]> git.karo-electronics.de Git - karo-tx-linux.git/blob - net/tipc/bcast.c
tipc: Improve accuracy of link transmit queue maximum size statistic
[karo-tx-linux.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, 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
122 static u32 buf_seqno(struct sk_buff *buf)
123 {
124         return msg_seqno(buf_msg(buf));
125 }
126
127 static u32 bcbuf_acks(struct sk_buff *buf)
128 {
129         return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
130 }
131
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
137 static void bcbuf_decr_acks(struct sk_buff *buf)
138 {
139         bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
140 }
141
142
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
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
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
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
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
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
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
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
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_mc_netid(msg, tipc_net_id);
302                 msg_set_bcast_ack(msg, mod(n_ptr->bclink.last_in));
303                 msg_set_bcgap_after(msg, n_ptr->bclink.gap_after);
304                 msg_set_bcgap_to(msg, n_ptr->bclink.gap_to);
305                 msg_set_bcast_tag(msg, tipc_own_tag);
306
307                 if (tipc_bearer_send(&bcbearer->bearer, buf, NULL)) {
308                         bcl->stats.sent_nacks++;
309                         buf_discard(buf);
310                 } else {
311                         tipc_bearer_schedule(bcl->b_ptr, bcl);
312                         bcl->proto_msg_queue = buf;
313                         bcl->stats.bearer_congs++;
314                 }
315
316                 /*
317                  * Ensure we doesn't send another NACK msg to the node
318                  * until 16 more deferred messages arrive from it
319                  * (i.e. helps prevent all nodes from NACK'ing at same time)
320                  */
321
322                 n_ptr->bclink.nack_sync = tipc_own_tag;
323         }
324 }
325
326 /**
327  * tipc_bclink_check_gap - send a NACK if a sequence gap exists
328  *
329  * tipc_net_lock and node lock set
330  */
331
332 void tipc_bclink_check_gap(struct tipc_node *n_ptr, u32 last_sent)
333 {
334         if (!n_ptr->bclink.supported ||
335             less_eq(last_sent, mod(n_ptr->bclink.last_in)))
336                 return;
337
338         bclink_set_gap(n_ptr);
339         if (n_ptr->bclink.gap_after == n_ptr->bclink.gap_to)
340                 n_ptr->bclink.gap_to = last_sent;
341         bclink_send_nack(n_ptr);
342 }
343
344 /**
345  * tipc_bclink_peek_nack - process a NACK msg meant for another node
346  *
347  * Only tipc_net_lock set.
348  */
349
350 static void tipc_bclink_peek_nack(u32 dest, u32 sender_tag, u32 gap_after, u32 gap_to)
351 {
352         struct tipc_node *n_ptr = tipc_node_find(dest);
353         u32 my_after, my_to;
354
355         if (unlikely(!n_ptr || !tipc_node_is_up(n_ptr)))
356                 return;
357         tipc_node_lock(n_ptr);
358         /*
359          * Modify gap to suppress unnecessary NACKs from this node
360          */
361         my_after = n_ptr->bclink.gap_after;
362         my_to = n_ptr->bclink.gap_to;
363
364         if (less_eq(gap_after, my_after)) {
365                 if (less(my_after, gap_to) && less(gap_to, my_to))
366                         n_ptr->bclink.gap_after = gap_to;
367                 else if (less_eq(my_to, gap_to))
368                         n_ptr->bclink.gap_to = n_ptr->bclink.gap_after;
369         } else if (less_eq(gap_after, my_to)) {
370                 if (less_eq(my_to, gap_to))
371                         n_ptr->bclink.gap_to = gap_after;
372         } else {
373                 /*
374                  * Expand gap if missing bufs not in deferred queue:
375                  */
376                 struct sk_buff *buf = n_ptr->bclink.deferred_head;
377                 u32 prev = n_ptr->bclink.gap_to;
378
379                 for (; buf; buf = buf->next) {
380                         u32 seqno = buf_seqno(buf);
381
382                         if (mod(seqno - prev) != 1) {
383                                 buf = NULL;
384                                 break;
385                         }
386                         if (seqno == gap_after)
387                                 break;
388                         prev = seqno;
389                 }
390                 if (buf == NULL)
391                         n_ptr->bclink.gap_to = gap_after;
392         }
393         /*
394          * Some nodes may send a complementary NACK now:
395          */
396         if (bclink_ack_allowed(sender_tag + 1)) {
397                 if (n_ptr->bclink.gap_to != n_ptr->bclink.gap_after) {
398                         bclink_send_nack(n_ptr);
399                         bclink_set_gap(n_ptr);
400                 }
401         }
402         tipc_node_unlock(n_ptr);
403 }
404
405 /**
406  * tipc_bclink_send_msg - broadcast a packet to all nodes in cluster
407  */
408
409 int tipc_bclink_send_msg(struct sk_buff *buf)
410 {
411         int res;
412
413         spin_lock_bh(&bc_lock);
414
415         res = tipc_link_send_buf(bcl, buf);
416         if (unlikely(res == -ELINKCONG))
417                 buf_discard(buf);
418         else
419                 bclink_set_last_sent();
420
421         bcl->stats.queue_sz_counts++;
422         bcl->stats.accu_queue_sz += bcl->out_queue_size;
423
424         spin_unlock_bh(&bc_lock);
425         return res;
426 }
427
428 /**
429  * tipc_bclink_recv_pkt - receive a broadcast packet, and deliver upwards
430  *
431  * tipc_net_lock is read_locked, no other locks set
432  */
433
434 void tipc_bclink_recv_pkt(struct sk_buff *buf)
435 {
436 #if (TIPC_BCAST_LOSS_RATE)
437         static int rx_count;
438 #endif
439         struct tipc_msg *msg = buf_msg(buf);
440         struct tipc_node *node = tipc_node_find(msg_prevnode(msg));
441         u32 next_in;
442         u32 seqno;
443         struct sk_buff *deferred;
444
445         if (unlikely(!node || !tipc_node_is_up(node) || !node->bclink.supported ||
446                      (msg_mc_netid(msg) != tipc_net_id))) {
447                 buf_discard(buf);
448                 return;
449         }
450
451         if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
452                 if (msg_destnode(msg) == tipc_own_addr) {
453                         tipc_node_lock(node);
454                         tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
455                         tipc_node_unlock(node);
456                         spin_lock_bh(&bc_lock);
457                         bcl->stats.recv_nacks++;
458                         bclink->retransmit_to = node;
459                         bclink_retransmit_pkt(msg_bcgap_after(msg),
460                                               msg_bcgap_to(msg));
461                         spin_unlock_bh(&bc_lock);
462                 } else {
463                         tipc_bclink_peek_nack(msg_destnode(msg),
464                                               msg_bcast_tag(msg),
465                                               msg_bcgap_after(msg),
466                                               msg_bcgap_to(msg));
467                 }
468                 buf_discard(buf);
469                 return;
470         }
471
472 #if (TIPC_BCAST_LOSS_RATE)
473         if (++rx_count == TIPC_BCAST_LOSS_RATE) {
474                 rx_count = 0;
475                 buf_discard(buf);
476                 return;
477         }
478 #endif
479
480         tipc_node_lock(node);
481 receive:
482         deferred = node->bclink.deferred_head;
483         next_in = mod(node->bclink.last_in + 1);
484         seqno = msg_seqno(msg);
485
486         if (likely(seqno == next_in)) {
487                 bcl->stats.recv_info++;
488                 node->bclink.last_in++;
489                 bclink_set_gap(node);
490                 if (unlikely(bclink_ack_allowed(seqno))) {
491                         bclink_send_ack(node);
492                         bcl->stats.sent_acks++;
493                 }
494                 if (likely(msg_isdata(msg))) {
495                         tipc_node_unlock(node);
496                         tipc_port_recv_mcast(buf, NULL);
497                 } else if (msg_user(msg) == MSG_BUNDLER) {
498                         bcl->stats.recv_bundles++;
499                         bcl->stats.recv_bundled += msg_msgcnt(msg);
500                         tipc_node_unlock(node);
501                         tipc_link_recv_bundle(buf);
502                 } else if (msg_user(msg) == MSG_FRAGMENTER) {
503                         bcl->stats.recv_fragments++;
504                         if (tipc_link_recv_fragment(&node->bclink.defragm,
505                                                     &buf, &msg))
506                                 bcl->stats.recv_fragmented++;
507                         tipc_node_unlock(node);
508                         tipc_net_route_msg(buf);
509                 } else {
510                         tipc_node_unlock(node);
511                         tipc_net_route_msg(buf);
512                 }
513                 if (deferred && (buf_seqno(deferred) == mod(next_in + 1))) {
514                         tipc_node_lock(node);
515                         buf = deferred;
516                         msg = buf_msg(buf);
517                         node->bclink.deferred_head = deferred->next;
518                         goto receive;
519                 }
520                 return;
521         } else if (less(next_in, seqno)) {
522                 u32 gap_after = node->bclink.gap_after;
523                 u32 gap_to = node->bclink.gap_to;
524
525                 if (tipc_link_defer_pkt(&node->bclink.deferred_head,
526                                         &node->bclink.deferred_tail,
527                                         buf)) {
528                         node->bclink.nack_sync++;
529                         bcl->stats.deferred_recv++;
530                         if (seqno == mod(gap_after + 1))
531                                 node->bclink.gap_after = seqno;
532                         else if (less(gap_after, seqno) && less(seqno, gap_to))
533                                 node->bclink.gap_to = seqno;
534                 }
535                 if (bclink_ack_allowed(node->bclink.nack_sync)) {
536                         if (gap_to != gap_after)
537                                 bclink_send_nack(node);
538                         bclink_set_gap(node);
539                 }
540         } else {
541                 bcl->stats.duplicates++;
542                 buf_discard(buf);
543         }
544         tipc_node_unlock(node);
545 }
546
547 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
548 {
549         return (n_ptr->bclink.supported &&
550                 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
551 }
552
553
554 /**
555  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
556  *
557  * Send through as many bearers as necessary to reach all nodes
558  * that support TIPC multicasting.
559  *
560  * Returns 0 if packet sent successfully, non-zero if not
561  */
562
563 static int tipc_bcbearer_send(struct sk_buff *buf,
564                               struct tipc_bearer *unused1,
565                               struct tipc_media_addr *unused2)
566 {
567         int bp_index;
568
569         /* Prepare buffer for broadcasting (if first time trying to send it) */
570
571         if (likely(!msg_non_seq(buf_msg(buf)))) {
572                 struct tipc_msg *msg;
573
574                 assert(tipc_bcast_nmap.count != 0);
575                 bcbuf_set_acks(buf, tipc_bcast_nmap.count);
576                 msg = buf_msg(buf);
577                 msg_set_non_seq(msg, 1);
578                 msg_set_mc_netid(msg, tipc_net_id);
579                 bcl->stats.sent_info++;
580         }
581
582         /* Send buffer over bearers until all targets reached */
583
584         bcbearer->remains = tipc_bcast_nmap;
585
586         for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
587                 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
588                 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
589
590                 if (!p)
591                         break;  /* no more bearers to try */
592
593                 tipc_nmap_diff(&bcbearer->remains, &p->nodes, &bcbearer->remains_new);
594                 if (bcbearer->remains_new.count == bcbearer->remains.count)
595                         continue;       /* bearer pair doesn't add anything */
596
597                 if (p->blocked ||
598                     p->media->send_msg(buf, p, &p->media->bcast_addr)) {
599                         /* unable to send on primary bearer */
600                         if (!s || s->blocked ||
601                             s->media->send_msg(buf, s,
602                                                &s->media->bcast_addr)) {
603                                 /* unable to send on either bearer */
604                                 continue;
605                         }
606                 }
607
608                 if (s) {
609                         bcbearer->bpairs[bp_index].primary = s;
610                         bcbearer->bpairs[bp_index].secondary = p;
611                 }
612
613                 if (bcbearer->remains_new.count == 0)
614                         return 0;
615
616                 bcbearer->remains = bcbearer->remains_new;
617         }
618
619         /*
620          * Unable to reach all targets (indicate success, since currently
621          * there isn't code in place to properly block & unblock the
622          * pseudo-bearer used by the broadcast link)
623          */
624
625         return TIPC_OK;
626 }
627
628 /**
629  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
630  */
631
632 void tipc_bcbearer_sort(void)
633 {
634         struct bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
635         struct bcbearer_pair *bp_curr;
636         int b_index;
637         int pri;
638
639         spin_lock_bh(&bc_lock);
640
641         /* Group bearers by priority (can assume max of two per priority) */
642
643         memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
644
645         for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
646                 struct tipc_bearer *b = &tipc_bearers[b_index];
647
648                 if (!b->active || !b->nodes.count)
649                         continue;
650
651                 if (!bp_temp[b->priority].primary)
652                         bp_temp[b->priority].primary = b;
653                 else
654                         bp_temp[b->priority].secondary = b;
655         }
656
657         /* Create array of bearer pairs for broadcasting */
658
659         bp_curr = bcbearer->bpairs;
660         memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
661
662         for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
663
664                 if (!bp_temp[pri].primary)
665                         continue;
666
667                 bp_curr->primary = bp_temp[pri].primary;
668
669                 if (bp_temp[pri].secondary) {
670                         if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
671                                             &bp_temp[pri].secondary->nodes)) {
672                                 bp_curr->secondary = bp_temp[pri].secondary;
673                         } else {
674                                 bp_curr++;
675                                 bp_curr->primary = bp_temp[pri].secondary;
676                         }
677                 }
678
679                 bp_curr++;
680         }
681
682         spin_unlock_bh(&bc_lock);
683 }
684
685 /**
686  * tipc_bcbearer_push - resolve bearer congestion
687  *
688  * Forces bclink to push out any unsent packets, until all packets are gone
689  * or congestion reoccurs.
690  * No locks set when function called
691  */
692
693 void tipc_bcbearer_push(void)
694 {
695         struct tipc_bearer *b_ptr;
696
697         spin_lock_bh(&bc_lock);
698         b_ptr = &bcbearer->bearer;
699         if (b_ptr->blocked) {
700                 b_ptr->blocked = 0;
701                 tipc_bearer_lock_push(b_ptr);
702         }
703         spin_unlock_bh(&bc_lock);
704 }
705
706
707 int tipc_bclink_stats(char *buf, const u32 buf_size)
708 {
709         struct print_buf pb;
710
711         if (!bcl)
712                 return 0;
713
714         tipc_printbuf_init(&pb, buf, buf_size);
715
716         spin_lock_bh(&bc_lock);
717
718         tipc_printf(&pb, "Link <%s>\n"
719                          "  Window:%u packets\n",
720                     bcl->name, bcl->queue_limit[0]);
721         tipc_printf(&pb, "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
722                     bcl->stats.recv_info,
723                     bcl->stats.recv_fragments,
724                     bcl->stats.recv_fragmented,
725                     bcl->stats.recv_bundles,
726                     bcl->stats.recv_bundled);
727         tipc_printf(&pb, "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
728                     bcl->stats.sent_info,
729                     bcl->stats.sent_fragments,
730                     bcl->stats.sent_fragmented,
731                     bcl->stats.sent_bundles,
732                     bcl->stats.sent_bundled);
733         tipc_printf(&pb, "  RX naks:%u defs:%u dups:%u\n",
734                     bcl->stats.recv_nacks,
735                     bcl->stats.deferred_recv,
736                     bcl->stats.duplicates);
737         tipc_printf(&pb, "  TX naks:%u acks:%u dups:%u\n",
738                     bcl->stats.sent_nacks,
739                     bcl->stats.sent_acks,
740                     bcl->stats.retransmitted);
741         tipc_printf(&pb, "  Congestion bearer:%u link:%u  Send queue max:%u avg:%u\n",
742                     bcl->stats.bearer_congs,
743                     bcl->stats.link_congs,
744                     bcl->stats.max_queue_sz,
745                     bcl->stats.queue_sz_counts
746                     ? (bcl->stats.accu_queue_sz / bcl->stats.queue_sz_counts)
747                     : 0);
748
749         spin_unlock_bh(&bc_lock);
750         return tipc_printbuf_validate(&pb);
751 }
752
753 int tipc_bclink_reset_stats(void)
754 {
755         if (!bcl)
756                 return -ENOPROTOOPT;
757
758         spin_lock_bh(&bc_lock);
759         memset(&bcl->stats, 0, sizeof(bcl->stats));
760         spin_unlock_bh(&bc_lock);
761         return 0;
762 }
763
764 int tipc_bclink_set_queue_limits(u32 limit)
765 {
766         if (!bcl)
767                 return -ENOPROTOOPT;
768         if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
769                 return -EINVAL;
770
771         spin_lock_bh(&bc_lock);
772         tipc_link_set_queue_limits(bcl, limit);
773         spin_unlock_bh(&bc_lock);
774         return 0;
775 }
776
777 int tipc_bclink_init(void)
778 {
779         bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
780         bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
781         if (!bcbearer || !bclink) {
782                 warn("Multicast link creation failed, no memory\n");
783                 kfree(bcbearer);
784                 bcbearer = NULL;
785                 kfree(bclink);
786                 bclink = NULL;
787                 return -ENOMEM;
788         }
789
790         INIT_LIST_HEAD(&bcbearer->bearer.cong_links);
791         bcbearer->bearer.media = &bcbearer->media;
792         bcbearer->media.send_msg = tipc_bcbearer_send;
793         sprintf(bcbearer->media.name, "tipc-multicast");
794
795         bcl = &bclink->link;
796         INIT_LIST_HEAD(&bcl->waiting_ports);
797         bcl->next_out_no = 1;
798         spin_lock_init(&bclink->node.lock);
799         bcl->owner = &bclink->node;
800         bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
801         tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
802         bcl->b_ptr = &bcbearer->bearer;
803         bcl->state = WORKING_WORKING;
804         strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
805
806         return 0;
807 }
808
809 void tipc_bclink_stop(void)
810 {
811         spin_lock_bh(&bc_lock);
812         if (bcbearer) {
813                 tipc_link_stop(bcl);
814                 bcl = NULL;
815                 kfree(bclink);
816                 bclink = NULL;
817                 kfree(bcbearer);
818                 bcbearer = NULL;
819         }
820         spin_unlock_bh(&bc_lock);
821 }
822
823
824 /**
825  * tipc_nmap_add - add a node to a node map
826  */
827
828 void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
829 {
830         int n = tipc_node(node);
831         int w = n / WSIZE;
832         u32 mask = (1 << (n % WSIZE));
833
834         if ((nm_ptr->map[w] & mask) == 0) {
835                 nm_ptr->count++;
836                 nm_ptr->map[w] |= mask;
837         }
838 }
839
840 /**
841  * tipc_nmap_remove - remove a node from a node map
842  */
843
844 void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
845 {
846         int n = tipc_node(node);
847         int w = n / WSIZE;
848         u32 mask = (1 << (n % WSIZE));
849
850         if ((nm_ptr->map[w] & mask) != 0) {
851                 nm_ptr->map[w] &= ~mask;
852                 nm_ptr->count--;
853         }
854 }
855
856 /**
857  * tipc_nmap_diff - find differences between node maps
858  * @nm_a: input node map A
859  * @nm_b: input node map B
860  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
861  */
862
863 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
864                            struct tipc_node_map *nm_b,
865                            struct tipc_node_map *nm_diff)
866 {
867         int stop = ARRAY_SIZE(nm_a->map);
868         int w;
869         int b;
870         u32 map;
871
872         memset(nm_diff, 0, sizeof(*nm_diff));
873         for (w = 0; w < stop; w++) {
874                 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
875                 nm_diff->map[w] = map;
876                 if (map != 0) {
877                         for (b = 0 ; b < WSIZE; b++) {
878                                 if (map & (1 << b))
879                                         nm_diff->count++;
880                         }
881                 }
882         }
883 }
884
885 /**
886  * tipc_port_list_add - add a port to a port list, ensuring no duplicates
887  */
888
889 void tipc_port_list_add(struct port_list *pl_ptr, u32 port)
890 {
891         struct port_list *item = pl_ptr;
892         int i;
893         int item_sz = PLSIZE;
894         int cnt = pl_ptr->count;
895
896         for (; ; cnt -= item_sz, item = item->next) {
897                 if (cnt < PLSIZE)
898                         item_sz = cnt;
899                 for (i = 0; i < item_sz; i++)
900                         if (item->ports[i] == port)
901                                 return;
902                 if (i < PLSIZE) {
903                         item->ports[i] = port;
904                         pl_ptr->count++;
905                         return;
906                 }
907                 if (!item->next) {
908                         item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
909                         if (!item->next) {
910                                 warn("Incomplete multicast delivery, no memory\n");
911                                 return;
912                         }
913                         item->next->next = NULL;
914                 }
915         }
916 }
917
918 /**
919  * tipc_port_list_free - free dynamically created entries in port_list chain
920  *
921  */
922
923 void tipc_port_list_free(struct port_list *pl_ptr)
924 {
925         struct port_list *item;
926         struct port_list *next;
927
928         for (item = pl_ptr->next; item; item = next) {
929                 next = item->next;
930                 kfree(item);
931         }
932 }
933