1 /*
2  * net/tipc/bcast.c: TIPC broadcast code
3  *
4  * Copyright (c) 2004-2006, 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 "port.h"
41 #include "bcast.h"
42 
43 #define MAX_PKT_DEFAULT_MCAST 1500	/* bcast link max packet size (fixed) */
44 
45 #define BCLINK_WIN_DEFAULT 20		/* bcast link window size (default) */
46 
47 /*
48  * Loss rate for incoming broadcast frames; used to test retransmission code.
49  * Set to N to cause every N'th frame to be discarded; 0 => don't discard any.
50  */
51 
52 #define TIPC_BCAST_LOSS_RATE 0
53 
54 /**
55  * struct bcbearer_pair - a pair of bearers used by broadcast link
56  * @primary: pointer to primary bearer
57  * @secondary: pointer to secondary bearer
58  *
59  * Bearers must have same priority and same set of reachable destinations
60  * to be paired.
61  */
62 
63 struct bcbearer_pair {
64 	struct tipc_bearer *primary;
65 	struct tipc_bearer *secondary;
66 };
67 
68 /**
69  * struct bcbearer - bearer used by broadcast link
70  * @bearer: (non-standard) broadcast bearer structure
71  * @media: (non-standard) broadcast media structure
72  * @bpairs: array of bearer pairs
73  * @bpairs_temp: temporary array of bearer pairs used by tipc_bcbearer_sort()
74  * @remains: temporary node map used by tipc_bcbearer_send()
75  * @remains_new: temporary node map used tipc_bcbearer_send()
76  *
77  * Note: The fields labelled "temporary" are incorporated into the bearer
78  * to avoid consuming potentially limited stack space through the use of
79  * large local variables within multicast routines.  Concurrent access is
80  * prevented through use of the spinlock "bc_lock".
81  */
82 
83 struct bcbearer {
84 	struct tipc_bearer bearer;
85 	struct media media;
86 	struct bcbearer_pair bpairs[MAX_BEARERS];
87 	struct bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1];
88 	struct tipc_node_map remains;
89 	struct tipc_node_map remains_new;
90 };
91 
92 /**
93  * struct bclink - link used for broadcast messages
94  * @link: (non-standard) broadcast link structure
95  * @node: (non-standard) node structure representing b'cast link's peer node
96  * @retransmit_to: node that most recently requested a retransmit
97  *
98  * Handles sequence numbering, fragmentation, bundling, etc.
99  */
100 
101 struct bclink {
102 	struct link link;
103 	struct tipc_node node;
104 	struct tipc_node *retransmit_to;
105 };
106 
107 
108 static struct bcbearer *bcbearer;
109 static struct bclink *bclink;
110 static struct link *bcl;
111 static DEFINE_SPINLOCK(bc_lock);
112 
113 /* broadcast-capable node map */
114 struct tipc_node_map tipc_bcast_nmap;
115 
116 const char tipc_bclink_name[] = "broadcast-link";
117 
118 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
119 			   struct tipc_node_map *nm_b,
120 			   struct tipc_node_map *nm_diff);
121 
buf_seqno(struct sk_buff * buf)122 static u32 buf_seqno(struct sk_buff *buf)
123 {
124 	return msg_seqno(buf_msg(buf));
125 }
126 
bcbuf_acks(struct sk_buff * buf)127 static u32 bcbuf_acks(struct sk_buff *buf)
128 {
129 	return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
130 }
131 
bcbuf_set_acks(struct sk_buff * buf,u32 acks)132 static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
133 {
134 	TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
135 }
136 
bcbuf_decr_acks(struct sk_buff * buf)137 static void bcbuf_decr_acks(struct sk_buff *buf)
138 {
139 	bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
140 }
141 
142 
bclink_set_last_sent(void)143 static void bclink_set_last_sent(void)
144 {
145 	if (bcl->next_out)
146 		bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
147 	else
148 		bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
149 }
150 
tipc_bclink_get_last_sent(void)151 u32 tipc_bclink_get_last_sent(void)
152 {
153 	return bcl->fsm_msg_cnt;
154 }
155 
156 /**
157  * bclink_set_gap - set gap according to contents of current deferred pkt queue
158  *
159  * Called with 'node' locked, bc_lock unlocked
160  */
161 
bclink_set_gap(struct tipc_node * n_ptr)162 static void bclink_set_gap(struct tipc_node *n_ptr)
163 {
164 	struct sk_buff *buf = n_ptr->bclink.deferred_head;
165 
166 	n_ptr->bclink.gap_after = n_ptr->bclink.gap_to =
167 		mod(n_ptr->bclink.last_in);
168 	if (unlikely(buf != NULL))
169 		n_ptr->bclink.gap_to = mod(buf_seqno(buf) - 1);
170 }
171 
172 /**
173  * bclink_ack_allowed - test if ACK or NACK message can be sent at this moment
174  *
175  * This mechanism endeavours to prevent all nodes in network from trying
176  * to ACK or NACK at the same time.
177  *
178  * Note: TIPC uses a different trigger to distribute ACKs than it does to
179  *       distribute NACKs, but tries to use the same spacing (divide by 16).
180  */
181 
bclink_ack_allowed(u32 n)182 static int bclink_ack_allowed(u32 n)
183 {
184 	return (n % TIPC_MIN_LINK_WIN) == tipc_own_tag;
185 }
186 
187 
188 /**
189  * tipc_bclink_retransmit_to - get most recent node to request retransmission
190  *
191  * Called with bc_lock locked
192  */
193 
tipc_bclink_retransmit_to(void)194 struct tipc_node *tipc_bclink_retransmit_to(void)
195 {
196 	return bclink->retransmit_to;
197 }
198 
199 /**
200  * bclink_retransmit_pkt - retransmit broadcast packets
201  * @after: sequence number of last packet to *not* retransmit
202  * @to: sequence number of last packet to retransmit
203  *
204  * Called with bc_lock locked
205  */
206 
bclink_retransmit_pkt(u32 after,u32 to)207 static void bclink_retransmit_pkt(u32 after, u32 to)
208 {
209 	struct sk_buff *buf;
210 
211 	buf = bcl->first_out;
212 	while (buf && less_eq(buf_seqno(buf), after))
213 		buf = buf->next;
214 	tipc_link_retransmit(bcl, buf, mod(to - after));
215 }
216 
217 /**
218  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
219  * @n_ptr: node that sent acknowledgement info
220  * @acked: broadcast sequence # that has been acknowledged
221  *
222  * Node is locked, bc_lock unlocked.
223  */
224 
tipc_bclink_acknowledge(struct tipc_node * n_ptr,u32 acked)225 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
226 {
227 	struct sk_buff *crs;
228 	struct sk_buff *next;
229 	unsigned int released = 0;
230 
231 	if (less_eq(acked, n_ptr->bclink.acked))
232 		return;
233 
234 	spin_lock_bh(&bc_lock);
235 
236 	/* Skip over packets that node has previously acknowledged */
237 
238 	crs = bcl->first_out;
239 	while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked))
240 		crs = crs->next;
241 
242 	/* Update packets that node is now acknowledging */
243 
244 	while (crs && less_eq(buf_seqno(crs), acked)) {
245 		next = crs->next;
246 		bcbuf_decr_acks(crs);
247 		if (bcbuf_acks(crs) == 0) {
248 			bcl->first_out = next;
249 			bcl->out_queue_size--;
250 			buf_discard(crs);
251 			released = 1;
252 		}
253 		crs = next;
254 	}
255 	n_ptr->bclink.acked = acked;
256 
257 	/* Try resolving broadcast link congestion, if necessary */
258 
259 	if (unlikely(bcl->next_out)) {
260 		tipc_link_push_queue(bcl);
261 		bclink_set_last_sent();
262 	}
263 	if (unlikely(released && !list_empty(&bcl->waiting_ports)))
264 		tipc_link_wakeup_ports(bcl, 0);
265 	spin_unlock_bh(&bc_lock);
266 }
267 
268 /**
269  * bclink_send_ack - unicast an ACK msg
270  *
271  * tipc_net_lock and node lock set
272  */
273 
bclink_send_ack(struct tipc_node * n_ptr)274 static void bclink_send_ack(struct tipc_node *n_ptr)
275 {
276 	struct link *l_ptr = n_ptr->active_links[n_ptr->addr & 1];
277 
278 	if (l_ptr != NULL)
279 		tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
280 }
281 
282 /**
283  * bclink_send_nack- broadcast a NACK msg
284  *
285  * tipc_net_lock and node lock set
286  */
287 
bclink_send_nack(struct tipc_node * n_ptr)288 static void bclink_send_nack(struct tipc_node *n_ptr)
289 {
290 	struct sk_buff *buf;
291 	struct tipc_msg *msg;
292 
293 	if (!less(n_ptr->bclink.gap_after, n_ptr->bclink.gap_to))
294 		return;
295 
296 	buf = tipc_buf_acquire(INT_H_SIZE);
297 	if (buf) {
298 		msg = buf_msg(buf);
299 		tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
300 			 INT_H_SIZE, n_ptr->addr);
301 		msg_set_non_seq(msg, 1);
302 		msg_set_mc_netid(msg, tipc_net_id);
303 		msg_set_bcast_ack(msg, mod(n_ptr->bclink.last_in));
304 		msg_set_bcgap_after(msg, n_ptr->bclink.gap_after);
305 		msg_set_bcgap_to(msg, n_ptr->bclink.gap_to);
306 		msg_set_bcast_tag(msg, tipc_own_tag);
307 
308 		if (tipc_bearer_send(&bcbearer->bearer, buf, NULL)) {
309 			bcl->stats.sent_nacks++;
310 			buf_discard(buf);
311 		} else {
312 			tipc_bearer_schedule(bcl->b_ptr, bcl);
313 			bcl->proto_msg_queue = buf;
314 			bcl->stats.bearer_congs++;
315 		}
316 
317 		/*
318 		 * Ensure we doesn't send another NACK msg to the node
319 		 * until 16 more deferred messages arrive from it
320 		 * (i.e. helps prevent all nodes from NACK'ing at same time)
321 		 */
322 
323 		n_ptr->bclink.nack_sync = tipc_own_tag;
324 	}
325 }
326 
327 /**
328  * tipc_bclink_check_gap - send a NACK if a sequence gap exists
329  *
330  * tipc_net_lock and node lock set
331  */
332 
tipc_bclink_check_gap(struct tipc_node * n_ptr,u32 last_sent)333 void tipc_bclink_check_gap(struct tipc_node *n_ptr, u32 last_sent)
334 {
335 	if (!n_ptr->bclink.supported ||
336 	    less_eq(last_sent, mod(n_ptr->bclink.last_in)))
337 		return;
338 
339 	bclink_set_gap(n_ptr);
340 	if (n_ptr->bclink.gap_after == n_ptr->bclink.gap_to)
341 		n_ptr->bclink.gap_to = last_sent;
342 	bclink_send_nack(n_ptr);
343 }
344 
345 /**
346  * tipc_bclink_peek_nack - process a NACK msg meant for another node
347  *
348  * Only tipc_net_lock set.
349  */
350 
tipc_bclink_peek_nack(u32 dest,u32 sender_tag,u32 gap_after,u32 gap_to)351 static void tipc_bclink_peek_nack(u32 dest, u32 sender_tag, u32 gap_after, u32 gap_to)
352 {
353 	struct tipc_node *n_ptr = tipc_node_find(dest);
354 	u32 my_after, my_to;
355 
356 	if (unlikely(!n_ptr || !tipc_node_is_up(n_ptr)))
357 		return;
358 	tipc_node_lock(n_ptr);
359 	/*
360 	 * Modify gap to suppress unnecessary NACKs from this node
361 	 */
362 	my_after = n_ptr->bclink.gap_after;
363 	my_to = n_ptr->bclink.gap_to;
364 
365 	if (less_eq(gap_after, my_after)) {
366 		if (less(my_after, gap_to) && less(gap_to, my_to))
367 			n_ptr->bclink.gap_after = gap_to;
368 		else if (less_eq(my_to, gap_to))
369 			n_ptr->bclink.gap_to = n_ptr->bclink.gap_after;
370 	} else if (less_eq(gap_after, my_to)) {
371 		if (less_eq(my_to, gap_to))
372 			n_ptr->bclink.gap_to = gap_after;
373 	} else {
374 		/*
375 		 * Expand gap if missing bufs not in deferred queue:
376 		 */
377 		struct sk_buff *buf = n_ptr->bclink.deferred_head;
378 		u32 prev = n_ptr->bclink.gap_to;
379 
380 		for (; buf; buf = buf->next) {
381 			u32 seqno = buf_seqno(buf);
382 
383 			if (mod(seqno - prev) != 1) {
384 				buf = NULL;
385 				break;
386 			}
387 			if (seqno == gap_after)
388 				break;
389 			prev = seqno;
390 		}
391 		if (buf == NULL)
392 			n_ptr->bclink.gap_to = gap_after;
393 	}
394 	/*
395 	 * Some nodes may send a complementary NACK now:
396 	 */
397 	if (bclink_ack_allowed(sender_tag + 1)) {
398 		if (n_ptr->bclink.gap_to != n_ptr->bclink.gap_after) {
399 			bclink_send_nack(n_ptr);
400 			bclink_set_gap(n_ptr);
401 		}
402 	}
403 	tipc_node_unlock(n_ptr);
404 }
405 
406 /**
407  * tipc_bclink_send_msg - broadcast a packet to all nodes in cluster
408  */
409 
tipc_bclink_send_msg(struct sk_buff * buf)410 int tipc_bclink_send_msg(struct sk_buff *buf)
411 {
412 	int res;
413 
414 	spin_lock_bh(&bc_lock);
415 
416 	res = tipc_link_send_buf(bcl, buf);
417 	if (unlikely(res == -ELINKCONG))
418 		buf_discard(buf);
419 	else
420 		bclink_set_last_sent();
421 
422 	bcl->stats.queue_sz_counts++;
423 	bcl->stats.accu_queue_sz += bcl->out_queue_size;
424 
425 	spin_unlock_bh(&bc_lock);
426 	return res;
427 }
428 
429 /**
430  * tipc_bclink_recv_pkt - receive a broadcast packet, and deliver upwards
431  *
432  * tipc_net_lock is read_locked, no other locks set
433  */
434 
tipc_bclink_recv_pkt(struct sk_buff * buf)435 void tipc_bclink_recv_pkt(struct sk_buff *buf)
436 {
437 #if (TIPC_BCAST_LOSS_RATE)
438 	static int rx_count;
439 #endif
440 	struct tipc_msg *msg = buf_msg(buf);
441 	struct tipc_node *node = tipc_node_find(msg_prevnode(msg));
442 	u32 next_in;
443 	u32 seqno;
444 	struct sk_buff *deferred;
445 
446 	if (unlikely(!node || !tipc_node_is_up(node) || !node->bclink.supported ||
447 		     (msg_mc_netid(msg) != tipc_net_id))) {
448 		buf_discard(buf);
449 		return;
450 	}
451 
452 	if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
453 		if (msg_destnode(msg) == tipc_own_addr) {
454 			tipc_node_lock(node);
455 			tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
456 			tipc_node_unlock(node);
457 			spin_lock_bh(&bc_lock);
458 			bcl->stats.recv_nacks++;
459 			bclink->retransmit_to = node;
460 			bclink_retransmit_pkt(msg_bcgap_after(msg),
461 					      msg_bcgap_to(msg));
462 			spin_unlock_bh(&bc_lock);
463 		} else {
464 			tipc_bclink_peek_nack(msg_destnode(msg),
465 					      msg_bcast_tag(msg),
466 					      msg_bcgap_after(msg),
467 					      msg_bcgap_to(msg));
468 		}
469 		buf_discard(buf);
470 		return;
471 	}
472 
473 #if (TIPC_BCAST_LOSS_RATE)
474 	if (++rx_count == TIPC_BCAST_LOSS_RATE) {
475 		rx_count = 0;
476 		buf_discard(buf);
477 		return;
478 	}
479 #endif
480 
481 	tipc_node_lock(node);
482 receive:
483 	deferred = node->bclink.deferred_head;
484 	next_in = mod(node->bclink.last_in + 1);
485 	seqno = msg_seqno(msg);
486 
487 	if (likely(seqno == next_in)) {
488 		bcl->stats.recv_info++;
489 		node->bclink.last_in++;
490 		bclink_set_gap(node);
491 		if (unlikely(bclink_ack_allowed(seqno))) {
492 			bclink_send_ack(node);
493 			bcl->stats.sent_acks++;
494 		}
495 		if (likely(msg_isdata(msg))) {
496 			tipc_node_unlock(node);
497 			tipc_port_recv_mcast(buf, NULL);
498 		} else if (msg_user(msg) == MSG_BUNDLER) {
499 			bcl->stats.recv_bundles++;
500 			bcl->stats.recv_bundled += msg_msgcnt(msg);
501 			tipc_node_unlock(node);
502 			tipc_link_recv_bundle(buf);
503 		} else if (msg_user(msg) == MSG_FRAGMENTER) {
504 			bcl->stats.recv_fragments++;
505 			if (tipc_link_recv_fragment(&node->bclink.defragm,
506 						    &buf, &msg))
507 				bcl->stats.recv_fragmented++;
508 			tipc_node_unlock(node);
509 			tipc_net_route_msg(buf);
510 		} else {
511 			tipc_node_unlock(node);
512 			tipc_net_route_msg(buf);
513 		}
514 		if (deferred && (buf_seqno(deferred) == mod(next_in + 1))) {
515 			tipc_node_lock(node);
516 			buf = deferred;
517 			msg = buf_msg(buf);
518 			node->bclink.deferred_head = deferred->next;
519 			goto receive;
520 		}
521 		return;
522 	} else if (less(next_in, seqno)) {
523 		u32 gap_after = node->bclink.gap_after;
524 		u32 gap_to = node->bclink.gap_to;
525 
526 		if (tipc_link_defer_pkt(&node->bclink.deferred_head,
527 					&node->bclink.deferred_tail,
528 					buf)) {
529 			node->bclink.nack_sync++;
530 			bcl->stats.deferred_recv++;
531 			if (seqno == mod(gap_after + 1))
532 				node->bclink.gap_after = seqno;
533 			else if (less(gap_after, seqno) && less(seqno, gap_to))
534 				node->bclink.gap_to = seqno;
535 		}
536 		if (bclink_ack_allowed(node->bclink.nack_sync)) {
537 			if (gap_to != gap_after)
538 				bclink_send_nack(node);
539 			bclink_set_gap(node);
540 		}
541 	} else {
542 		bcl->stats.duplicates++;
543 		buf_discard(buf);
544 	}
545 	tipc_node_unlock(node);
546 }
547 
tipc_bclink_acks_missing(struct tipc_node * n_ptr)548 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
549 {
550 	return (n_ptr->bclink.supported &&
551 		(tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
552 }
553 
554 
555 /**
556  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
557  *
558  * Send through as many bearers as necessary to reach all nodes
559  * that support TIPC multicasting.
560  *
561  * Returns 0 if packet sent successfully, non-zero if not
562  */
563 
tipc_bcbearer_send(struct sk_buff * buf,struct tipc_bearer * unused1,struct tipc_media_addr * unused2)564 static int tipc_bcbearer_send(struct sk_buff *buf,
565 			      struct tipc_bearer *unused1,
566 			      struct tipc_media_addr *unused2)
567 {
568 	int bp_index;
569 
570 	/* Prepare buffer for broadcasting (if first time trying to send it) */
571 
572 	if (likely(!msg_non_seq(buf_msg(buf)))) {
573 		struct tipc_msg *msg;
574 
575 		assert(tipc_bcast_nmap.count != 0);
576 		bcbuf_set_acks(buf, tipc_bcast_nmap.count);
577 		msg = buf_msg(buf);
578 		msg_set_non_seq(msg, 1);
579 		msg_set_mc_netid(msg, tipc_net_id);
580 		bcl->stats.sent_info++;
581 	}
582 
583 	/* Send buffer over bearers until all targets reached */
584 
585 	bcbearer->remains = tipc_bcast_nmap;
586 
587 	for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
588 		struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
589 		struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
590 
591 		if (!p)
592 			break;	/* no more bearers to try */
593 
594 		tipc_nmap_diff(&bcbearer->remains, &p->nodes, &bcbearer->remains_new);
595 		if (bcbearer->remains_new.count == bcbearer->remains.count)
596 			continue;	/* bearer pair doesn't add anything */
597 
598 		if (p->blocked ||
599 		    p->media->send_msg(buf, p, &p->media->bcast_addr)) {
600 			/* unable to send on primary bearer */
601 			if (!s || s->blocked ||
602 			    s->media->send_msg(buf, s,
603 					       &s->media->bcast_addr)) {
604 				/* unable to send on either bearer */
605 				continue;
606 			}
607 		}
608 
609 		if (s) {
610 			bcbearer->bpairs[bp_index].primary = s;
611 			bcbearer->bpairs[bp_index].secondary = p;
612 		}
613 
614 		if (bcbearer->remains_new.count == 0)
615 			return 0;
616 
617 		bcbearer->remains = bcbearer->remains_new;
618 	}
619 
620 	/*
621 	 * Unable to reach all targets (indicate success, since currently
622 	 * there isn't code in place to properly block & unblock the
623 	 * pseudo-bearer used by the broadcast link)
624 	 */
625 
626 	return TIPC_OK;
627 }
628 
629 /**
630  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
631  */
632 
tipc_bcbearer_sort(void)633 void tipc_bcbearer_sort(void)
634 {
635 	struct bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
636 	struct bcbearer_pair *bp_curr;
637 	int b_index;
638 	int pri;
639 
640 	spin_lock_bh(&bc_lock);
641 
642 	/* Group bearers by priority (can assume max of two per priority) */
643 
644 	memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
645 
646 	for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
647 		struct tipc_bearer *b = &tipc_bearers[b_index];
648 
649 		if (!b->active || !b->nodes.count)
650 			continue;
651 
652 		if (!bp_temp[b->priority].primary)
653 			bp_temp[b->priority].primary = b;
654 		else
655 			bp_temp[b->priority].secondary = b;
656 	}
657 
658 	/* Create array of bearer pairs for broadcasting */
659 
660 	bp_curr = bcbearer->bpairs;
661 	memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
662 
663 	for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
664 
665 		if (!bp_temp[pri].primary)
666 			continue;
667 
668 		bp_curr->primary = bp_temp[pri].primary;
669 
670 		if (bp_temp[pri].secondary) {
671 			if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
672 					    &bp_temp[pri].secondary->nodes)) {
673 				bp_curr->secondary = bp_temp[pri].secondary;
674 			} else {
675 				bp_curr++;
676 				bp_curr->primary = bp_temp[pri].secondary;
677 			}
678 		}
679 
680 		bp_curr++;
681 	}
682 
683 	spin_unlock_bh(&bc_lock);
684 }
685 
686 /**
687  * tipc_bcbearer_push - resolve bearer congestion
688  *
689  * Forces bclink to push out any unsent packets, until all packets are gone
690  * or congestion reoccurs.
691  * No locks set when function called
692  */
693 
tipc_bcbearer_push(void)694 void tipc_bcbearer_push(void)
695 {
696 	struct tipc_bearer *b_ptr;
697 
698 	spin_lock_bh(&bc_lock);
699 	b_ptr = &bcbearer->bearer;
700 	if (b_ptr->blocked) {
701 		b_ptr->blocked = 0;
702 		tipc_bearer_lock_push(b_ptr);
703 	}
704 	spin_unlock_bh(&bc_lock);
705 }
706 
707 
tipc_bclink_stats(char * buf,const u32 buf_size)708 int tipc_bclink_stats(char *buf, const u32 buf_size)
709 {
710 	struct print_buf pb;
711 
712 	if (!bcl)
713 		return 0;
714 
715 	tipc_printbuf_init(&pb, buf, buf_size);
716 
717 	spin_lock_bh(&bc_lock);
718 
719 	tipc_printf(&pb, "Link <%s>\n"
720 			 "  Window:%u packets\n",
721 		    bcl->name, bcl->queue_limit[0]);
722 	tipc_printf(&pb, "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
723 		    bcl->stats.recv_info,
724 		    bcl->stats.recv_fragments,
725 		    bcl->stats.recv_fragmented,
726 		    bcl->stats.recv_bundles,
727 		    bcl->stats.recv_bundled);
728 	tipc_printf(&pb, "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
729 		    bcl->stats.sent_info,
730 		    bcl->stats.sent_fragments,
731 		    bcl->stats.sent_fragmented,
732 		    bcl->stats.sent_bundles,
733 		    bcl->stats.sent_bundled);
734 	tipc_printf(&pb, "  RX naks:%u defs:%u dups:%u\n",
735 		    bcl->stats.recv_nacks,
736 		    bcl->stats.deferred_recv,
737 		    bcl->stats.duplicates);
738 	tipc_printf(&pb, "  TX naks:%u acks:%u dups:%u\n",
739 		    bcl->stats.sent_nacks,
740 		    bcl->stats.sent_acks,
741 		    bcl->stats.retransmitted);
742 	tipc_printf(&pb, "  Congestion bearer:%u link:%u  Send queue max:%u avg:%u\n",
743 		    bcl->stats.bearer_congs,
744 		    bcl->stats.link_congs,
745 		    bcl->stats.max_queue_sz,
746 		    bcl->stats.queue_sz_counts
747 		    ? (bcl->stats.accu_queue_sz / bcl->stats.queue_sz_counts)
748 		    : 0);
749 
750 	spin_unlock_bh(&bc_lock);
751 	return tipc_printbuf_validate(&pb);
752 }
753 
tipc_bclink_reset_stats(void)754 int tipc_bclink_reset_stats(void)
755 {
756 	if (!bcl)
757 		return -ENOPROTOOPT;
758 
759 	spin_lock_bh(&bc_lock);
760 	memset(&bcl->stats, 0, sizeof(bcl->stats));
761 	spin_unlock_bh(&bc_lock);
762 	return 0;
763 }
764 
tipc_bclink_set_queue_limits(u32 limit)765 int tipc_bclink_set_queue_limits(u32 limit)
766 {
767 	if (!bcl)
768 		return -ENOPROTOOPT;
769 	if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
770 		return -EINVAL;
771 
772 	spin_lock_bh(&bc_lock);
773 	tipc_link_set_queue_limits(bcl, limit);
774 	spin_unlock_bh(&bc_lock);
775 	return 0;
776 }
777 
tipc_bclink_init(void)778 int tipc_bclink_init(void)
779 {
780 	bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
781 	bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
782 	if (!bcbearer || !bclink) {
783 		warn("Multicast link creation failed, no memory\n");
784 		kfree(bcbearer);
785 		bcbearer = NULL;
786 		kfree(bclink);
787 		bclink = NULL;
788 		return -ENOMEM;
789 	}
790 
791 	INIT_LIST_HEAD(&bcbearer->bearer.cong_links);
792 	bcbearer->bearer.media = &bcbearer->media;
793 	bcbearer->media.send_msg = tipc_bcbearer_send;
794 	sprintf(bcbearer->media.name, "tipc-multicast");
795 
796 	bcl = &bclink->link;
797 	INIT_LIST_HEAD(&bcl->waiting_ports);
798 	bcl->next_out_no = 1;
799 	spin_lock_init(&bclink->node.lock);
800 	bcl->owner = &bclink->node;
801 	bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
802 	tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
803 	bcl->b_ptr = &bcbearer->bearer;
804 	bcl->state = WORKING_WORKING;
805 	strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
806 
807 	return 0;
808 }
809 
tipc_bclink_stop(void)810 void tipc_bclink_stop(void)
811 {
812 	spin_lock_bh(&bc_lock);
813 	if (bcbearer) {
814 		tipc_link_stop(bcl);
815 		bcl = NULL;
816 		kfree(bclink);
817 		bclink = NULL;
818 		kfree(bcbearer);
819 		bcbearer = NULL;
820 	}
821 	spin_unlock_bh(&bc_lock);
822 }
823 
824 
825 /**
826  * tipc_nmap_add - add a node to a node map
827  */
828 
tipc_nmap_add(struct tipc_node_map * nm_ptr,u32 node)829 void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
830 {
831 	int n = tipc_node(node);
832 	int w = n / WSIZE;
833 	u32 mask = (1 << (n % WSIZE));
834 
835 	if ((nm_ptr->map[w] & mask) == 0) {
836 		nm_ptr->count++;
837 		nm_ptr->map[w] |= mask;
838 	}
839 }
840 
841 /**
842  * tipc_nmap_remove - remove a node from a node map
843  */
844 
tipc_nmap_remove(struct tipc_node_map * nm_ptr,u32 node)845 void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
846 {
847 	int n = tipc_node(node);
848 	int w = n / WSIZE;
849 	u32 mask = (1 << (n % WSIZE));
850 
851 	if ((nm_ptr->map[w] & mask) != 0) {
852 		nm_ptr->map[w] &= ~mask;
853 		nm_ptr->count--;
854 	}
855 }
856 
857 /**
858  * tipc_nmap_diff - find differences between node maps
859  * @nm_a: input node map A
860  * @nm_b: input node map B
861  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
862  */
863 
tipc_nmap_diff(struct tipc_node_map * nm_a,struct tipc_node_map * nm_b,struct tipc_node_map * nm_diff)864 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
865 			   struct tipc_node_map *nm_b,
866 			   struct tipc_node_map *nm_diff)
867 {
868 	int stop = ARRAY_SIZE(nm_a->map);
869 	int w;
870 	int b;
871 	u32 map;
872 
873 	memset(nm_diff, 0, sizeof(*nm_diff));
874 	for (w = 0; w < stop; w++) {
875 		map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
876 		nm_diff->map[w] = map;
877 		if (map != 0) {
878 			for (b = 0 ; b < WSIZE; b++) {
879 				if (map & (1 << b))
880 					nm_diff->count++;
881 			}
882 		}
883 	}
884 }
885 
886 /**
887  * tipc_port_list_add - add a port to a port list, ensuring no duplicates
888  */
889 
tipc_port_list_add(struct port_list * pl_ptr,u32 port)890 void tipc_port_list_add(struct port_list *pl_ptr, u32 port)
891 {
892 	struct port_list *item = pl_ptr;
893 	int i;
894 	int item_sz = PLSIZE;
895 	int cnt = pl_ptr->count;
896 
897 	for (; ; cnt -= item_sz, item = item->next) {
898 		if (cnt < PLSIZE)
899 			item_sz = cnt;
900 		for (i = 0; i < item_sz; i++)
901 			if (item->ports[i] == port)
902 				return;
903 		if (i < PLSIZE) {
904 			item->ports[i] = port;
905 			pl_ptr->count++;
906 			return;
907 		}
908 		if (!item->next) {
909 			item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
910 			if (!item->next) {
911 				warn("Incomplete multicast delivery, no memory\n");
912 				return;
913 			}
914 			item->next->next = NULL;
915 		}
916 	}
917 }
918 
919 /**
920  * tipc_port_list_free - free dynamically created entries in port_list chain
921  *
922  */
923 
tipc_port_list_free(struct port_list * pl_ptr)924 void tipc_port_list_free(struct port_list *pl_ptr)
925 {
926 	struct port_list *item;
927 	struct port_list *next;
928 
929 	for (item = pl_ptr->next; item; item = next) {
930 		next = item->next;
931 		kfree(item);
932 	}
933 }
934 
935