xref: /linux/net/tipc/bcast.c (revision 93d90ad708b8da6efc0e487b66111aa9db7f70c7)
1 /*
2  * net/tipc/bcast.c: TIPC broadcast code
3  *
4  * Copyright (c) 2004-2006, 2014, Ericsson AB
5  * Copyright (c) 2004, Intel Corporation.
6  * Copyright (c) 2005, 2010-2011, Wind River Systems
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions are met:
11  *
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the names of the copyright holders nor the names of its
18  *    contributors may be used to endorse or promote products derived from
19  *    this software without specific prior written permission.
20  *
21  * Alternatively, this software may be distributed under the terms of the
22  * GNU General Public License ("GPL") version 2 as published by the Free
23  * Software Foundation.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
26  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
29  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35  * POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include "core.h"
39 #include "link.h"
40 #include "socket.h"
41 #include "msg.h"
42 #include "bcast.h"
43 #include "name_distr.h"
44 
45 #define	MAX_PKT_DEFAULT_MCAST	1500	/* bcast link max packet size (fixed) */
46 #define	BCLINK_WIN_DEFAULT	20	/* bcast link window size (default) */
47 #define	BCBEARER		MAX_BEARERS
48 
49 /**
50  * struct tipc_bcbearer_pair - a pair of bearers used by broadcast link
51  * @primary: pointer to primary bearer
52  * @secondary: pointer to secondary bearer
53  *
54  * Bearers must have same priority and same set of reachable destinations
55  * to be paired.
56  */
57 
58 struct tipc_bcbearer_pair {
59 	struct tipc_bearer *primary;
60 	struct tipc_bearer *secondary;
61 };
62 
63 /**
64  * struct tipc_bcbearer - bearer used by broadcast link
65  * @bearer: (non-standard) broadcast bearer structure
66  * @media: (non-standard) broadcast media structure
67  * @bpairs: array of bearer pairs
68  * @bpairs_temp: temporary array of bearer pairs used by tipc_bcbearer_sort()
69  * @remains: temporary node map used by tipc_bcbearer_send()
70  * @remains_new: temporary node map used tipc_bcbearer_send()
71  *
72  * Note: The fields labelled "temporary" are incorporated into the bearer
73  * to avoid consuming potentially limited stack space through the use of
74  * large local variables within multicast routines.  Concurrent access is
75  * prevented through use of the spinlock "bclink_lock".
76  */
77 struct tipc_bcbearer {
78 	struct tipc_bearer bearer;
79 	struct tipc_media media;
80 	struct tipc_bcbearer_pair bpairs[MAX_BEARERS];
81 	struct tipc_bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1];
82 	struct tipc_node_map remains;
83 	struct tipc_node_map remains_new;
84 };
85 
86 /**
87  * struct tipc_bclink - link used for broadcast messages
88  * @lock: spinlock governing access to structure
89  * @link: (non-standard) broadcast link structure
90  * @node: (non-standard) node structure representing b'cast link's peer node
91  * @flags: represent bclink states
92  * @bcast_nodes: map of broadcast-capable nodes
93  * @retransmit_to: node that most recently requested a retransmit
94  *
95  * Handles sequence numbering, fragmentation, bundling, etc.
96  */
97 struct tipc_bclink {
98 	spinlock_t lock;
99 	struct tipc_link link;
100 	struct tipc_node node;
101 	unsigned int flags;
102 	struct tipc_node_map bcast_nodes;
103 	struct tipc_node *retransmit_to;
104 };
105 
106 static struct tipc_bcbearer *bcbearer;
107 static struct tipc_bclink *bclink;
108 static struct tipc_link *bcl;
109 
110 const char tipc_bclink_name[] = "broadcast-link";
111 
112 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
113 			   struct tipc_node_map *nm_b,
114 			   struct tipc_node_map *nm_diff);
115 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node);
116 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node);
117 
118 static void tipc_bclink_lock(void)
119 {
120 	spin_lock_bh(&bclink->lock);
121 }
122 
123 static void tipc_bclink_unlock(void)
124 {
125 	struct tipc_node *node = NULL;
126 
127 	if (likely(!bclink->flags)) {
128 		spin_unlock_bh(&bclink->lock);
129 		return;
130 	}
131 
132 	if (bclink->flags & TIPC_BCLINK_RESET) {
133 		bclink->flags &= ~TIPC_BCLINK_RESET;
134 		node = tipc_bclink_retransmit_to();
135 	}
136 	spin_unlock_bh(&bclink->lock);
137 
138 	if (node)
139 		tipc_link_reset_all(node);
140 }
141 
142 uint  tipc_bclink_get_mtu(void)
143 {
144 	return MAX_PKT_DEFAULT_MCAST;
145 }
146 
147 void tipc_bclink_set_flags(unsigned int flags)
148 {
149 	bclink->flags |= flags;
150 }
151 
152 static u32 bcbuf_acks(struct sk_buff *buf)
153 {
154 	return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
155 }
156 
157 static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
158 {
159 	TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
160 }
161 
162 static void bcbuf_decr_acks(struct sk_buff *buf)
163 {
164 	bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
165 }
166 
167 void tipc_bclink_add_node(u32 addr)
168 {
169 	tipc_bclink_lock();
170 	tipc_nmap_add(&bclink->bcast_nodes, addr);
171 	tipc_bclink_unlock();
172 }
173 
174 void tipc_bclink_remove_node(u32 addr)
175 {
176 	tipc_bclink_lock();
177 	tipc_nmap_remove(&bclink->bcast_nodes, addr);
178 	tipc_bclink_unlock();
179 }
180 
181 static void bclink_set_last_sent(void)
182 {
183 	if (bcl->next_out)
184 		bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
185 	else
186 		bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
187 }
188 
189 u32 tipc_bclink_get_last_sent(void)
190 {
191 	return bcl->fsm_msg_cnt;
192 }
193 
194 static void bclink_update_last_sent(struct tipc_node *node, u32 seqno)
195 {
196 	node->bclink.last_sent = less_eq(node->bclink.last_sent, seqno) ?
197 						seqno : node->bclink.last_sent;
198 }
199 
200 
201 /**
202  * tipc_bclink_retransmit_to - get most recent node to request retransmission
203  *
204  * Called with bclink_lock locked
205  */
206 struct tipc_node *tipc_bclink_retransmit_to(void)
207 {
208 	return bclink->retransmit_to;
209 }
210 
211 /**
212  * bclink_retransmit_pkt - retransmit broadcast packets
213  * @after: sequence number of last packet to *not* retransmit
214  * @to: sequence number of last packet to retransmit
215  *
216  * Called with bclink_lock locked
217  */
218 static void bclink_retransmit_pkt(u32 after, u32 to)
219 {
220 	struct sk_buff *skb;
221 
222 	skb_queue_walk(&bcl->outqueue, skb) {
223 		if (more(buf_seqno(skb), after)) {
224 			tipc_link_retransmit(bcl, skb, mod(to - after));
225 			break;
226 		}
227 	}
228 }
229 
230 /**
231  * tipc_bclink_wakeup_users - wake up pending users
232  *
233  * Called with no locks taken
234  */
235 void tipc_bclink_wakeup_users(void)
236 {
237 	struct sk_buff *skb;
238 
239 	while ((skb = skb_dequeue(&bclink->link.waiting_sks)))
240 		tipc_sk_rcv(skb);
241 
242 }
243 
244 /**
245  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
246  * @n_ptr: node that sent acknowledgement info
247  * @acked: broadcast sequence # that has been acknowledged
248  *
249  * Node is locked, bclink_lock unlocked.
250  */
251 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
252 {
253 	struct sk_buff *skb, *tmp;
254 	struct sk_buff *next;
255 	unsigned int released = 0;
256 
257 	tipc_bclink_lock();
258 	/* Bail out if tx queue is empty (no clean up is required) */
259 	skb = skb_peek(&bcl->outqueue);
260 	if (!skb)
261 		goto exit;
262 
263 	/* Determine which messages need to be acknowledged */
264 	if (acked == INVALID_LINK_SEQ) {
265 		/*
266 		 * Contact with specified node has been lost, so need to
267 		 * acknowledge sent messages only (if other nodes still exist)
268 		 * or both sent and unsent messages (otherwise)
269 		 */
270 		if (bclink->bcast_nodes.count)
271 			acked = bcl->fsm_msg_cnt;
272 		else
273 			acked = bcl->next_out_no;
274 	} else {
275 		/*
276 		 * Bail out if specified sequence number does not correspond
277 		 * to a message that has been sent and not yet acknowledged
278 		 */
279 		if (less(acked, buf_seqno(skb)) ||
280 		    less(bcl->fsm_msg_cnt, acked) ||
281 		    less_eq(acked, n_ptr->bclink.acked))
282 			goto exit;
283 	}
284 
285 	/* Skip over packets that node has previously acknowledged */
286 	skb_queue_walk(&bcl->outqueue, skb) {
287 		if (more(buf_seqno(skb), n_ptr->bclink.acked))
288 			break;
289 	}
290 
291 	/* Update packets that node is now acknowledging */
292 	skb_queue_walk_from_safe(&bcl->outqueue, skb, tmp) {
293 		if (more(buf_seqno(skb), acked))
294 			break;
295 
296 		next = tipc_skb_queue_next(&bcl->outqueue, skb);
297 		if (skb != bcl->next_out) {
298 			bcbuf_decr_acks(skb);
299 		} else {
300 			bcbuf_set_acks(skb, 0);
301 			bcl->next_out = next;
302 			bclink_set_last_sent();
303 		}
304 
305 		if (bcbuf_acks(skb) == 0) {
306 			__skb_unlink(skb, &bcl->outqueue);
307 			kfree_skb(skb);
308 			released = 1;
309 		}
310 	}
311 	n_ptr->bclink.acked = acked;
312 
313 	/* Try resolving broadcast link congestion, if necessary */
314 	if (unlikely(bcl->next_out)) {
315 		tipc_link_push_packets(bcl);
316 		bclink_set_last_sent();
317 	}
318 	if (unlikely(released && !skb_queue_empty(&bcl->waiting_sks)))
319 		n_ptr->action_flags |= TIPC_WAKEUP_BCAST_USERS;
320 
321 exit:
322 	tipc_bclink_unlock();
323 }
324 
325 /**
326  * tipc_bclink_update_link_state - update broadcast link state
327  *
328  * RCU and node lock set
329  */
330 void tipc_bclink_update_link_state(struct tipc_node *n_ptr, u32 last_sent)
331 {
332 	struct sk_buff *buf;
333 
334 	/* Ignore "stale" link state info */
335 	if (less_eq(last_sent, n_ptr->bclink.last_in))
336 		return;
337 
338 	/* Update link synchronization state; quit if in sync */
339 	bclink_update_last_sent(n_ptr, last_sent);
340 
341 	if (n_ptr->bclink.last_sent == n_ptr->bclink.last_in)
342 		return;
343 
344 	/* Update out-of-sync state; quit if loss is still unconfirmed */
345 	if ((++n_ptr->bclink.oos_state) == 1) {
346 		if (n_ptr->bclink.deferred_size < (TIPC_MIN_LINK_WIN / 2))
347 			return;
348 		n_ptr->bclink.oos_state++;
349 	}
350 
351 	/* Don't NACK if one has been recently sent (or seen) */
352 	if (n_ptr->bclink.oos_state & 0x1)
353 		return;
354 
355 	/* Send NACK */
356 	buf = tipc_buf_acquire(INT_H_SIZE);
357 	if (buf) {
358 		struct tipc_msg *msg = buf_msg(buf);
359 		struct sk_buff *skb = skb_peek(&n_ptr->bclink.deferred_queue);
360 		u32 to = skb ? buf_seqno(skb) - 1 : n_ptr->bclink.last_sent;
361 
362 		tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
363 			      INT_H_SIZE, n_ptr->addr);
364 		msg_set_non_seq(msg, 1);
365 		msg_set_mc_netid(msg, tipc_net_id);
366 		msg_set_bcast_ack(msg, n_ptr->bclink.last_in);
367 		msg_set_bcgap_after(msg, n_ptr->bclink.last_in);
368 		msg_set_bcgap_to(msg, to);
369 
370 		tipc_bclink_lock();
371 		tipc_bearer_send(MAX_BEARERS, buf, NULL);
372 		bcl->stats.sent_nacks++;
373 		tipc_bclink_unlock();
374 		kfree_skb(buf);
375 
376 		n_ptr->bclink.oos_state++;
377 	}
378 }
379 
380 /**
381  * bclink_peek_nack - monitor retransmission requests sent by other nodes
382  *
383  * Delay any upcoming NACK by this node if another node has already
384  * requested the first message this node is going to ask for.
385  */
386 static void bclink_peek_nack(struct tipc_msg *msg)
387 {
388 	struct tipc_node *n_ptr = tipc_node_find(msg_destnode(msg));
389 
390 	if (unlikely(!n_ptr))
391 		return;
392 
393 	tipc_node_lock(n_ptr);
394 
395 	if (n_ptr->bclink.recv_permitted &&
396 	    (n_ptr->bclink.last_in != n_ptr->bclink.last_sent) &&
397 	    (n_ptr->bclink.last_in == msg_bcgap_after(msg)))
398 		n_ptr->bclink.oos_state = 2;
399 
400 	tipc_node_unlock(n_ptr);
401 }
402 
403 /* tipc_bclink_xmit - broadcast buffer chain to all nodes in cluster
404  *                    and to identified node local sockets
405  * @list: chain of buffers containing message
406  * Consumes the buffer chain, except when returning -ELINKCONG
407  * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
408  */
409 int tipc_bclink_xmit(struct sk_buff_head *list)
410 {
411 	int rc = 0;
412 	int bc = 0;
413 	struct sk_buff *skb;
414 
415 	/* Prepare clone of message for local node */
416 	skb = tipc_msg_reassemble(list);
417 	if (unlikely(!skb)) {
418 		__skb_queue_purge(list);
419 		return -EHOSTUNREACH;
420 	}
421 
422 	/* Broadcast to all other nodes */
423 	if (likely(bclink)) {
424 		tipc_bclink_lock();
425 		if (likely(bclink->bcast_nodes.count)) {
426 			rc = __tipc_link_xmit(bcl, list);
427 			if (likely(!rc)) {
428 				u32 len = skb_queue_len(&bcl->outqueue);
429 
430 				bclink_set_last_sent();
431 				bcl->stats.queue_sz_counts++;
432 				bcl->stats.accu_queue_sz += len;
433 			}
434 			bc = 1;
435 		}
436 		tipc_bclink_unlock();
437 	}
438 
439 	if (unlikely(!bc))
440 		__skb_queue_purge(list);
441 
442 	/* Deliver message clone */
443 	if (likely(!rc))
444 		tipc_sk_mcast_rcv(skb);
445 	else
446 		kfree_skb(skb);
447 
448 	return rc;
449 }
450 
451 /**
452  * bclink_accept_pkt - accept an incoming, in-sequence broadcast packet
453  *
454  * Called with both sending node's lock and bclink_lock taken.
455  */
456 static void bclink_accept_pkt(struct tipc_node *node, u32 seqno)
457 {
458 	bclink_update_last_sent(node, seqno);
459 	node->bclink.last_in = seqno;
460 	node->bclink.oos_state = 0;
461 	bcl->stats.recv_info++;
462 
463 	/*
464 	 * Unicast an ACK periodically, ensuring that
465 	 * all nodes in the cluster don't ACK at the same time
466 	 */
467 	if (((seqno - tipc_own_addr) % TIPC_MIN_LINK_WIN) == 0) {
468 		tipc_link_proto_xmit(node->active_links[node->addr & 1],
469 				     STATE_MSG, 0, 0, 0, 0, 0);
470 		bcl->stats.sent_acks++;
471 	}
472 }
473 
474 /**
475  * tipc_bclink_rcv - receive a broadcast packet, and deliver upwards
476  *
477  * RCU is locked, no other locks set
478  */
479 void tipc_bclink_rcv(struct sk_buff *buf)
480 {
481 	struct tipc_msg *msg = buf_msg(buf);
482 	struct tipc_node *node;
483 	u32 next_in;
484 	u32 seqno;
485 	int deferred = 0;
486 
487 	/* Screen out unwanted broadcast messages */
488 	if (msg_mc_netid(msg) != tipc_net_id)
489 		goto exit;
490 
491 	node = tipc_node_find(msg_prevnode(msg));
492 	if (unlikely(!node))
493 		goto exit;
494 
495 	tipc_node_lock(node);
496 	if (unlikely(!node->bclink.recv_permitted))
497 		goto unlock;
498 
499 	/* Handle broadcast protocol message */
500 	if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
501 		if (msg_type(msg) != STATE_MSG)
502 			goto unlock;
503 		if (msg_destnode(msg) == tipc_own_addr) {
504 			tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
505 			tipc_node_unlock(node);
506 			tipc_bclink_lock();
507 			bcl->stats.recv_nacks++;
508 			bclink->retransmit_to = node;
509 			bclink_retransmit_pkt(msg_bcgap_after(msg),
510 					      msg_bcgap_to(msg));
511 			tipc_bclink_unlock();
512 		} else {
513 			tipc_node_unlock(node);
514 			bclink_peek_nack(msg);
515 		}
516 		goto exit;
517 	}
518 
519 	/* Handle in-sequence broadcast message */
520 	seqno = msg_seqno(msg);
521 	next_in = mod(node->bclink.last_in + 1);
522 
523 	if (likely(seqno == next_in)) {
524 receive:
525 		/* Deliver message to destination */
526 		if (likely(msg_isdata(msg))) {
527 			tipc_bclink_lock();
528 			bclink_accept_pkt(node, seqno);
529 			tipc_bclink_unlock();
530 			tipc_node_unlock(node);
531 			if (likely(msg_mcast(msg)))
532 				tipc_sk_mcast_rcv(buf);
533 			else
534 				kfree_skb(buf);
535 		} else if (msg_user(msg) == MSG_BUNDLER) {
536 			tipc_bclink_lock();
537 			bclink_accept_pkt(node, seqno);
538 			bcl->stats.recv_bundles++;
539 			bcl->stats.recv_bundled += msg_msgcnt(msg);
540 			tipc_bclink_unlock();
541 			tipc_node_unlock(node);
542 			tipc_link_bundle_rcv(buf);
543 		} else if (msg_user(msg) == MSG_FRAGMENTER) {
544 			tipc_buf_append(&node->bclink.reasm_buf, &buf);
545 			if (unlikely(!buf && !node->bclink.reasm_buf))
546 				goto unlock;
547 			tipc_bclink_lock();
548 			bclink_accept_pkt(node, seqno);
549 			bcl->stats.recv_fragments++;
550 			if (buf) {
551 				bcl->stats.recv_fragmented++;
552 				msg = buf_msg(buf);
553 				tipc_bclink_unlock();
554 				goto receive;
555 			}
556 			tipc_bclink_unlock();
557 			tipc_node_unlock(node);
558 		} else if (msg_user(msg) == NAME_DISTRIBUTOR) {
559 			tipc_bclink_lock();
560 			bclink_accept_pkt(node, seqno);
561 			tipc_bclink_unlock();
562 			tipc_node_unlock(node);
563 			tipc_named_rcv(buf);
564 		} else {
565 			tipc_bclink_lock();
566 			bclink_accept_pkt(node, seqno);
567 			tipc_bclink_unlock();
568 			tipc_node_unlock(node);
569 			kfree_skb(buf);
570 		}
571 		buf = NULL;
572 
573 		/* Determine new synchronization state */
574 		tipc_node_lock(node);
575 		if (unlikely(!tipc_node_is_up(node)))
576 			goto unlock;
577 
578 		if (node->bclink.last_in == node->bclink.last_sent)
579 			goto unlock;
580 
581 		if (skb_queue_empty(&node->bclink.deferred_queue)) {
582 			node->bclink.oos_state = 1;
583 			goto unlock;
584 		}
585 
586 		msg = buf_msg(skb_peek(&node->bclink.deferred_queue));
587 		seqno = msg_seqno(msg);
588 		next_in = mod(next_in + 1);
589 		if (seqno != next_in)
590 			goto unlock;
591 
592 		/* Take in-sequence message from deferred queue & deliver it */
593 		buf = __skb_dequeue(&node->bclink.deferred_queue);
594 		goto receive;
595 	}
596 
597 	/* Handle out-of-sequence broadcast message */
598 	if (less(next_in, seqno)) {
599 		deferred = tipc_link_defer_pkt(&node->bclink.deferred_queue,
600 					       buf);
601 		bclink_update_last_sent(node, seqno);
602 		buf = NULL;
603 	}
604 
605 	tipc_bclink_lock();
606 
607 	if (deferred)
608 		bcl->stats.deferred_recv++;
609 	else
610 		bcl->stats.duplicates++;
611 
612 	tipc_bclink_unlock();
613 
614 unlock:
615 	tipc_node_unlock(node);
616 exit:
617 	kfree_skb(buf);
618 }
619 
620 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
621 {
622 	return (n_ptr->bclink.recv_permitted &&
623 		(tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
624 }
625 
626 
627 /**
628  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
629  *
630  * Send packet over as many bearers as necessary to reach all nodes
631  * that have joined the broadcast link.
632  *
633  * Returns 0 (packet sent successfully) under all circumstances,
634  * since the broadcast link's pseudo-bearer never blocks
635  */
636 static int tipc_bcbearer_send(struct sk_buff *buf, struct tipc_bearer *unused1,
637 			      struct tipc_media_addr *unused2)
638 {
639 	int bp_index;
640 	struct tipc_msg *msg = buf_msg(buf);
641 
642 	/* Prepare broadcast link message for reliable transmission,
643 	 * if first time trying to send it;
644 	 * preparation is skipped for broadcast link protocol messages
645 	 * since they are sent in an unreliable manner and don't need it
646 	 */
647 	if (likely(!msg_non_seq(buf_msg(buf)))) {
648 		bcbuf_set_acks(buf, bclink->bcast_nodes.count);
649 		msg_set_non_seq(msg, 1);
650 		msg_set_mc_netid(msg, tipc_net_id);
651 		bcl->stats.sent_info++;
652 
653 		if (WARN_ON(!bclink->bcast_nodes.count)) {
654 			dump_stack();
655 			return 0;
656 		}
657 	}
658 
659 	/* Send buffer over bearers until all targets reached */
660 	bcbearer->remains = bclink->bcast_nodes;
661 
662 	for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
663 		struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
664 		struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
665 		struct tipc_bearer *bp[2] = {p, s};
666 		struct tipc_bearer *b = bp[msg_link_selector(msg)];
667 		struct sk_buff *tbuf;
668 
669 		if (!p)
670 			break; /* No more bearers to try */
671 		if (!b)
672 			b = p;
673 		tipc_nmap_diff(&bcbearer->remains, &b->nodes,
674 			       &bcbearer->remains_new);
675 		if (bcbearer->remains_new.count == bcbearer->remains.count)
676 			continue; /* Nothing added by bearer pair */
677 
678 		if (bp_index == 0) {
679 			/* Use original buffer for first bearer */
680 			tipc_bearer_send(b->identity, buf, &b->bcast_addr);
681 		} else {
682 			/* Avoid concurrent buffer access */
683 			tbuf = pskb_copy_for_clone(buf, GFP_ATOMIC);
684 			if (!tbuf)
685 				break;
686 			tipc_bearer_send(b->identity, tbuf, &b->bcast_addr);
687 			kfree_skb(tbuf); /* Bearer keeps a clone */
688 		}
689 		if (bcbearer->remains_new.count == 0)
690 			break; /* All targets reached */
691 
692 		bcbearer->remains = bcbearer->remains_new;
693 	}
694 
695 	return 0;
696 }
697 
698 /**
699  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
700  */
701 void tipc_bcbearer_sort(struct tipc_node_map *nm_ptr, u32 node, bool action)
702 {
703 	struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
704 	struct tipc_bcbearer_pair *bp_curr;
705 	struct tipc_bearer *b;
706 	int b_index;
707 	int pri;
708 
709 	tipc_bclink_lock();
710 
711 	if (action)
712 		tipc_nmap_add(nm_ptr, node);
713 	else
714 		tipc_nmap_remove(nm_ptr, node);
715 
716 	/* Group bearers by priority (can assume max of two per priority) */
717 	memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
718 
719 	rcu_read_lock();
720 	for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
721 		b = rcu_dereference_rtnl(bearer_list[b_index]);
722 		if (!b || !b->nodes.count)
723 			continue;
724 
725 		if (!bp_temp[b->priority].primary)
726 			bp_temp[b->priority].primary = b;
727 		else
728 			bp_temp[b->priority].secondary = b;
729 	}
730 	rcu_read_unlock();
731 
732 	/* Create array of bearer pairs for broadcasting */
733 	bp_curr = bcbearer->bpairs;
734 	memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
735 
736 	for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
737 
738 		if (!bp_temp[pri].primary)
739 			continue;
740 
741 		bp_curr->primary = bp_temp[pri].primary;
742 
743 		if (bp_temp[pri].secondary) {
744 			if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
745 					    &bp_temp[pri].secondary->nodes)) {
746 				bp_curr->secondary = bp_temp[pri].secondary;
747 			} else {
748 				bp_curr++;
749 				bp_curr->primary = bp_temp[pri].secondary;
750 			}
751 		}
752 
753 		bp_curr++;
754 	}
755 
756 	tipc_bclink_unlock();
757 }
758 
759 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
760 				      struct tipc_stats *stats)
761 {
762 	int i;
763 	struct nlattr *nest;
764 
765 	struct nla_map {
766 		__u32 key;
767 		__u32 val;
768 	};
769 
770 	struct nla_map map[] = {
771 		{TIPC_NLA_STATS_RX_INFO, stats->recv_info},
772 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
773 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
774 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
775 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
776 		{TIPC_NLA_STATS_TX_INFO, stats->sent_info},
777 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
778 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
779 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
780 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
781 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
782 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
783 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
784 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
785 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
786 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
787 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
788 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
789 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
790 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
791 	};
792 
793 	nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
794 	if (!nest)
795 		return -EMSGSIZE;
796 
797 	for (i = 0; i <  ARRAY_SIZE(map); i++)
798 		if (nla_put_u32(skb, map[i].key, map[i].val))
799 			goto msg_full;
800 
801 	nla_nest_end(skb, nest);
802 
803 	return 0;
804 msg_full:
805 	nla_nest_cancel(skb, nest);
806 
807 	return -EMSGSIZE;
808 }
809 
810 int tipc_nl_add_bc_link(struct tipc_nl_msg *msg)
811 {
812 	int err;
813 	void *hdr;
814 	struct nlattr *attrs;
815 	struct nlattr *prop;
816 
817 	if (!bcl)
818 		return 0;
819 
820 	tipc_bclink_lock();
821 
822 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_v2_family,
823 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
824 	if (!hdr)
825 		return -EMSGSIZE;
826 
827 	attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
828 	if (!attrs)
829 		goto msg_full;
830 
831 	/* The broadcast link is always up */
832 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
833 		goto attr_msg_full;
834 
835 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
836 		goto attr_msg_full;
837 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
838 		goto attr_msg_full;
839 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, bcl->next_in_no))
840 		goto attr_msg_full;
841 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, bcl->next_out_no))
842 		goto attr_msg_full;
843 
844 	prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
845 	if (!prop)
846 		goto attr_msg_full;
847 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->queue_limit[0]))
848 		goto prop_msg_full;
849 	nla_nest_end(msg->skb, prop);
850 
851 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
852 	if (err)
853 		goto attr_msg_full;
854 
855 	tipc_bclink_unlock();
856 	nla_nest_end(msg->skb, attrs);
857 	genlmsg_end(msg->skb, hdr);
858 
859 	return 0;
860 
861 prop_msg_full:
862 	nla_nest_cancel(msg->skb, prop);
863 attr_msg_full:
864 	nla_nest_cancel(msg->skb, attrs);
865 msg_full:
866 	tipc_bclink_unlock();
867 	genlmsg_cancel(msg->skb, hdr);
868 
869 	return -EMSGSIZE;
870 }
871 
872 int tipc_bclink_stats(char *buf, const u32 buf_size)
873 {
874 	int ret;
875 	struct tipc_stats *s;
876 
877 	if (!bcl)
878 		return 0;
879 
880 	tipc_bclink_lock();
881 
882 	s = &bcl->stats;
883 
884 	ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
885 			    "  Window:%u packets\n",
886 			    bcl->name, bcl->queue_limit[0]);
887 	ret += tipc_snprintf(buf + ret, buf_size - ret,
888 			     "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
889 			     s->recv_info, s->recv_fragments,
890 			     s->recv_fragmented, s->recv_bundles,
891 			     s->recv_bundled);
892 	ret += tipc_snprintf(buf + ret, buf_size - ret,
893 			     "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
894 			     s->sent_info, s->sent_fragments,
895 			     s->sent_fragmented, s->sent_bundles,
896 			     s->sent_bundled);
897 	ret += tipc_snprintf(buf + ret, buf_size - ret,
898 			     "  RX naks:%u defs:%u dups:%u\n",
899 			     s->recv_nacks, s->deferred_recv, s->duplicates);
900 	ret += tipc_snprintf(buf + ret, buf_size - ret,
901 			     "  TX naks:%u acks:%u dups:%u\n",
902 			     s->sent_nacks, s->sent_acks, s->retransmitted);
903 	ret += tipc_snprintf(buf + ret, buf_size - ret,
904 			     "  Congestion link:%u  Send queue max:%u avg:%u\n",
905 			     s->link_congs, s->max_queue_sz,
906 			     s->queue_sz_counts ?
907 			     (s->accu_queue_sz / s->queue_sz_counts) : 0);
908 
909 	tipc_bclink_unlock();
910 	return ret;
911 }
912 
913 int tipc_bclink_reset_stats(void)
914 {
915 	if (!bcl)
916 		return -ENOPROTOOPT;
917 
918 	tipc_bclink_lock();
919 	memset(&bcl->stats, 0, sizeof(bcl->stats));
920 	tipc_bclink_unlock();
921 	return 0;
922 }
923 
924 int tipc_bclink_set_queue_limits(u32 limit)
925 {
926 	if (!bcl)
927 		return -ENOPROTOOPT;
928 	if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
929 		return -EINVAL;
930 
931 	tipc_bclink_lock();
932 	tipc_link_set_queue_limits(bcl, limit);
933 	tipc_bclink_unlock();
934 	return 0;
935 }
936 
937 int tipc_bclink_init(void)
938 {
939 	bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
940 	if (!bcbearer)
941 		return -ENOMEM;
942 
943 	bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
944 	if (!bclink) {
945 		kfree(bcbearer);
946 		return -ENOMEM;
947 	}
948 
949 	bcl = &bclink->link;
950 	bcbearer->bearer.media = &bcbearer->media;
951 	bcbearer->media.send_msg = tipc_bcbearer_send;
952 	sprintf(bcbearer->media.name, "tipc-broadcast");
953 
954 	spin_lock_init(&bclink->lock);
955 	__skb_queue_head_init(&bcl->outqueue);
956 	__skb_queue_head_init(&bcl->deferred_queue);
957 	skb_queue_head_init(&bcl->waiting_sks);
958 	bcl->next_out_no = 1;
959 	spin_lock_init(&bclink->node.lock);
960 	__skb_queue_head_init(&bclink->node.waiting_sks);
961 	bcl->owner = &bclink->node;
962 	bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
963 	tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
964 	bcl->bearer_id = MAX_BEARERS;
965 	rcu_assign_pointer(bearer_list[MAX_BEARERS], &bcbearer->bearer);
966 	bcl->state = WORKING_WORKING;
967 	strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
968 	return 0;
969 }
970 
971 void tipc_bclink_stop(void)
972 {
973 	tipc_bclink_lock();
974 	tipc_link_purge_queues(bcl);
975 	tipc_bclink_unlock();
976 
977 	RCU_INIT_POINTER(bearer_list[BCBEARER], NULL);
978 	synchronize_net();
979 	kfree(bcbearer);
980 	kfree(bclink);
981 }
982 
983 /**
984  * tipc_nmap_add - add a node to a node map
985  */
986 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
987 {
988 	int n = tipc_node(node);
989 	int w = n / WSIZE;
990 	u32 mask = (1 << (n % WSIZE));
991 
992 	if ((nm_ptr->map[w] & mask) == 0) {
993 		nm_ptr->count++;
994 		nm_ptr->map[w] |= mask;
995 	}
996 }
997 
998 /**
999  * tipc_nmap_remove - remove a node from a node map
1000  */
1001 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
1002 {
1003 	int n = tipc_node(node);
1004 	int w = n / WSIZE;
1005 	u32 mask = (1 << (n % WSIZE));
1006 
1007 	if ((nm_ptr->map[w] & mask) != 0) {
1008 		nm_ptr->map[w] &= ~mask;
1009 		nm_ptr->count--;
1010 	}
1011 }
1012 
1013 /**
1014  * tipc_nmap_diff - find differences between node maps
1015  * @nm_a: input node map A
1016  * @nm_b: input node map B
1017  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
1018  */
1019 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
1020 			   struct tipc_node_map *nm_b,
1021 			   struct tipc_node_map *nm_diff)
1022 {
1023 	int stop = ARRAY_SIZE(nm_a->map);
1024 	int w;
1025 	int b;
1026 	u32 map;
1027 
1028 	memset(nm_diff, 0, sizeof(*nm_diff));
1029 	for (w = 0; w < stop; w++) {
1030 		map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
1031 		nm_diff->map[w] = map;
1032 		if (map != 0) {
1033 			for (b = 0 ; b < WSIZE; b++) {
1034 				if (map & (1 << b))
1035 					nm_diff->count++;
1036 			}
1037 		}
1038 	}
1039 }
1040 
1041 /**
1042  * tipc_port_list_add - add a port to a port list, ensuring no duplicates
1043  */
1044 void tipc_port_list_add(struct tipc_port_list *pl_ptr, u32 port)
1045 {
1046 	struct tipc_port_list *item = pl_ptr;
1047 	int i;
1048 	int item_sz = PLSIZE;
1049 	int cnt = pl_ptr->count;
1050 
1051 	for (; ; cnt -= item_sz, item = item->next) {
1052 		if (cnt < PLSIZE)
1053 			item_sz = cnt;
1054 		for (i = 0; i < item_sz; i++)
1055 			if (item->ports[i] == port)
1056 				return;
1057 		if (i < PLSIZE) {
1058 			item->ports[i] = port;
1059 			pl_ptr->count++;
1060 			return;
1061 		}
1062 		if (!item->next) {
1063 			item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
1064 			if (!item->next) {
1065 				pr_warn("Incomplete multicast delivery, no memory\n");
1066 				return;
1067 			}
1068 			item->next->next = NULL;
1069 		}
1070 	}
1071 }
1072 
1073 /**
1074  * tipc_port_list_free - free dynamically created entries in port_list chain
1075  *
1076  */
1077 void tipc_port_list_free(struct tipc_port_list *pl_ptr)
1078 {
1079 	struct tipc_port_list *item;
1080 	struct tipc_port_list *next;
1081 
1082 	for (item = pl_ptr->next; item; item = next) {
1083 		next = item->next;
1084 		kfree(item);
1085 	}
1086 }
1087