.connect	= tipc_connect,
3377	.socketpair	= tipc_socketpair,
3378	.accept		= tipc_accept,
3379	.getname	= tipc_getname,
3380	.poll		= tipc_poll,
3381	.ioctl		= tipc_ioctl,
3382	.listen		= tipc_listen,
3383	.shutdown	= tipc_shutdown,
3384	.setsockopt	= tipc_setsockopt,
3385	.getsockopt	= tipc_getsockopt,
3386	.sendmsg	= tipc_send_packet,
3387	.recvmsg	= tipc_recvmsg,
3388	.mmap		= sock_no_mmap,
3389	.sendpage	= sock_no_sendpage
3390};
3391
3392static const struct proto_ops stream_ops = {
3393	.owner		= THIS_MODULE,
3394	.family		= AF_TIPC,
3395	.release	= tipc_release,
3396	.bind		= tipc_bind,
3397	.connect	= tipc_connect,
3398	.socketpair	= tipc_socketpair,
3399	.accept		= tipc_accept,
3400	.getname	= tipc_getname,
3401	.poll		= tipc_poll,
3402	.ioctl		= tipc_ioctl,
3403	.listen		= tipc_listen,
3404	.shutdown	= tipc_shutdown,
3405	.setsockopt	= tipc_setsockopt,
3406	.getsockopt	= tipc_getsockopt,
3407	.sendmsg	= tipc_sendstream,
3408	.recvmsg	= tipc_recvstream,
3409	.mmap		= sock_no_mmap,
3410	.sendpage	= sock_no_sendpage
3411};
3412
3413static const struct net_proto_family tipc_family_ops = {
3414	.owner		= THIS_MODULE,
3415	.family		= AF_TIPC,
3416	.create		= tipc_sk_create
3417};
3418
3419static struct proto tipc_proto = {
3420	.name		= "TIPC",
3421	.owner		= THIS_MODULE,
3422	.obj_size	= sizeof(struct tipc_sock),
3423	.sysctl_rmem	= sysctl_tipc_rmem
3424};
3425
3426/**
3427 * tipc_socket_init - initialize TIPC socket interface
3428 *
3429 * Returns 0 on success, errno otherwise
3430 */
3431int tipc_socket_init(void)
3432{
3433	int res;
3434
3435	res = proto_register(&tipc_proto, 1);
3436	if (res) {
3437		pr_err("Failed to register TIPC protocol type\n");
3438		goto out;
3439	}
3440
3441	res = sock_register(&tipc_family_ops);
3442	if (res) {
3443		pr_err("Failed to register TIPC socket type\n");
3444		proto_unregister(&tipc_proto);
3445		goto out;
3446	}
3447 out:
3448	return res;
3449}
3450
3451/**
3452 * tipc_socket_stop - stop TIPC socket interface
3453 */
3454void tipc_socket_stop(void)
3455{
3456	sock_unregister(tipc_family_ops.family);
3457	proto_unregister(&tipc_proto);
3458}
3459
3460/* Caller should hold socket lock for the passed tipc socket. */
3461static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3462{
3463	u32 peer_node;
3464	u32 peer_port;
3465	struct nlattr *nest;
3466
3467	peer_node = tsk_peer_node(tsk);
3468	peer_port = tsk_peer_port(tsk);
3469
3470	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3471	if (!nest)
3472		return -EMSGSIZE;
3473
3474	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3475		goto msg_full;
3476	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3477		goto msg_full;
3478
3479	if (tsk->conn_type != 0) {
3480		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3481			goto msg_full;
3482		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3483			goto msg_full;
3484		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3485			goto msg_full;
3486	}
3487	nla_nest_end(skb, nest);
3488
3489	return 0;
3490
3491msg_full:
3492	nla_nest_cancel(skb, nest);
3493
3494	return -EMSGSIZE;
3495}
3496
3497static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3498			  *tsk)
3499{
3500	struct net *net = sock_net(skb->sk);
3501	struct sock *sk = &tsk->sk;
3502
3503	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3504	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3505		return -EMSGSIZE;
3506
3507	if (tipc_sk_connected(sk)) {
3508		if (__tipc_nl_add_sk_con(skb, tsk))
3509			return -EMSGSIZE;
3510	} else if (!list_empty(&tsk->publications)) {
3511		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3512			return -EMSGSIZE;
3513	}
3514	return 0;
3515}
3516
3517/* Caller should hold socket lock for the passed tipc socket. */
3518static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3519			    struct tipc_sock *tsk)
3520{
3521	struct nlattr *attrs;
3522	void *hdr;
3523
3524	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3525			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3526	if (!hdr)
3527		goto msg_cancel;
3528
3529	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3530	if (!attrs)
3531		goto genlmsg_cancel;
3532
3533	if (__tipc_nl_add_sk_info(skb, tsk))
3534		goto attr_msg_cancel;
3535
3536	nla_nest_end(skb, attrs);
3537	genlmsg_end(skb, hdr);
3538
3539	return 0;
3540
3541attr_msg_cancel:
3542	nla_nest_cancel(skb, attrs);
3543genlmsg_cancel:
3544	genlmsg_cancel(skb, hdr);
3545msg_cancel:
3546	return -EMSGSIZE;
3547}
3548
3549int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3550		    int (*skb_handler)(struct sk_buff *skb,
3551				       struct netlink_callback *cb,
3552				       struct tipc_sock *tsk))
3553{
3554	struct rhashtable_iter *iter = (void *)cb->args[4];
3555	struct tipc_sock *tsk;
3556	int err;
3557
3558	rhashtable_walk_start(iter);
3559	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3560		if (IS_ERR(tsk)) {
3561			err = PTR_ERR(tsk);
3562			if (err == -EAGAIN) {
3563				err = 0;
3564				continue;
3565			}
3566			break;
3567		}
3568
3569		sock_hold(&tsk->sk);
3570		rhashtable_walk_stop(iter);
3571		lock_sock(&tsk->sk);
3572		err = skb_handler(skb, cb, tsk);
3573		if (err) {
3574			release_sock(&tsk->sk);
3575			sock_put(&tsk->sk);
3576			goto out;
3577		}
3578		release_sock(&tsk->sk);
3579		rhashtable_walk_start(iter);
3580		sock_put(&tsk->sk);
3581	}
3582	rhashtable_walk_stop(iter);
3583out:
3584	return skb->len;
3585}
3586EXPORT_SYMBOL(tipc_nl_sk_walk);
3587
3588int tipc_dump_start(struct netlink_callback *cb)
3589{
3590	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3591}
3592EXPORT_SYMBOL(tipc_dump_start);
3593
3594int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3595{
3596	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3597	struct rhashtable_iter *iter = (void *)cb->args[4];
3598	struct tipc_net *tn = tipc_net(net);
3599
3600	if (!iter) {
3601		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3602		if (!iter)
3603			return -ENOMEM;
3604
3605		cb->args[4] = (long)iter;
3606	}
3607
3608	rhashtable_walk_enter(&tn->sk_rht, iter);
3609	return 0;
3610}
3611
3612int tipc_dump_done(struct netlink_callback *cb)
3613{
3614	struct rhashtable_iter *hti = (void *)cb->args[4];
3615
3616	rhashtable_walk_exit(hti);
3617	kfree(hti);
3618	return 0;
3619}
3620EXPORT_SYMBOL(tipc_dump_done);
3621
3622int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3623			   struct tipc_sock *tsk, u32 sk_filter_state,
3624			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3625{
3626	struct sock *sk = &tsk->sk;
3627	struct nlattr *attrs;
3628	struct nlattr *stat;
3629
3630	/*filter response w.r.t sk_state*/
3631	if (!(sk_filter_state & (1 << sk->sk_state)))
3632		return 0;
3633
3634	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3635	if (!attrs)
3636		goto msg_cancel;
3637
3638	if (__tipc_nl_add_sk_info(skb, tsk))
3639		goto attr_msg_cancel;
3640
3641	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3642	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3643	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3644	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3645			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3646					 sock_i_uid(sk))) ||
3647	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3648			      tipc_diag_gen_cookie(sk),
3649			      TIPC_NLA_SOCK_PAD))
3650		goto attr_msg_cancel;
3651
3652	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3653	if (!stat)
3654		goto attr_msg_cancel;
3655
3656	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3657			skb_queue_len(&sk->sk_receive_queue)) ||
3658	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3659			skb_queue_len(&sk->sk_write_queue)) ||
3660	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3661			atomic_read(&sk->sk_drops)))
3662		goto stat_msg_cancel;
3663
3664	if (tsk->cong_link_cnt &&
3665	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3666		goto stat_msg_cancel;
3667
3668	if (tsk_conn_cong(tsk) &&
3669	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3670		goto stat_msg_cancel;
3671
3672	nla_nest_end(skb, stat);
3673
3674	if (tsk->group)
3675		if (tipc_group_fill_sock_diag(tsk->group, skb))
3676			goto stat_msg_cancel;
3677
3678	nla_nest_end(skb, attrs);
3679
3680	return 0;
3681
3682stat_msg_cancel:
3683	nla_nest_cancel(skb, stat);
3684attr_msg_cancel:
3685	nla_nest_cancel(skb, attrs);
3686msg_cancel:
3687	return -EMSGSIZE;
3688}
3689EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3690
3691int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3692{
3693	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3694}
3695
3696/* Caller should hold socket lock for the passed tipc socket. */
3697static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3698				 struct netlink_callback *cb,
3699				 struct publication *publ)
3700{
3701	void *hdr;
3702	struct nlattr *attrs;
3703
3704	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3705			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3706	if (!hdr)
3707		goto msg_cancel;
3708
3709	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3710	if (!attrs)
3711		goto genlmsg_cancel;
3712
3713	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3714		goto attr_msg_cancel;
3715	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3716		goto attr_msg_cancel;
3717	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3718		goto attr_msg_cancel;
3719	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3720		goto attr_msg_cancel;
3721
3722	nla_nest_end(skb, attrs);
3723	genlmsg_end(skb, hdr);
3724
3725	return 0;
3726
3727attr_msg_cancel:
3728	nla_nest_cancel(skb, attrs);
3729genlmsg_cancel:
3730	genlmsg_cancel(skb, hdr);
3731msg_cancel:
3732	return -EMSGSIZE;
3733}
3734
3735/* Caller should hold socket lock for the passed tipc socket. */
3736static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3737				  struct netlink_callback *cb,
3738				  struct tipc_sock *tsk, u32 *last_publ)
3739{
3740	int err;
3741	struct publication *p;
3742
3743	if (*last_publ) {
3744		list_for_each_entry(p, &tsk->publications, binding_sock) {
3745			if (p->key == *last_publ)
3746				break;
3747		}
3748		if (list_entry_is_head(p, &tsk->publications, binding_sock)) {
3749			/* We never set seq or call nl_dump_check_consistent()
3750			 * this means that setting prev_seq here will cause the
3751			 * consistence check to fail in the netlink callback
3752			 * handler. Resulting in the last NLMSG_DONE message
3753			 * having the NLM_F_DUMP_INTR flag set.
3754			 */
3755			cb->prev_seq = 1;
3756			*last_publ = 0;
3757			return -EPIPE;
3758		}
3759	} else {
3760		p = list_first_entry(&tsk->publications, struct publication,
3761				     binding_sock);
3762	}
3763
3764	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3765		err = __tipc_nl_add_sk_publ(skb, cb, p);
3766		if (err) {
3767			*last_publ = p->key;
3768			return err;
3769		}
3770	}
3771	*last_publ = 0;
3772
3773	return 0;
3774}
3775
3776int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3777{
3778	int err;
3779	u32 tsk_portid = cb->args[0];
3780	u32 last_publ = cb->args[1];
3781	u32 done = cb->args[2];
3782	struct net *net = sock_net(skb->sk);
3783	struct tipc_sock *tsk;
3784
3785	if (!tsk_portid) {
3786		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3787		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3788
3789		if (!attrs[TIPC_NLA_SOCK])
3790			return -EINVAL;
3791
3792		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3793						  attrs[TIPC_NLA_SOCK],
3794						  tipc_nl_sock_policy, NULL);
3795		if (err)
3796			return err;
3797
3798		if (!sock[TIPC_NLA_SOCK_REF])
3799			return -EINVAL;
3800
3801		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3802	}
3803
3804	if (done)
3805		return 0;
3806
3807	tsk = tipc_sk_lookup(net, tsk_portid);
3808	if (!tsk)
3809		return -EINVAL;
3810
3811	lock_sock(&tsk->sk);
3812	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3813	if (!err)
3814		done = 1;
3815	release_sock(&tsk->sk);
3816	sock_put(&tsk->sk);
3817
3818	cb->args[0] = tsk_portid;
3819	cb->args[1] = last_publ;
3820	cb->args[2] = done;
3821
3822	return skb->len;
3823}
3824
3825/**
3826 * tipc_sk_filtering - check if a socket should be traced
3827 * @sk: the socket to be examined
3828 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3829 *  (portid, sock type, name type, name lower, name upper)
3830 *
3831 * Returns true if the socket meets the socket tuple data
3832 * (value 0 = 'any') or when there is no tuple set (all = 0),
3833 * otherwise false
3834 */
3835bool tipc_sk_filtering(struct sock *sk)
3836{
3837	struct tipc_sock *tsk;
3838	struct publication *p;
3839	u32 _port, _sktype, _type, _lower, _upper;
3840	u32 type = 0, lower = 0, upper = 0;
3841
3842	if (!sk)
3843		return true;
3844
3845	tsk = tipc_sk(sk);
3846
3847	_port = sysctl_tipc_sk_filter[0];
3848	_sktype = sysctl_tipc_sk_filter[1];
3849	_type = sysctl_tipc_sk_filter[2];
3850	_lower = sysctl_tipc_sk_filter[3];
3851	_upper = sysctl_tipc_sk_filter[4];
3852
3853	if (!_port && !_sktype && !_type && !_lower && !_upper)
3854		return true;
3855
3856	if (_port)
3857		return (_port == tsk->portid);
3858
3859	if (_sktype && _sktype != sk->sk_type)
3860		return false;
3861
3862	if (tsk->published) {
3863		p = list_first_entry_or_null(&tsk->publications,
3864					     struct publication, binding_sock);
3865		if (p) {
3866			type = p->type;
3867			lower = p->lower;
3868			upper = p->upper;
3869		}
3870	}
3871
3872	if (!tipc_sk_type_connectionless(sk)) {
3873		type = tsk->conn_type;
3874		lower = tsk->conn_instance;
3875		upper = tsk->conn_instance;
3876	}
3877
3878	if ((_type && _type != type) || (_lower && _lower != lower) ||
3879	    (_upper && _upper != upper))
3880		return false;
3881
3882	return true;
3883}
3884
3885u32 tipc_sock_get_portid(struct sock *sk)
3886{
3887	return (sk) ? (tipc_sk(sk))->portid : 0;
3888}
3889
3890/**
3891 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3892 *			both the rcv and backlog queues are considered
3893 * @sk: tipc sk to be checked
3894 * @skb: tipc msg to be checked
3895 *
3896 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3897 */
3898
3899bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3900{
3901	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3902	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3903	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3904
3905	return (qsize > lim * 90 / 100);
3906}
3907
3908/**
3909 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3910 *			only the rcv queue is considered
3911 * @sk: tipc sk to be checked
3912 * @skb: tipc msg to be checked
3913 *
3914 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3915 */
3916
3917bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3918{
3919	unsigned int lim = rcvbuf_limit(sk, skb);
3920	unsigned int qsize = sk_rmem_alloc_get(sk);
3921
3922	return (qsize > lim * 90 / 100);
3923}
3924
3925/**
3926 * tipc_sk_dump - dump TIPC socket
3927 * @sk: tipc sk to be dumped
3928 * @dqueues: bitmask to decide if any socket queue to be dumped?
3929 *           - TIPC_DUMP_NONE: don't dump socket queues
3930 *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3931 *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3932 *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3933 *           - TIPC_DUMP_ALL: dump all the socket queues above
3934 * @buf: returned buffer of dump data in format
3935 */
3936int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3937{
3938	int i = 0;
3939	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3940	struct tipc_sock *tsk;
3941	struct publication *p;
3942	bool tsk_connected;
3943
3944	if (!sk) {
3945		i += scnprintf(buf, sz, "sk data: (null)\n");
3946		return i;
3947	}
3948
3949	tsk = tipc_sk(sk);
3950	tsk_connected = !tipc_sk_type_connectionless(sk);
3951
3952	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3953	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3954	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3955	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3956	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3957	if (tsk_connected) {
3958		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3959		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3960		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3961		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3962	}
3963	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3964	if (tsk->published) {
3965		p = list_first_entry_or_null(&tsk->publications,
3966					     struct publication, binding_sock);
3967		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3968		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3969		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3970	}
3971	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3972	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3973	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3974	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3975	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3976	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3977	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3978	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3979	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3980	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3981	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3982	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3983	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3984	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3985
3986	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3987		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3988		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3989	}
3990
3991	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3992		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3993		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3994	}
3995
3996	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3997		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
3998		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3999		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4000			i += scnprintf(buf + i, sz - i, "  tail ");
4001			i += tipc_skb_dump(sk->sk_backlog.tail, false,
4002					   buf + i);
4003		}
4004	}
4005
4006	return i;
4007}
4008