]> git.karo-electronics.de Git - karo-tx-linux.git/blob - net/tipc/bcast.c
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
[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 #include "name_distr.h"
43
44 #define MAX_PKT_DEFAULT_MCAST   1500    /* bcast link max packet size (fixed) */
45 #define BCLINK_WIN_DEFAULT      20      /* bcast link window size (default) */
46 #define BCBEARER                MAX_BEARERS
47
48 /**
49  * struct tipc_bcbearer_pair - a pair of bearers used by broadcast link
50  * @primary: pointer to primary bearer
51  * @secondary: pointer to secondary bearer
52  *
53  * Bearers must have same priority and same set of reachable destinations
54  * to be paired.
55  */
56
57 struct tipc_bcbearer_pair {
58         struct tipc_bearer *primary;
59         struct tipc_bearer *secondary;
60 };
61
62 /**
63  * struct tipc_bcbearer - bearer used by broadcast link
64  * @bearer: (non-standard) broadcast bearer structure
65  * @media: (non-standard) broadcast media structure
66  * @bpairs: array of bearer pairs
67  * @bpairs_temp: temporary array of bearer pairs used by tipc_bcbearer_sort()
68  * @remains: temporary node map used by tipc_bcbearer_send()
69  * @remains_new: temporary node map used tipc_bcbearer_send()
70  *
71  * Note: The fields labelled "temporary" are incorporated into the bearer
72  * to avoid consuming potentially limited stack space through the use of
73  * large local variables within multicast routines.  Concurrent access is
74  * prevented through use of the spinlock "bclink_lock".
75  */
76 struct tipc_bcbearer {
77         struct tipc_bearer bearer;
78         struct tipc_media media;
79         struct tipc_bcbearer_pair bpairs[MAX_BEARERS];
80         struct tipc_bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1];
81         struct tipc_node_map remains;
82         struct tipc_node_map remains_new;
83 };
84
85 /**
86  * struct tipc_bclink - link used for broadcast messages
87  * @lock: spinlock governing access to structure
88  * @link: (non-standard) broadcast link structure
89  * @node: (non-standard) node structure representing b'cast link's peer node
90  * @flags: represent bclink states
91  * @bcast_nodes: map of broadcast-capable nodes
92  * @retransmit_to: node that most recently requested a retransmit
93  *
94  * Handles sequence numbering, fragmentation, bundling, etc.
95  */
96 struct tipc_bclink {
97         spinlock_t lock;
98         struct tipc_link link;
99         struct tipc_node node;
100         unsigned int flags;
101         struct tipc_node_map bcast_nodes;
102         struct tipc_node *retransmit_to;
103 };
104
105 static struct tipc_bcbearer *bcbearer;
106 static struct tipc_bclink *bclink;
107 static struct tipc_link *bcl;
108
109 const char tipc_bclink_name[] = "broadcast-link";
110
111 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
112                            struct tipc_node_map *nm_b,
113                            struct tipc_node_map *nm_diff);
114 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node);
115 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node);
116
117 static void tipc_bclink_lock(void)
118 {
119         spin_lock_bh(&bclink->lock);
120 }
121
122 static void tipc_bclink_unlock(void)
123 {
124         struct tipc_node *node = NULL;
125
126         if (likely(!bclink->flags)) {
127                 spin_unlock_bh(&bclink->lock);
128                 return;
129         }
130
131         if (bclink->flags & TIPC_BCLINK_RESET) {
132                 bclink->flags &= ~TIPC_BCLINK_RESET;
133                 node = tipc_bclink_retransmit_to();
134         }
135         spin_unlock_bh(&bclink->lock);
136
137         if (node)
138                 tipc_link_reset_all(node);
139 }
140
141 void tipc_bclink_set_flags(unsigned int flags)
142 {
143         bclink->flags |= flags;
144 }
145
146 static u32 bcbuf_acks(struct sk_buff *buf)
147 {
148         return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
149 }
150
151 static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
152 {
153         TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
154 }
155
156 static void bcbuf_decr_acks(struct sk_buff *buf)
157 {
158         bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
159 }
160
161 void tipc_bclink_add_node(u32 addr)
162 {
163         tipc_bclink_lock();
164         tipc_nmap_add(&bclink->bcast_nodes, addr);
165         tipc_bclink_unlock();
166 }
167
168 void tipc_bclink_remove_node(u32 addr)
169 {
170         tipc_bclink_lock();
171         tipc_nmap_remove(&bclink->bcast_nodes, addr);
172         tipc_bclink_unlock();
173 }
174
175 static void bclink_set_last_sent(void)
176 {
177         if (bcl->next_out)
178                 bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
179         else
180                 bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
181 }
182
183 u32 tipc_bclink_get_last_sent(void)
184 {
185         return bcl->fsm_msg_cnt;
186 }
187
188 static void bclink_update_last_sent(struct tipc_node *node, u32 seqno)
189 {
190         node->bclink.last_sent = less_eq(node->bclink.last_sent, seqno) ?
191                                                 seqno : node->bclink.last_sent;
192 }
193
194
195 /**
196  * tipc_bclink_retransmit_to - get most recent node to request retransmission
197  *
198  * Called with bclink_lock locked
199  */
200 struct tipc_node *tipc_bclink_retransmit_to(void)
201 {
202         return bclink->retransmit_to;
203 }
204
205 /**
206  * bclink_retransmit_pkt - retransmit broadcast packets
207  * @after: sequence number of last packet to *not* retransmit
208  * @to: sequence number of last packet to retransmit
209  *
210  * Called with bclink_lock locked
211  */
212 static void bclink_retransmit_pkt(u32 after, u32 to)
213 {
214         struct sk_buff *buf;
215
216         buf = bcl->first_out;
217         while (buf && less_eq(buf_seqno(buf), after))
218                 buf = buf->next;
219         tipc_link_retransmit(bcl, buf, mod(to - after));
220 }
221
222 /**
223  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
224  * @n_ptr: node that sent acknowledgement info
225  * @acked: broadcast sequence # that has been acknowledged
226  *
227  * Node is locked, bclink_lock unlocked.
228  */
229 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
230 {
231         struct sk_buff *crs;
232         struct sk_buff *next;
233         unsigned int released = 0;
234
235         tipc_bclink_lock();
236         /* Bail out if tx queue is empty (no clean up is required) */
237         crs = bcl->first_out;
238         if (!crs)
239                 goto exit;
240
241         /* Determine which messages need to be acknowledged */
242         if (acked == INVALID_LINK_SEQ) {
243                 /*
244                  * Contact with specified node has been lost, so need to
245                  * acknowledge sent messages only (if other nodes still exist)
246                  * or both sent and unsent messages (otherwise)
247                  */
248                 if (bclink->bcast_nodes.count)
249                         acked = bcl->fsm_msg_cnt;
250                 else
251                         acked = bcl->next_out_no;
252         } else {
253                 /*
254                  * Bail out if specified sequence number does not correspond
255                  * to a message that has been sent and not yet acknowledged
256                  */
257                 if (less(acked, buf_seqno(crs)) ||
258                     less(bcl->fsm_msg_cnt, acked) ||
259                     less_eq(acked, n_ptr->bclink.acked))
260                         goto exit;
261         }
262
263         /* Skip over packets that node has previously acknowledged */
264         while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked))
265                 crs = crs->next;
266
267         /* Update packets that node is now acknowledging */
268
269         while (crs && less_eq(buf_seqno(crs), acked)) {
270                 next = crs->next;
271
272                 if (crs != bcl->next_out)
273                         bcbuf_decr_acks(crs);
274                 else {
275                         bcbuf_set_acks(crs, 0);
276                         bcl->next_out = next;
277                         bclink_set_last_sent();
278                 }
279
280                 if (bcbuf_acks(crs) == 0) {
281                         bcl->first_out = next;
282                         bcl->out_queue_size--;
283                         kfree_skb(crs);
284                         released = 1;
285                 }
286                 crs = next;
287         }
288         n_ptr->bclink.acked = acked;
289
290         /* Try resolving broadcast link congestion, if necessary */
291
292         if (unlikely(bcl->next_out)) {
293                 tipc_link_push_queue(bcl);
294                 bclink_set_last_sent();
295         }
296         if (unlikely(released && !list_empty(&bcl->waiting_ports)))
297                 tipc_link_wakeup_ports(bcl, 0);
298 exit:
299         tipc_bclink_unlock();
300 }
301
302 /**
303  * tipc_bclink_update_link_state - update broadcast link state
304  *
305  * RCU and node lock set
306  */
307 void tipc_bclink_update_link_state(struct tipc_node *n_ptr, u32 last_sent)
308 {
309         struct sk_buff *buf;
310
311         /* Ignore "stale" link state info */
312
313         if (less_eq(last_sent, n_ptr->bclink.last_in))
314                 return;
315
316         /* Update link synchronization state; quit if in sync */
317
318         bclink_update_last_sent(n_ptr, last_sent);
319
320         if (n_ptr->bclink.last_sent == n_ptr->bclink.last_in)
321                 return;
322
323         /* Update out-of-sync state; quit if loss is still unconfirmed */
324
325         if ((++n_ptr->bclink.oos_state) == 1) {
326                 if (n_ptr->bclink.deferred_size < (TIPC_MIN_LINK_WIN / 2))
327                         return;
328                 n_ptr->bclink.oos_state++;
329         }
330
331         /* Don't NACK if one has been recently sent (or seen) */
332
333         if (n_ptr->bclink.oos_state & 0x1)
334                 return;
335
336         /* Send NACK */
337
338         buf = tipc_buf_acquire(INT_H_SIZE);
339         if (buf) {
340                 struct tipc_msg *msg = buf_msg(buf);
341
342                 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
343                               INT_H_SIZE, n_ptr->addr);
344                 msg_set_non_seq(msg, 1);
345                 msg_set_mc_netid(msg, tipc_net_id);
346                 msg_set_bcast_ack(msg, n_ptr->bclink.last_in);
347                 msg_set_bcgap_after(msg, n_ptr->bclink.last_in);
348                 msg_set_bcgap_to(msg, n_ptr->bclink.deferred_head
349                                  ? buf_seqno(n_ptr->bclink.deferred_head) - 1
350                                  : n_ptr->bclink.last_sent);
351
352                 tipc_bclink_lock();
353                 tipc_bearer_send(MAX_BEARERS, buf, NULL);
354                 bcl->stats.sent_nacks++;
355                 tipc_bclink_unlock();
356                 kfree_skb(buf);
357
358                 n_ptr->bclink.oos_state++;
359         }
360 }
361
362 /**
363  * bclink_peek_nack - monitor retransmission requests sent by other nodes
364  *
365  * Delay any upcoming NACK by this node if another node has already
366  * requested the first message this node is going to ask for.
367  */
368 static void bclink_peek_nack(struct tipc_msg *msg)
369 {
370         struct tipc_node *n_ptr = tipc_node_find(msg_destnode(msg));
371
372         if (unlikely(!n_ptr))
373                 return;
374
375         tipc_node_lock(n_ptr);
376
377         if (n_ptr->bclink.recv_permitted &&
378             (n_ptr->bclink.last_in != n_ptr->bclink.last_sent) &&
379             (n_ptr->bclink.last_in == msg_bcgap_after(msg)))
380                 n_ptr->bclink.oos_state = 2;
381
382         tipc_node_unlock(n_ptr);
383 }
384
385 /*
386  * tipc_bclink_xmit - broadcast a packet to all nodes in cluster
387  */
388 int tipc_bclink_xmit(struct sk_buff *buf)
389 {
390         int res;
391
392         tipc_bclink_lock();
393
394         if (!bclink->bcast_nodes.count) {
395                 res = msg_data_sz(buf_msg(buf));
396                 kfree_skb(buf);
397                 goto exit;
398         }
399
400         res = __tipc_link_xmit(bcl, buf);
401         if (likely(res >= 0)) {
402                 bclink_set_last_sent();
403                 bcl->stats.queue_sz_counts++;
404                 bcl->stats.accu_queue_sz += bcl->out_queue_size;
405         }
406 exit:
407         tipc_bclink_unlock();
408         return res;
409 }
410
411 /**
412  * bclink_accept_pkt - accept an incoming, in-sequence broadcast packet
413  *
414  * Called with both sending node's lock and bclink_lock taken.
415  */
416 static void bclink_accept_pkt(struct tipc_node *node, u32 seqno)
417 {
418         bclink_update_last_sent(node, seqno);
419         node->bclink.last_in = seqno;
420         node->bclink.oos_state = 0;
421         bcl->stats.recv_info++;
422
423         /*
424          * Unicast an ACK periodically, ensuring that
425          * all nodes in the cluster don't ACK at the same time
426          */
427
428         if (((seqno - tipc_own_addr) % TIPC_MIN_LINK_WIN) == 0) {
429                 tipc_link_proto_xmit(node->active_links[node->addr & 1],
430                                      STATE_MSG, 0, 0, 0, 0, 0);
431                 bcl->stats.sent_acks++;
432         }
433 }
434
435 /**
436  * tipc_bclink_rcv - receive a broadcast packet, and deliver upwards
437  *
438  * RCU is locked, no other locks set
439  */
440 void tipc_bclink_rcv(struct sk_buff *buf)
441 {
442         struct tipc_msg *msg = buf_msg(buf);
443         struct tipc_node *node;
444         u32 next_in;
445         u32 seqno;
446         int deferred;
447
448         /* Screen out unwanted broadcast messages */
449
450         if (msg_mc_netid(msg) != tipc_net_id)
451                 goto exit;
452
453         node = tipc_node_find(msg_prevnode(msg));
454         if (unlikely(!node))
455                 goto exit;
456
457         tipc_node_lock(node);
458         if (unlikely(!node->bclink.recv_permitted))
459                 goto unlock;
460
461         /* Handle broadcast protocol message */
462
463         if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
464                 if (msg_type(msg) != STATE_MSG)
465                         goto unlock;
466                 if (msg_destnode(msg) == tipc_own_addr) {
467                         tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
468                         tipc_node_unlock(node);
469                         tipc_bclink_lock();
470                         bcl->stats.recv_nacks++;
471                         bclink->retransmit_to = node;
472                         bclink_retransmit_pkt(msg_bcgap_after(msg),
473                                               msg_bcgap_to(msg));
474                         tipc_bclink_unlock();
475                 } else {
476                         tipc_node_unlock(node);
477                         bclink_peek_nack(msg);
478                 }
479                 goto exit;
480         }
481
482         /* Handle in-sequence broadcast message */
483
484         seqno = msg_seqno(msg);
485         next_in = mod(node->bclink.last_in + 1);
486
487         if (likely(seqno == next_in)) {
488 receive:
489                 /* Deliver message to destination */
490
491                 if (likely(msg_isdata(msg))) {
492                         tipc_bclink_lock();
493                         bclink_accept_pkt(node, seqno);
494                         tipc_bclink_unlock();
495                         tipc_node_unlock(node);
496                         if (likely(msg_mcast(msg)))
497                                 tipc_port_mcast_rcv(buf, NULL);
498                         else
499                                 kfree_skb(buf);
500                 } else if (msg_user(msg) == MSG_BUNDLER) {
501                         tipc_bclink_lock();
502                         bclink_accept_pkt(node, seqno);
503                         bcl->stats.recv_bundles++;
504                         bcl->stats.recv_bundled += msg_msgcnt(msg);
505                         tipc_bclink_unlock();
506                         tipc_node_unlock(node);
507                         tipc_link_bundle_rcv(buf);
508                 } else if (msg_user(msg) == MSG_FRAGMENTER) {
509                         tipc_buf_append(&node->bclink.reasm_buf, &buf);
510                         if (unlikely(!buf && !node->bclink.reasm_buf))
511                                 goto unlock;
512                         tipc_bclink_lock();
513                         bclink_accept_pkt(node, seqno);
514                         bcl->stats.recv_fragments++;
515                         if (buf) {
516                                 bcl->stats.recv_fragmented++;
517                                 msg = buf_msg(buf);
518                                 tipc_bclink_unlock();
519                                 goto receive;
520                         }
521                         tipc_bclink_unlock();
522                         tipc_node_unlock(node);
523                 } else if (msg_user(msg) == NAME_DISTRIBUTOR) {
524                         tipc_bclink_lock();
525                         bclink_accept_pkt(node, seqno);
526                         tipc_bclink_unlock();
527                         tipc_node_unlock(node);
528                         tipc_named_rcv(buf);
529                 } else {
530                         tipc_bclink_lock();
531                         bclink_accept_pkt(node, seqno);
532                         tipc_bclink_unlock();
533                         tipc_node_unlock(node);
534                         kfree_skb(buf);
535                 }
536                 buf = NULL;
537
538                 /* Determine new synchronization state */
539
540                 tipc_node_lock(node);
541                 if (unlikely(!tipc_node_is_up(node)))
542                         goto unlock;
543
544                 if (node->bclink.last_in == node->bclink.last_sent)
545                         goto unlock;
546
547                 if (!node->bclink.deferred_head) {
548                         node->bclink.oos_state = 1;
549                         goto unlock;
550                 }
551
552                 msg = buf_msg(node->bclink.deferred_head);
553                 seqno = msg_seqno(msg);
554                 next_in = mod(next_in + 1);
555                 if (seqno != next_in)
556                         goto unlock;
557
558                 /* Take in-sequence message from deferred queue & deliver it */
559
560                 buf = node->bclink.deferred_head;
561                 node->bclink.deferred_head = buf->next;
562                 buf->next = NULL;
563                 node->bclink.deferred_size--;
564                 goto receive;
565         }
566
567         /* Handle out-of-sequence broadcast message */
568
569         if (less(next_in, seqno)) {
570                 deferred = tipc_link_defer_pkt(&node->bclink.deferred_head,
571                                                &node->bclink.deferred_tail,
572                                                buf);
573                 node->bclink.deferred_size += deferred;
574                 bclink_update_last_sent(node, seqno);
575                 buf = NULL;
576         } else
577                 deferred = 0;
578
579         tipc_bclink_lock();
580
581         if (deferred)
582                 bcl->stats.deferred_recv++;
583         else
584                 bcl->stats.duplicates++;
585
586         tipc_bclink_unlock();
587
588 unlock:
589         tipc_node_unlock(node);
590 exit:
591         kfree_skb(buf);
592 }
593
594 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
595 {
596         return (n_ptr->bclink.recv_permitted &&
597                 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
598 }
599
600
601 /**
602  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
603  *
604  * Send packet over as many bearers as necessary to reach all nodes
605  * that have joined the broadcast link.
606  *
607  * Returns 0 (packet sent successfully) under all circumstances,
608  * since the broadcast link's pseudo-bearer never blocks
609  */
610 static int tipc_bcbearer_send(struct sk_buff *buf, struct tipc_bearer *unused1,
611                               struct tipc_media_addr *unused2)
612 {
613         int bp_index;
614
615         /* Prepare broadcast link message for reliable transmission,
616          * if first time trying to send it;
617          * preparation is skipped for broadcast link protocol messages
618          * since they are sent in an unreliable manner and don't need it
619          */
620         if (likely(!msg_non_seq(buf_msg(buf)))) {
621                 struct tipc_msg *msg;
622
623                 bcbuf_set_acks(buf, bclink->bcast_nodes.count);
624                 msg = buf_msg(buf);
625                 msg_set_non_seq(msg, 1);
626                 msg_set_mc_netid(msg, tipc_net_id);
627                 bcl->stats.sent_info++;
628
629                 if (WARN_ON(!bclink->bcast_nodes.count)) {
630                         dump_stack();
631                         return 0;
632                 }
633         }
634
635         /* Send buffer over bearers until all targets reached */
636         bcbearer->remains = bclink->bcast_nodes;
637
638         for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
639                 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
640                 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
641                 struct tipc_bearer *b = p;
642                 struct sk_buff *tbuf;
643
644                 if (!p)
645                         break; /* No more bearers to try */
646
647                 tipc_nmap_diff(&bcbearer->remains, &b->nodes,
648                                &bcbearer->remains_new);
649                 if (bcbearer->remains_new.count == bcbearer->remains.count)
650                         continue; /* Nothing added by bearer pair */
651
652                 if (bp_index == 0) {
653                         /* Use original buffer for first bearer */
654                         tipc_bearer_send(b->identity, buf, &b->bcast_addr);
655                 } else {
656                         /* Avoid concurrent buffer access */
657                         tbuf = pskb_copy_for_clone(buf, GFP_ATOMIC);
658                         if (!tbuf)
659                                 break;
660                         tipc_bearer_send(b->identity, tbuf, &b->bcast_addr);
661                         kfree_skb(tbuf); /* Bearer keeps a clone */
662                 }
663
664                 /* Swap bearers for next packet */
665                 if (s) {
666                         bcbearer->bpairs[bp_index].primary = s;
667                         bcbearer->bpairs[bp_index].secondary = p;
668                 }
669
670                 if (bcbearer->remains_new.count == 0)
671                         break; /* All targets reached */
672
673                 bcbearer->remains = bcbearer->remains_new;
674         }
675
676         return 0;
677 }
678
679 /**
680  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
681  */
682 void tipc_bcbearer_sort(struct tipc_node_map *nm_ptr, u32 node, bool action)
683 {
684         struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
685         struct tipc_bcbearer_pair *bp_curr;
686         struct tipc_bearer *b;
687         int b_index;
688         int pri;
689
690         tipc_bclink_lock();
691
692         if (action)
693                 tipc_nmap_add(nm_ptr, node);
694         else
695                 tipc_nmap_remove(nm_ptr, node);
696
697         /* Group bearers by priority (can assume max of two per priority) */
698         memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
699
700         rcu_read_lock();
701         for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
702                 b = rcu_dereference_rtnl(bearer_list[b_index]);
703                 if (!b || !b->nodes.count)
704                         continue;
705
706                 if (!bp_temp[b->priority].primary)
707                         bp_temp[b->priority].primary = b;
708                 else
709                         bp_temp[b->priority].secondary = b;
710         }
711         rcu_read_unlock();
712
713         /* Create array of bearer pairs for broadcasting */
714         bp_curr = bcbearer->bpairs;
715         memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
716
717         for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
718
719                 if (!bp_temp[pri].primary)
720                         continue;
721
722                 bp_curr->primary = bp_temp[pri].primary;
723
724                 if (bp_temp[pri].secondary) {
725                         if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
726                                             &bp_temp[pri].secondary->nodes)) {
727                                 bp_curr->secondary = bp_temp[pri].secondary;
728                         } else {
729                                 bp_curr++;
730                                 bp_curr->primary = bp_temp[pri].secondary;
731                         }
732                 }
733
734                 bp_curr++;
735         }
736
737         tipc_bclink_unlock();
738 }
739
740
741 int tipc_bclink_stats(char *buf, const u32 buf_size)
742 {
743         int ret;
744         struct tipc_stats *s;
745
746         if (!bcl)
747                 return 0;
748
749         tipc_bclink_lock();
750
751         s = &bcl->stats;
752
753         ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
754                             "  Window:%u packets\n",
755                             bcl->name, bcl->queue_limit[0]);
756         ret += tipc_snprintf(buf + ret, buf_size - ret,
757                              "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
758                              s->recv_info, s->recv_fragments,
759                              s->recv_fragmented, s->recv_bundles,
760                              s->recv_bundled);
761         ret += tipc_snprintf(buf + ret, buf_size - ret,
762                              "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
763                              s->sent_info, s->sent_fragments,
764                              s->sent_fragmented, s->sent_bundles,
765                              s->sent_bundled);
766         ret += tipc_snprintf(buf + ret, buf_size - ret,
767                              "  RX naks:%u defs:%u dups:%u\n",
768                              s->recv_nacks, s->deferred_recv, s->duplicates);
769         ret += tipc_snprintf(buf + ret, buf_size - ret,
770                              "  TX naks:%u acks:%u dups:%u\n",
771                              s->sent_nacks, s->sent_acks, s->retransmitted);
772         ret += tipc_snprintf(buf + ret, buf_size - ret,
773                              "  Congestion link:%u  Send queue max:%u avg:%u\n",
774                              s->link_congs, s->max_queue_sz,
775                              s->queue_sz_counts ?
776                              (s->accu_queue_sz / s->queue_sz_counts) : 0);
777
778         tipc_bclink_unlock();
779         return ret;
780 }
781
782 int tipc_bclink_reset_stats(void)
783 {
784         if (!bcl)
785                 return -ENOPROTOOPT;
786
787         tipc_bclink_lock();
788         memset(&bcl->stats, 0, sizeof(bcl->stats));
789         tipc_bclink_unlock();
790         return 0;
791 }
792
793 int tipc_bclink_set_queue_limits(u32 limit)
794 {
795         if (!bcl)
796                 return -ENOPROTOOPT;
797         if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
798                 return -EINVAL;
799
800         tipc_bclink_lock();
801         tipc_link_set_queue_limits(bcl, limit);
802         tipc_bclink_unlock();
803         return 0;
804 }
805
806 int tipc_bclink_init(void)
807 {
808         bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
809         if (!bcbearer)
810                 return -ENOMEM;
811
812         bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
813         if (!bclink) {
814                 kfree(bcbearer);
815                 return -ENOMEM;
816         }
817
818         bcl = &bclink->link;
819         bcbearer->bearer.media = &bcbearer->media;
820         bcbearer->media.send_msg = tipc_bcbearer_send;
821         sprintf(bcbearer->media.name, "tipc-broadcast");
822
823         spin_lock_init(&bclink->lock);
824         INIT_LIST_HEAD(&bcl->waiting_ports);
825         bcl->next_out_no = 1;
826         spin_lock_init(&bclink->node.lock);
827         bcl->owner = &bclink->node;
828         bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
829         tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
830         bcl->bearer_id = MAX_BEARERS;
831         rcu_assign_pointer(bearer_list[MAX_BEARERS], &bcbearer->bearer);
832         bcl->state = WORKING_WORKING;
833         strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
834         return 0;
835 }
836
837 void tipc_bclink_stop(void)
838 {
839         tipc_bclink_lock();
840         tipc_link_purge_queues(bcl);
841         tipc_bclink_unlock();
842
843         RCU_INIT_POINTER(bearer_list[BCBEARER], NULL);
844         synchronize_net();
845         kfree(bcbearer);
846         kfree(bclink);
847 }
848
849 /**
850  * tipc_nmap_add - add a node to a node map
851  */
852 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
853 {
854         int n = tipc_node(node);
855         int w = n / WSIZE;
856         u32 mask = (1 << (n % WSIZE));
857
858         if ((nm_ptr->map[w] & mask) == 0) {
859                 nm_ptr->count++;
860                 nm_ptr->map[w] |= mask;
861         }
862 }
863
864 /**
865  * tipc_nmap_remove - remove a node from a node map
866  */
867 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
868 {
869         int n = tipc_node(node);
870         int w = n / WSIZE;
871         u32 mask = (1 << (n % WSIZE));
872
873         if ((nm_ptr->map[w] & mask) != 0) {
874                 nm_ptr->map[w] &= ~mask;
875                 nm_ptr->count--;
876         }
877 }
878
879 /**
880  * tipc_nmap_diff - find differences between node maps
881  * @nm_a: input node map A
882  * @nm_b: input node map B
883  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
884  */
885 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
886                            struct tipc_node_map *nm_b,
887                            struct tipc_node_map *nm_diff)
888 {
889         int stop = ARRAY_SIZE(nm_a->map);
890         int w;
891         int b;
892         u32 map;
893
894         memset(nm_diff, 0, sizeof(*nm_diff));
895         for (w = 0; w < stop; w++) {
896                 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
897                 nm_diff->map[w] = map;
898                 if (map != 0) {
899                         for (b = 0 ; b < WSIZE; b++) {
900                                 if (map & (1 << b))
901                                         nm_diff->count++;
902                         }
903                 }
904         }
905 }
906
907 /**
908  * tipc_port_list_add - add a port to a port list, ensuring no duplicates
909  */
910 void tipc_port_list_add(struct tipc_port_list *pl_ptr, u32 port)
911 {
912         struct tipc_port_list *item = pl_ptr;
913         int i;
914         int item_sz = PLSIZE;
915         int cnt = pl_ptr->count;
916
917         for (; ; cnt -= item_sz, item = item->next) {
918                 if (cnt < PLSIZE)
919                         item_sz = cnt;
920                 for (i = 0; i < item_sz; i++)
921                         if (item->ports[i] == port)
922                                 return;
923                 if (i < PLSIZE) {
924                         item->ports[i] = port;
925                         pl_ptr->count++;
926                         return;
927                 }
928                 if (!item->next) {
929                         item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
930                         if (!item->next) {
931                                 pr_warn("Incomplete multicast delivery, no memory\n");
932                                 return;
933                         }
934                         item->next->next = NULL;
935                 }
936         }
937 }
938
939 /**
940  * tipc_port_list_free - free dynamically created entries in port_list chain
941  *
942  */
943 void tipc_port_list_free(struct tipc_port_list *pl_ptr)
944 {
945         struct tipc_port_list *item;
946         struct tipc_port_list *next;
947
948         for (item = pl_ptr->next; item; item = next) {
949                 next = item->next;
950                 kfree(item);
951         }
952 }