[TIPC]: Multicast link failure now resets all links to "nacking" node.
[linux-2.6.git] / net / tipc / bcast.c
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, 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 "msg.h"
40 #include "dbg.h"
41 #include "link.h"
42 #include "net.h"
43 #include "node.h"
44 #include "port.h"
45 #include "addr.h"
46 #include "node_subscr.h"
47 #include "name_distr.h"
48 #include "bearer.h"
49 #include "name_table.h"
50 #include "bcast.h"
51
52 #define MAX_PKT_DEFAULT_MCAST 1500      /* bcast link max packet size (fixed) */
53
54 #define BCLINK_WIN_DEFAULT 20           /* bcast link window size (default) */
55
56 #define BCLINK_LOG_BUF_SIZE 0
57
58 /*
59  * Loss rate for incoming broadcast frames; used to test retransmission code.
60  * Set to N to cause every N'th frame to be discarded; 0 => don't discard any.
61  */
62  
63 #define TIPC_BCAST_LOSS_RATE 0
64
65 /**
66  * struct bcbearer_pair - a pair of bearers used by broadcast link
67  * @primary: pointer to primary bearer
68  * @secondary: pointer to secondary bearer
69  * 
70  * Bearers must have same priority and same set of reachable destinations 
71  * to be paired.
72  */
73
74 struct bcbearer_pair {
75         struct bearer *primary;
76         struct bearer *secondary;
77 };
78
79 /**
80  * struct bcbearer - bearer used by broadcast link
81  * @bearer: (non-standard) broadcast bearer structure
82  * @media: (non-standard) broadcast media structure
83  * @bpairs: array of bearer pairs
84  * @bpairs_temp: array of bearer pairs used during creation of "bpairs"
85  */
86
87 struct bcbearer {
88         struct bearer bearer;
89         struct media media;
90         struct bcbearer_pair bpairs[MAX_BEARERS];
91         struct bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1];
92 };
93
94 /**
95  * struct bclink - link used for broadcast messages
96  * @link: (non-standard) broadcast link structure
97  * @node: (non-standard) node structure representing b'cast link's peer node
98  * 
99  * Handles sequence numbering, fragmentation, bundling, etc.
100  */
101
102 struct bclink {
103         struct link link;
104         struct node node;
105 };
106
107
108 static struct bcbearer *bcbearer = NULL;
109 static struct bclink *bclink = NULL;
110 static struct link *bcl = NULL;
111 static spinlock_t bc_lock = SPIN_LOCK_UNLOCKED;
112
113 char tipc_bclink_name[] = "multicast-link";
114
115
116 static u32 buf_seqno(struct sk_buff *buf)
117 {
118         return msg_seqno(buf_msg(buf));
119
120
121 static u32 bcbuf_acks(struct sk_buff *buf)
122 {
123         return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
124 }
125
126 static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
127 {
128         TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
129 }
130
131 static void bcbuf_decr_acks(struct sk_buff *buf)
132 {
133         bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
134 }
135
136
137 /** 
138  * bclink_set_gap - set gap according to contents of current deferred pkt queue
139  * 
140  * Called with 'node' locked, bc_lock unlocked
141  */
142
143 static void bclink_set_gap(struct node *n_ptr)
144 {
145         struct sk_buff *buf = n_ptr->bclink.deferred_head;
146
147         n_ptr->bclink.gap_after = n_ptr->bclink.gap_to =
148                 mod(n_ptr->bclink.last_in);
149         if (unlikely(buf != NULL))
150                 n_ptr->bclink.gap_to = mod(buf_seqno(buf) - 1);
151 }
152
153 /** 
154  * bclink_ack_allowed - test if ACK or NACK message can be sent at this moment
155  * 
156  * This mechanism endeavours to prevent all nodes in network from trying
157  * to ACK or NACK at the same time.
158  * 
159  * Note: TIPC uses a different trigger to distribute ACKs than it does to
160  *       distribute NACKs, but tries to use the same spacing (divide by 16). 
161  */
162
163 static int bclink_ack_allowed(u32 n)
164 {
165         return((n % TIPC_MIN_LINK_WIN) == tipc_own_tag);
166 }
167
168
169 /** 
170  * bclink_retransmit_pkt - retransmit broadcast packets
171  * @after: sequence number of last packet to *not* retransmit
172  * @to: sequence number of last packet to retransmit
173  * 
174  * Called with bc_lock locked
175  */
176
177 static void bclink_retransmit_pkt(u32 after, u32 to)
178 {
179         struct sk_buff *buf;
180
181         buf = bcl->first_out;
182         while (buf && less_eq(buf_seqno(buf), after)) {
183                 buf = buf->next;                
184         }
185         tipc_link_retransmit(bcl, buf, mod(to - after));
186 }
187
188 /** 
189  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
190  * @n_ptr: node that sent acknowledgement info
191  * @acked: broadcast sequence # that has been acknowledged
192  * 
193  * Node is locked, bc_lock unlocked.
194  */
195
196 void tipc_bclink_acknowledge(struct node *n_ptr, u32 acked)
197 {
198         struct sk_buff *crs;
199         struct sk_buff *next;
200         unsigned int released = 0;
201
202         if (less_eq(acked, n_ptr->bclink.acked))
203                 return;
204
205         spin_lock_bh(&bc_lock);
206
207         /* Skip over packets that node has previously acknowledged */
208
209         crs = bcl->first_out;
210         while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked)) {
211                 crs = crs->next;
212         }
213
214         /* Update packets that node is now acknowledging */
215
216         while (crs && less_eq(buf_seqno(crs), acked)) {
217                 next = crs->next;
218                 bcbuf_decr_acks(crs);
219                 if (bcbuf_acks(crs) == 0) {
220                         bcl->first_out = next;
221                         bcl->out_queue_size--;
222                         buf_discard(crs);
223                         released = 1;
224                 }
225                 crs = next;
226         }
227         n_ptr->bclink.acked = acked;
228
229         /* Try resolving broadcast link congestion, if necessary */
230
231         if (unlikely(bcl->next_out))
232                 tipc_link_push_queue(bcl);
233         if (unlikely(released && !list_empty(&bcl->waiting_ports)))
234                 tipc_link_wakeup_ports(bcl, 0);
235         spin_unlock_bh(&bc_lock);
236 }
237
238 /** 
239  * bclink_send_ack - unicast an ACK msg
240  * 
241  * tipc_net_lock and node lock set
242  */
243
244 static void bclink_send_ack(struct node *n_ptr)
245 {
246         struct link *l_ptr = n_ptr->active_links[n_ptr->addr & 1];
247
248         if (l_ptr != NULL)
249                 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
250 }
251
252 /** 
253  * bclink_send_nack- broadcast a NACK msg
254  * 
255  * tipc_net_lock and node lock set
256  */
257
258 static void bclink_send_nack(struct node *n_ptr)
259 {
260         struct sk_buff *buf;
261         struct tipc_msg *msg;
262
263         if (!less(n_ptr->bclink.gap_after, n_ptr->bclink.gap_to))
264                 return;
265
266         buf = buf_acquire(INT_H_SIZE);
267         if (buf) {
268                 msg = buf_msg(buf);
269                 msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
270                          TIPC_OK, INT_H_SIZE, n_ptr->addr);
271                 msg_set_mc_netid(msg, tipc_net_id);
272                 msg_set_bcast_ack(msg, mod(n_ptr->bclink.last_in)); 
273                 msg_set_bcgap_after(msg, n_ptr->bclink.gap_after);
274                 msg_set_bcgap_to(msg, n_ptr->bclink.gap_to);
275                 msg_set_bcast_tag(msg, tipc_own_tag);
276
277                 if (tipc_bearer_send(&bcbearer->bearer, buf, NULL)) {
278                         bcl->stats.sent_nacks++;
279                         buf_discard(buf);
280                 } else {
281                         tipc_bearer_schedule(bcl->b_ptr, bcl);
282                         bcl->proto_msg_queue = buf;
283                         bcl->stats.bearer_congs++;
284                 }
285
286                 /* 
287                  * Ensure we doesn't send another NACK msg to the node
288                  * until 16 more deferred messages arrive from it
289                  * (i.e. helps prevent all nodes from NACK'ing at same time)
290                  */
291                 
292                 n_ptr->bclink.nack_sync = tipc_own_tag;
293         }
294 }
295
296 /** 
297  * tipc_bclink_check_gap - send a NACK if a sequence gap exists
298  *
299  * tipc_net_lock and node lock set
300  */
301
302 void tipc_bclink_check_gap(struct node *n_ptr, u32 last_sent)
303 {
304         if (!n_ptr->bclink.supported ||
305             less_eq(last_sent, mod(n_ptr->bclink.last_in)))
306                 return;
307
308         bclink_set_gap(n_ptr);
309         if (n_ptr->bclink.gap_after == n_ptr->bclink.gap_to)
310                 n_ptr->bclink.gap_to = last_sent;
311         bclink_send_nack(n_ptr);
312 }
313
314 /** 
315  * tipc_bclink_peek_nack - process a NACK msg meant for another node
316  * 
317  * Only tipc_net_lock set.
318  */
319
320 static void tipc_bclink_peek_nack(u32 dest, u32 sender_tag, u32 gap_after, u32 gap_to)
321 {
322         struct node *n_ptr = tipc_node_find(dest);
323         u32 my_after, my_to;
324
325         if (unlikely(!n_ptr || !tipc_node_is_up(n_ptr)))
326                 return;
327         tipc_node_lock(n_ptr);
328         /*
329          * Modify gap to suppress unnecessary NACKs from this node
330          */
331         my_after = n_ptr->bclink.gap_after;
332         my_to = n_ptr->bclink.gap_to;
333
334         if (less_eq(gap_after, my_after)) {
335                 if (less(my_after, gap_to) && less(gap_to, my_to))
336                         n_ptr->bclink.gap_after = gap_to;
337                 else if (less_eq(my_to, gap_to))
338                         n_ptr->bclink.gap_to = n_ptr->bclink.gap_after;
339         } else if (less_eq(gap_after, my_to)) {
340                 if (less_eq(my_to, gap_to))
341                         n_ptr->bclink.gap_to = gap_after;
342         } else {
343                 /* 
344                  * Expand gap if missing bufs not in deferred queue:
345                  */
346                 struct sk_buff *buf = n_ptr->bclink.deferred_head;
347                 u32 prev = n_ptr->bclink.gap_to;
348
349                 for (; buf; buf = buf->next) {
350                         u32 seqno = buf_seqno(buf);
351
352                         if (mod(seqno - prev) != 1)
353                                 buf = NULL;
354                         if (seqno == gap_after)
355                                 break;
356                         prev = seqno;
357                 }
358                 if (buf == NULL)
359                         n_ptr->bclink.gap_to = gap_after;
360         }
361         /*
362          * Some nodes may send a complementary NACK now:
363          */ 
364         if (bclink_ack_allowed(sender_tag + 1)) {
365                 if (n_ptr->bclink.gap_to != n_ptr->bclink.gap_after) {
366                         bclink_send_nack(n_ptr);
367                         bclink_set_gap(n_ptr);
368                 }
369         }
370         tipc_node_unlock(n_ptr);
371 }
372
373 /**
374  * tipc_bclink_send_msg - broadcast a packet to all nodes in cluster
375  */
376
377 int tipc_bclink_send_msg(struct sk_buff *buf)
378 {
379         int res;
380
381         spin_lock_bh(&bc_lock);
382
383         res = tipc_link_send_buf(bcl, buf);
384         if (unlikely(res == -ELINKCONG))
385                 buf_discard(buf);
386         else
387                 bcl->stats.sent_info++;
388
389         if (bcl->out_queue_size > bcl->stats.max_queue_sz)
390                 bcl->stats.max_queue_sz = bcl->out_queue_size;
391         bcl->stats.queue_sz_counts++;
392         bcl->stats.accu_queue_sz += bcl->out_queue_size;
393
394         spin_unlock_bh(&bc_lock);
395         return res;
396 }
397
398 /**
399  * tipc_bclink_recv_pkt - receive a broadcast packet, and deliver upwards
400  * 
401  * tipc_net_lock is read_locked, no other locks set
402  */
403
404 void tipc_bclink_recv_pkt(struct sk_buff *buf)
405 {
406 #if (TIPC_BCAST_LOSS_RATE)
407         static int rx_count = 0;
408 #endif
409         struct tipc_msg *msg = buf_msg(buf);
410         struct node* node = tipc_node_find(msg_prevnode(msg));
411         u32 next_in;
412         u32 seqno;
413         struct sk_buff *deferred;
414
415         msg_dbg(msg, "<BC<<<");
416
417         if (unlikely(!node || !tipc_node_is_up(node) || !node->bclink.supported || 
418                      (msg_mc_netid(msg) != tipc_net_id))) {
419                 buf_discard(buf);
420                 return;
421         }
422
423         if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
424                 msg_dbg(msg, "<BCNACK<<<");
425                 if (msg_destnode(msg) == tipc_own_addr) {
426                         tipc_node_lock(node);
427                         tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
428                         tipc_node_unlock(node);
429                         spin_lock_bh(&bc_lock);
430                         bcl->stats.recv_nacks++;
431                         bcl->owner->next = node;   /* remember requestor */
432                         bclink_retransmit_pkt(msg_bcgap_after(msg),
433                                               msg_bcgap_to(msg));
434                         bcl->owner->next = NULL;
435                         spin_unlock_bh(&bc_lock);              
436                 } else {
437                         tipc_bclink_peek_nack(msg_destnode(msg),
438                                               msg_bcast_tag(msg),
439                                               msg_bcgap_after(msg),
440                                               msg_bcgap_to(msg));
441                 }
442                 buf_discard(buf);
443                 return;
444         }
445
446 #if (TIPC_BCAST_LOSS_RATE)
447         if (++rx_count == TIPC_BCAST_LOSS_RATE) {
448                 rx_count = 0;
449                 buf_discard(buf);
450                 return;
451         }
452 #endif
453
454         tipc_node_lock(node);
455 receive:
456         deferred = node->bclink.deferred_head;
457         next_in = mod(node->bclink.last_in + 1);
458         seqno = msg_seqno(msg);
459
460         if (likely(seqno == next_in)) {
461                 bcl->stats.recv_info++;
462                 node->bclink.last_in++;
463                 bclink_set_gap(node);
464                 if (unlikely(bclink_ack_allowed(seqno))) {
465                         bclink_send_ack(node);
466                         bcl->stats.sent_acks++;
467                 }
468                 if (likely(msg_isdata(msg))) {
469                         tipc_node_unlock(node);
470                         tipc_port_recv_mcast(buf, NULL);
471                 } else if (msg_user(msg) == MSG_BUNDLER) {
472                         bcl->stats.recv_bundles++;
473                         bcl->stats.recv_bundled += msg_msgcnt(msg);
474                         tipc_node_unlock(node);
475                         tipc_link_recv_bundle(buf);
476                 } else if (msg_user(msg) == MSG_FRAGMENTER) {
477                         bcl->stats.recv_fragments++;
478                         if (tipc_link_recv_fragment(&node->bclink.defragm,
479                                                     &buf, &msg))
480                                 bcl->stats.recv_fragmented++;
481                         tipc_node_unlock(node);
482                         tipc_net_route_msg(buf);
483                 } else {
484                         tipc_node_unlock(node);
485                         tipc_net_route_msg(buf);
486                 }
487                 if (deferred && (buf_seqno(deferred) == mod(next_in + 1))) {
488                         tipc_node_lock(node);
489                         buf = deferred;
490                         msg = buf_msg(buf);
491                         node->bclink.deferred_head = deferred->next;
492                         goto receive;
493                 }
494                 return;
495         } else if (less(next_in, seqno)) {
496                 u32 gap_after = node->bclink.gap_after;
497                 u32 gap_to = node->bclink.gap_to;
498
499                 if (tipc_link_defer_pkt(&node->bclink.deferred_head,
500                                         &node->bclink.deferred_tail,
501                                         buf)) {
502                         node->bclink.nack_sync++;
503                         bcl->stats.deferred_recv++;
504                         if (seqno == mod(gap_after + 1))
505                                 node->bclink.gap_after = seqno;
506                         else if (less(gap_after, seqno) && less(seqno, gap_to))
507                                 node->bclink.gap_to = seqno;
508                 }
509                 if (bclink_ack_allowed(node->bclink.nack_sync)) {
510                         if (gap_to != gap_after)
511                                 bclink_send_nack(node);
512                         bclink_set_gap(node);
513                 }
514         } else {
515                 bcl->stats.duplicates++;
516                 buf_discard(buf);
517         }
518         tipc_node_unlock(node);
519 }
520
521 u32 tipc_bclink_get_last_sent(void)
522 {
523         u32 last_sent = mod(bcl->next_out_no - 1);
524
525         if (bcl->next_out)
526                 last_sent = mod(buf_seqno(bcl->next_out) - 1);
527         return last_sent;
528 }
529
530 u32 tipc_bclink_acks_missing(struct node *n_ptr)
531 {
532         return (n_ptr->bclink.supported &&
533                 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
534 }
535
536
537 /**
538  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
539  * 
540  * Send through as many bearers as necessary to reach all nodes
541  * that support TIPC multicasting.
542  * 
543  * Returns 0 if packet sent successfully, non-zero if not
544  */
545
546 static int tipc_bcbearer_send(struct sk_buff *buf,
547                               struct tipc_bearer *unused1,
548                               struct tipc_media_addr *unused2)
549 {
550         static int send_count = 0;
551
552         struct node_map *remains;
553         struct node_map *remains_new;
554         struct node_map *remains_tmp;
555         int bp_index;
556         int swap_time;
557         int err;
558
559         /* Prepare buffer for broadcasting (if first time trying to send it) */
560
561         if (likely(!msg_non_seq(buf_msg(buf)))) {
562                 struct tipc_msg *msg;
563
564                 assert(tipc_cltr_bcast_nodes.count != 0);
565                 bcbuf_set_acks(buf, tipc_cltr_bcast_nodes.count);
566                 msg = buf_msg(buf);
567                 msg_set_non_seq(msg);
568                 msg_set_mc_netid(msg, tipc_net_id);
569         }
570
571         /* Determine if bearer pairs should be swapped following this attempt */
572
573         if ((swap_time = (++send_count >= 10)))
574                 send_count = 0;
575
576         /* Send buffer over bearers until all targets reached */
577         
578         remains = kmalloc(sizeof(struct node_map), GFP_ATOMIC);
579         remains_new = kmalloc(sizeof(struct node_map), GFP_ATOMIC);
580         *remains = tipc_cltr_bcast_nodes;
581
582         for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
583                 struct bearer *p = bcbearer->bpairs[bp_index].primary;
584                 struct bearer *s = bcbearer->bpairs[bp_index].secondary;
585
586                 if (!p)
587                         break;  /* no more bearers to try */
588
589                 tipc_nmap_diff(remains, &p->nodes, remains_new);
590                 if (remains_new->count == remains->count)
591                         continue;       /* bearer pair doesn't add anything */
592
593                 if (!p->publ.blocked &&
594                     !p->media->send_msg(buf, &p->publ, &p->media->bcast_addr)) {
595                         if (swap_time && s && !s->publ.blocked)
596                                 goto swap;
597                         else
598                                 goto update;
599                 }
600
601                 if (!s || s->publ.blocked ||
602                     s->media->send_msg(buf, &s->publ, &s->media->bcast_addr))
603                         continue;       /* unable to send using bearer pair */
604 swap:
605                 bcbearer->bpairs[bp_index].primary = s;
606                 bcbearer->bpairs[bp_index].secondary = p;
607 update:
608                 if (remains_new->count == 0) {
609                         err = TIPC_OK;
610                         goto out;
611                 }
612
613                 /* swap map */
614                 remains_tmp = remains;
615                 remains = remains_new;
616                 remains_new = remains_tmp;
617         }
618         
619         /* Unable to reach all targets */
620
621         bcbearer->bearer.publ.blocked = 1;
622         bcl->stats.bearer_congs++;
623         err = ~TIPC_OK;
624
625  out:
626         kfree(remains_new);
627         kfree(remains);
628         return err;
629 }
630
631 /**
632  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
633  */
634
635 void tipc_bcbearer_sort(void)
636 {
637         struct bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
638         struct bcbearer_pair *bp_curr;
639         int b_index;
640         int pri;
641
642         spin_lock_bh(&bc_lock);
643
644         /* Group bearers by priority (can assume max of two per priority) */
645
646         memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
647
648         for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
649                 struct bearer *b = &tipc_bearers[b_index];
650
651                 if (!b->active || !b->nodes.count)
652                         continue;
653
654                 if (!bp_temp[b->priority].primary)
655                         bp_temp[b->priority].primary = b;
656                 else
657                         bp_temp[b->priority].secondary = b;
658         }
659
660         /* Create array of bearer pairs for broadcasting */
661
662         bp_curr = bcbearer->bpairs;
663         memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
664
665         for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
666
667                 if (!bp_temp[pri].primary)
668                         continue;
669
670                 bp_curr->primary = bp_temp[pri].primary;
671
672                 if (bp_temp[pri].secondary) {
673                         if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
674                                             &bp_temp[pri].secondary->nodes)) {
675                                 bp_curr->secondary = bp_temp[pri].secondary;
676                         } else {
677                                 bp_curr++;
678                                 bp_curr->primary = bp_temp[pri].secondary;
679                         }
680                 }
681
682                 bp_curr++;
683         }
684
685         spin_unlock_bh(&bc_lock);
686 }
687
688 /**
689  * tipc_bcbearer_push - resolve bearer congestion
690  * 
691  * Forces bclink to push out any unsent packets, until all packets are gone
692  * or congestion reoccurs.
693  * No locks set when function called
694  */
695
696 void tipc_bcbearer_push(void)
697 {
698         struct bearer *b_ptr;
699
700         spin_lock_bh(&bc_lock);
701         b_ptr = &bcbearer->bearer;
702         if (b_ptr->publ.blocked) {
703                 b_ptr->publ.blocked = 0;
704                 tipc_bearer_lock_push(b_ptr);
705         }
706         spin_unlock_bh(&bc_lock);
707 }
708
709
710 int tipc_bclink_stats(char *buf, const u32 buf_size)
711 {
712         struct print_buf pb;
713
714         if (!bcl)
715                 return 0;
716
717         tipc_printbuf_init(&pb, buf, buf_size);
718
719         spin_lock_bh(&bc_lock);
720
721         tipc_printf(&pb, "Link <%s>\n"
722                          "  Window:%u packets\n", 
723                     bcl->name, bcl->queue_limit[0]);
724         tipc_printf(&pb, "  RX packets:%u fragments:%u/%u bundles:%u/%u\n", 
725                     bcl->stats.recv_info,
726                     bcl->stats.recv_fragments,
727                     bcl->stats.recv_fragmented,
728                     bcl->stats.recv_bundles,
729                     bcl->stats.recv_bundled);
730         tipc_printf(&pb, "  TX packets:%u fragments:%u/%u bundles:%u/%u\n", 
731                     bcl->stats.sent_info,
732                     bcl->stats.sent_fragments,
733                     bcl->stats.sent_fragmented, 
734                     bcl->stats.sent_bundles,
735                     bcl->stats.sent_bundled);
736         tipc_printf(&pb, "  RX naks:%u defs:%u dups:%u\n", 
737                     bcl->stats.recv_nacks,
738                     bcl->stats.deferred_recv, 
739                     bcl->stats.duplicates);
740         tipc_printf(&pb, "  TX naks:%u acks:%u dups:%u\n", 
741                     bcl->stats.sent_nacks, 
742                     bcl->stats.sent_acks, 
743                     bcl->stats.retransmitted);
744         tipc_printf(&pb, "  Congestion bearer:%u link:%u  Send queue max:%u avg:%u\n",
745                     bcl->stats.bearer_congs,
746                     bcl->stats.link_congs,
747                     bcl->stats.max_queue_sz,
748                     bcl->stats.queue_sz_counts
749                     ? (bcl->stats.accu_queue_sz / bcl->stats.queue_sz_counts)
750                     : 0);
751
752         spin_unlock_bh(&bc_lock);
753         return tipc_printbuf_validate(&pb);
754 }
755
756 int tipc_bclink_reset_stats(void)
757 {
758         if (!bcl)
759                 return -ENOPROTOOPT;
760
761         spin_lock_bh(&bc_lock);
762         memset(&bcl->stats, 0, sizeof(bcl->stats));
763         spin_unlock_bh(&bc_lock);
764         return TIPC_OK;
765 }
766
767 int tipc_bclink_set_queue_limits(u32 limit)
768 {
769         if (!bcl)
770                 return -ENOPROTOOPT;
771         if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
772                 return -EINVAL;
773
774         spin_lock_bh(&bc_lock);
775         tipc_link_set_queue_limits(bcl, limit);
776         spin_unlock_bh(&bc_lock);
777         return TIPC_OK;
778 }
779
780 int tipc_bclink_init(void)
781 {
782         bcbearer = kmalloc(sizeof(*bcbearer), GFP_ATOMIC);
783         bclink = kmalloc(sizeof(*bclink), GFP_ATOMIC);
784         if (!bcbearer || !bclink) {
785  nomem:
786                 warn("Memory squeeze; Failed to create multicast link\n");
787                 kfree(bcbearer);
788                 bcbearer = NULL;
789                 kfree(bclink);
790                 bclink = NULL;
791                 return -ENOMEM;
792         }
793
794         memset(bcbearer, 0, sizeof(struct bcbearer));
795         INIT_LIST_HEAD(&bcbearer->bearer.cong_links);
796         bcbearer->bearer.media = &bcbearer->media;
797         bcbearer->media.send_msg = tipc_bcbearer_send;
798         sprintf(bcbearer->media.name, "tipc-multicast");
799
800         bcl = &bclink->link;
801         memset(bclink, 0, sizeof(struct bclink));
802         INIT_LIST_HEAD(&bcl->waiting_ports);
803         bcl->next_out_no = 1;
804         bclink->node.lock =  SPIN_LOCK_UNLOCKED;        
805         bcl->owner = &bclink->node;
806         bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
807         tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
808         bcl->b_ptr = &bcbearer->bearer;
809         bcl->state = WORKING_WORKING;
810         sprintf(bcl->name, tipc_bclink_name);
811
812         if (BCLINK_LOG_BUF_SIZE) {
813                 char *pb = kmalloc(BCLINK_LOG_BUF_SIZE, GFP_ATOMIC);
814
815                 if (!pb)
816                         goto nomem;
817                 tipc_printbuf_init(&bcl->print_buf, pb, BCLINK_LOG_BUF_SIZE);
818         }
819
820         return TIPC_OK;
821 }
822
823 void tipc_bclink_stop(void)
824 {
825         spin_lock_bh(&bc_lock);
826         if (bcbearer) {
827                 tipc_link_stop(bcl);
828                 if (BCLINK_LOG_BUF_SIZE)
829                         kfree(bcl->print_buf.buf);
830                 bcl = NULL;
831                 kfree(bclink);
832                 bclink = NULL;
833                 kfree(bcbearer);
834                 bcbearer = NULL;
835         }
836         spin_unlock_bh(&bc_lock);
837 }
838