]> bbs.cooldavid.org Git - net-next-2.6.git/blob - net/tipc/bcast.c
scm: lower SCM_MAX_FD
[net-next-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: temporary array of bearer pairs used by tipc_bcbearer_sort()
85  * @remains: temporary node map used by tipc_bcbearer_send()
86  * @remains_new: temporary node map used tipc_bcbearer_send()
87  *
88  * Note: The fields labelled "temporary" are incorporated into the bearer
89  * to avoid consuming potentially limited stack space through the use of
90  * large local variables within multicast routines.  Concurrent access is
91  * prevented through use of the spinlock "bc_lock".
92  */
93
94 struct bcbearer {
95         struct bearer bearer;
96         struct media media;
97         struct bcbearer_pair bpairs[MAX_BEARERS];
98         struct bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1];
99         struct tipc_node_map remains;
100         struct tipc_node_map remains_new;
101 };
102
103 /**
104  * struct bclink - link used for broadcast messages
105  * @link: (non-standard) broadcast link structure
106  * @node: (non-standard) node structure representing b'cast link's peer node
107  *
108  * Handles sequence numbering, fragmentation, bundling, etc.
109  */
110
111 struct bclink {
112         struct link link;
113         struct tipc_node node;
114 };
115
116
117 static struct bcbearer *bcbearer = NULL;
118 static struct bclink *bclink = NULL;
119 static struct link *bcl = NULL;
120 static DEFINE_SPINLOCK(bc_lock);
121
122 const char tipc_bclink_name[] = "broadcast-link";
123
124 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
125                            struct tipc_node_map *nm_b,
126                            struct tipc_node_map *nm_diff);
127
128 static u32 buf_seqno(struct sk_buff *buf)
129 {
130         return msg_seqno(buf_msg(buf));
131 }
132
133 static u32 bcbuf_acks(struct sk_buff *buf)
134 {
135         return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
136 }
137
138 static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
139 {
140         TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
141 }
142
143 static void bcbuf_decr_acks(struct sk_buff *buf)
144 {
145         bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
146 }
147
148
149 static void bclink_set_last_sent(void)
150 {
151         if (bcl->next_out)
152                 bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
153         else
154                 bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
155 }
156
157 u32 tipc_bclink_get_last_sent(void)
158 {
159         return bcl->fsm_msg_cnt;
160 }
161
162 /**
163  * bclink_set_gap - set gap according to contents of current deferred pkt queue
164  *
165  * Called with 'node' locked, bc_lock unlocked
166  */
167
168 static void bclink_set_gap(struct tipc_node *n_ptr)
169 {
170         struct sk_buff *buf = n_ptr->bclink.deferred_head;
171
172         n_ptr->bclink.gap_after = n_ptr->bclink.gap_to =
173                 mod(n_ptr->bclink.last_in);
174         if (unlikely(buf != NULL))
175                 n_ptr->bclink.gap_to = mod(buf_seqno(buf) - 1);
176 }
177
178 /**
179  * bclink_ack_allowed - test if ACK or NACK message can be sent at this moment
180  *
181  * This mechanism endeavours to prevent all nodes in network from trying
182  * to ACK or NACK at the same time.
183  *
184  * Note: TIPC uses a different trigger to distribute ACKs than it does to
185  *       distribute NACKs, but tries to use the same spacing (divide by 16).
186  */
187
188 static int bclink_ack_allowed(u32 n)
189 {
190         return (n % TIPC_MIN_LINK_WIN) == tipc_own_tag;
191 }
192
193
194 /**
195  * bclink_retransmit_pkt - retransmit broadcast packets
196  * @after: sequence number of last packet to *not* retransmit
197  * @to: sequence number of last packet to retransmit
198  *
199  * Called with bc_lock locked
200  */
201
202 static void bclink_retransmit_pkt(u32 after, u32 to)
203 {
204         struct sk_buff *buf;
205
206         buf = bcl->first_out;
207         while (buf && less_eq(buf_seqno(buf), after)) {
208                 buf = buf->next;
209         }
210         tipc_link_retransmit(bcl, buf, mod(to - after));
211 }
212
213 /**
214  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
215  * @n_ptr: node that sent acknowledgement info
216  * @acked: broadcast sequence # that has been acknowledged
217  *
218  * Node is locked, bc_lock unlocked.
219  */
220
221 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
222 {
223         struct sk_buff *crs;
224         struct sk_buff *next;
225         unsigned int released = 0;
226
227         if (less_eq(acked, n_ptr->bclink.acked))
228                 return;
229
230         spin_lock_bh(&bc_lock);
231
232         /* Skip over packets that node has previously acknowledged */
233
234         crs = bcl->first_out;
235         while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked)) {
236                 crs = crs->next;
237         }
238
239         /* Update packets that node is now acknowledging */
240
241         while (crs && less_eq(buf_seqno(crs), acked)) {
242                 next = crs->next;
243                 bcbuf_decr_acks(crs);
244                 if (bcbuf_acks(crs) == 0) {
245                         bcl->first_out = next;
246                         bcl->out_queue_size--;
247                         buf_discard(crs);
248                         released = 1;
249                 }
250                 crs = next;
251         }
252         n_ptr->bclink.acked = acked;
253
254         /* Try resolving broadcast link congestion, if necessary */
255
256         if (unlikely(bcl->next_out)) {
257                 tipc_link_push_queue(bcl);
258                 bclink_set_last_sent();
259         }
260         if (unlikely(released && !list_empty(&bcl->waiting_ports)))
261                 tipc_link_wakeup_ports(bcl, 0);
262         spin_unlock_bh(&bc_lock);
263 }
264
265 /**
266  * bclink_send_ack - unicast an ACK msg
267  *
268  * tipc_net_lock and node lock set
269  */
270
271 static void bclink_send_ack(struct tipc_node *n_ptr)
272 {
273         struct link *l_ptr = n_ptr->active_links[n_ptr->addr & 1];
274
275         if (l_ptr != NULL)
276                 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
277 }
278
279 /**
280  * bclink_send_nack- broadcast a NACK msg
281  *
282  * tipc_net_lock and node lock set
283  */
284
285 static void bclink_send_nack(struct tipc_node *n_ptr)
286 {
287         struct sk_buff *buf;
288         struct tipc_msg *msg;
289
290         if (!less(n_ptr->bclink.gap_after, n_ptr->bclink.gap_to))
291                 return;
292
293         buf = tipc_buf_acquire(INT_H_SIZE);
294         if (buf) {
295                 msg = buf_msg(buf);
296                 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
297                          INT_H_SIZE, n_ptr->addr);
298                 msg_set_mc_netid(msg, tipc_net_id);
299                 msg_set_bcast_ack(msg, mod(n_ptr->bclink.last_in));
300                 msg_set_bcgap_after(msg, n_ptr->bclink.gap_after);
301                 msg_set_bcgap_to(msg, n_ptr->bclink.gap_to);
302                 msg_set_bcast_tag(msg, tipc_own_tag);
303
304                 if (tipc_bearer_send(&bcbearer->bearer, buf, NULL)) {
305                         bcl->stats.sent_nacks++;
306                         buf_discard(buf);
307                 } else {
308                         tipc_bearer_schedule(bcl->b_ptr, bcl);
309                         bcl->proto_msg_queue = buf;
310                         bcl->stats.bearer_congs++;
311                 }
312
313                 /*
314                  * Ensure we doesn't send another NACK msg to the node
315                  * until 16 more deferred messages arrive from it
316                  * (i.e. helps prevent all nodes from NACK'ing at same time)
317                  */
318
319                 n_ptr->bclink.nack_sync = tipc_own_tag;
320         }
321 }
322
323 /**
324  * tipc_bclink_check_gap - send a NACK if a sequence gap exists
325  *
326  * tipc_net_lock and node lock set
327  */
328
329 void tipc_bclink_check_gap(struct tipc_node *n_ptr, u32 last_sent)
330 {
331         if (!n_ptr->bclink.supported ||
332             less_eq(last_sent, mod(n_ptr->bclink.last_in)))
333                 return;
334
335         bclink_set_gap(n_ptr);
336         if (n_ptr->bclink.gap_after == n_ptr->bclink.gap_to)
337                 n_ptr->bclink.gap_to = last_sent;
338         bclink_send_nack(n_ptr);
339 }
340
341 /**
342  * tipc_bclink_peek_nack - process a NACK msg meant for another node
343  *
344  * Only tipc_net_lock set.
345  */
346
347 static void tipc_bclink_peek_nack(u32 dest, u32 sender_tag, u32 gap_after, u32 gap_to)
348 {
349         struct tipc_node *n_ptr = tipc_node_find(dest);
350         u32 my_after, my_to;
351
352         if (unlikely(!n_ptr || !tipc_node_is_up(n_ptr)))
353                 return;
354         tipc_node_lock(n_ptr);
355         /*
356          * Modify gap to suppress unnecessary NACKs from this node
357          */
358         my_after = n_ptr->bclink.gap_after;
359         my_to = n_ptr->bclink.gap_to;
360
361         if (less_eq(gap_after, my_after)) {
362                 if (less(my_after, gap_to) && less(gap_to, my_to))
363                         n_ptr->bclink.gap_after = gap_to;
364                 else if (less_eq(my_to, gap_to))
365                         n_ptr->bclink.gap_to = n_ptr->bclink.gap_after;
366         } else if (less_eq(gap_after, my_to)) {
367                 if (less_eq(my_to, gap_to))
368                         n_ptr->bclink.gap_to = gap_after;
369         } else {
370                 /*
371                  * Expand gap if missing bufs not in deferred queue:
372                  */
373                 struct sk_buff *buf = n_ptr->bclink.deferred_head;
374                 u32 prev = n_ptr->bclink.gap_to;
375
376                 for (; buf; buf = buf->next) {
377                         u32 seqno = buf_seqno(buf);
378
379                         if (mod(seqno - prev) != 1) {
380                                 buf = NULL;
381                                 break;
382                         }
383                         if (seqno == gap_after)
384                                 break;
385                         prev = seqno;
386                 }
387                 if (buf == NULL)
388                         n_ptr->bclink.gap_to = gap_after;
389         }
390         /*
391          * Some nodes may send a complementary NACK now:
392          */
393         if (bclink_ack_allowed(sender_tag + 1)) {
394                 if (n_ptr->bclink.gap_to != n_ptr->bclink.gap_after) {
395                         bclink_send_nack(n_ptr);
396                         bclink_set_gap(n_ptr);
397                 }
398         }
399         tipc_node_unlock(n_ptr);
400 }
401
402 /**
403  * tipc_bclink_send_msg - broadcast a packet to all nodes in cluster
404  */
405
406 int tipc_bclink_send_msg(struct sk_buff *buf)
407 {
408         int res;
409
410         spin_lock_bh(&bc_lock);
411
412         res = tipc_link_send_buf(bcl, buf);
413         if (unlikely(res == -ELINKCONG))
414                 buf_discard(buf);
415         else
416                 bclink_set_last_sent();
417
418         if (bcl->out_queue_size > bcl->stats.max_queue_sz)
419                 bcl->stats.max_queue_sz = bcl->out_queue_size;
420         bcl->stats.queue_sz_counts++;
421         bcl->stats.accu_queue_sz += bcl->out_queue_size;
422
423         spin_unlock_bh(&bc_lock);
424         return res;
425 }
426
427 /**
428  * tipc_bclink_recv_pkt - receive a broadcast packet, and deliver upwards
429  *
430  * tipc_net_lock is read_locked, no other locks set
431  */
432
433 void tipc_bclink_recv_pkt(struct sk_buff *buf)
434 {
435 #if (TIPC_BCAST_LOSS_RATE)
436         static int rx_count = 0;
437 #endif
438         struct tipc_msg *msg = buf_msg(buf);
439         struct tipc_node* node = tipc_node_find(msg_prevnode(msg));
440         u32 next_in;
441         u32 seqno;
442         struct sk_buff *deferred;
443
444         msg_dbg(msg, "<BC<<<");
445
446         if (unlikely(!node || !tipc_node_is_up(node) || !node->bclink.supported ||
447                      (msg_mc_netid(msg) != tipc_net_id))) {
448                 buf_discard(buf);
449                 return;
450         }
451
452         if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
453                 msg_dbg(msg, "<BCNACK<<<");
454                 if (msg_destnode(msg) == tipc_own_addr) {
455                         tipc_node_lock(node);
456                         tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
457                         tipc_node_unlock(node);
458                         spin_lock_bh(&bc_lock);
459                         bcl->stats.recv_nacks++;
460                         bcl->owner->next = node;   /* remember requestor */
461                         bclink_retransmit_pkt(msg_bcgap_after(msg),
462                                               msg_bcgap_to(msg));
463                         bcl->owner->next = NULL;
464                         spin_unlock_bh(&bc_lock);
465                 } else {
466                         tipc_bclink_peek_nack(msg_destnode(msg),
467                                               msg_bcast_tag(msg),
468                                               msg_bcgap_after(msg),
469                                               msg_bcgap_to(msg));
470                 }
471                 buf_discard(buf);
472                 return;
473         }
474
475 #if (TIPC_BCAST_LOSS_RATE)
476         if (++rx_count == TIPC_BCAST_LOSS_RATE) {
477                 rx_count = 0;
478                 buf_discard(buf);
479                 return;
480         }
481 #endif
482
483         tipc_node_lock(node);
484 receive:
485         deferred = node->bclink.deferred_head;
486         next_in = mod(node->bclink.last_in + 1);
487         seqno = msg_seqno(msg);
488
489         if (likely(seqno == next_in)) {
490                 bcl->stats.recv_info++;
491                 node->bclink.last_in++;
492                 bclink_set_gap(node);
493                 if (unlikely(bclink_ack_allowed(seqno))) {
494                         bclink_send_ack(node);
495                         bcl->stats.sent_acks++;
496                 }
497                 if (likely(msg_isdata(msg))) {
498                         tipc_node_unlock(node);
499                         tipc_port_recv_mcast(buf, NULL);
500                 } else if (msg_user(msg) == MSG_BUNDLER) {
501                         bcl->stats.recv_bundles++;
502                         bcl->stats.recv_bundled += msg_msgcnt(msg);
503                         tipc_node_unlock(node);
504                         tipc_link_recv_bundle(buf);
505                 } else if (msg_user(msg) == MSG_FRAGMENTER) {
506                         bcl->stats.recv_fragments++;
507                         if (tipc_link_recv_fragment(&node->bclink.defragm,
508                                                     &buf, &msg))
509                                 bcl->stats.recv_fragmented++;
510                         tipc_node_unlock(node);
511                         tipc_net_route_msg(buf);
512                 } else {
513                         tipc_node_unlock(node);
514                         tipc_net_route_msg(buf);
515                 }
516                 if (deferred && (buf_seqno(deferred) == mod(next_in + 1))) {
517                         tipc_node_lock(node);
518                         buf = deferred;
519                         msg = buf_msg(buf);
520                         node->bclink.deferred_head = deferred->next;
521                         goto receive;
522                 }
523                 return;
524         } else if (less(next_in, seqno)) {
525                 u32 gap_after = node->bclink.gap_after;
526                 u32 gap_to = node->bclink.gap_to;
527
528                 if (tipc_link_defer_pkt(&node->bclink.deferred_head,
529                                         &node->bclink.deferred_tail,
530                                         buf)) {
531                         node->bclink.nack_sync++;
532                         bcl->stats.deferred_recv++;
533                         if (seqno == mod(gap_after + 1))
534                                 node->bclink.gap_after = seqno;
535                         else if (less(gap_after, seqno) && less(seqno, gap_to))
536                                 node->bclink.gap_to = seqno;
537                 }
538                 if (bclink_ack_allowed(node->bclink.nack_sync)) {
539                         if (gap_to != gap_after)
540                                 bclink_send_nack(node);
541                         bclink_set_gap(node);
542                 }
543         } else {
544                 bcl->stats.duplicates++;
545                 buf_discard(buf);
546         }
547         tipc_node_unlock(node);
548 }
549
550 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
551 {
552         return (n_ptr->bclink.supported &&
553                 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
554 }
555
556
557 /**
558  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
559  *
560  * Send through as many bearers as necessary to reach all nodes
561  * that support TIPC multicasting.
562  *
563  * Returns 0 if packet sent successfully, non-zero if not
564  */
565
566 static int tipc_bcbearer_send(struct sk_buff *buf,
567                               struct tipc_bearer *unused1,
568                               struct tipc_media_addr *unused2)
569 {
570         int bp_index;
571
572         /* Prepare buffer for broadcasting (if first time trying to send it) */
573
574         if (likely(!msg_non_seq(buf_msg(buf)))) {
575                 struct tipc_msg *msg;
576
577                 assert(tipc_cltr_bcast_nodes.count != 0);
578                 bcbuf_set_acks(buf, tipc_cltr_bcast_nodes.count);
579                 msg = buf_msg(buf);
580                 msg_set_non_seq(msg, 1);
581                 msg_set_mc_netid(msg, tipc_net_id);
582                 bcl->stats.sent_info++;
583         }
584
585         /* Send buffer over bearers until all targets reached */
586
587         bcbearer->remains = tipc_cltr_bcast_nodes;
588
589         for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
590                 struct bearer *p = bcbearer->bpairs[bp_index].primary;
591                 struct bearer *s = bcbearer->bpairs[bp_index].secondary;
592
593                 if (!p)
594                         break;  /* no more bearers to try */
595
596                 tipc_nmap_diff(&bcbearer->remains, &p->nodes, &bcbearer->remains_new);
597                 if (bcbearer->remains_new.count == bcbearer->remains.count)
598                         continue;       /* bearer pair doesn't add anything */
599
600                 if (p->publ.blocked ||
601                     p->media->send_msg(buf, &p->publ, &p->media->bcast_addr)) {
602                         /* unable to send on primary bearer */
603                         if (!s || s->publ.blocked ||
604                             s->media->send_msg(buf, &s->publ,
605                                                &s->media->bcast_addr)) {
606                                 /* unable to send on either bearer */
607                                 continue;
608                         }
609                 }
610
611                 if (s) {
612                         bcbearer->bpairs[bp_index].primary = s;
613                         bcbearer->bpairs[bp_index].secondary = p;
614                 }
615
616                 if (bcbearer->remains_new.count == 0)
617                         return 0;
618
619                 bcbearer->remains = bcbearer->remains_new;
620         }
621
622         /*
623          * Unable to reach all targets (indicate success, since currently
624          * there isn't code in place to properly block & unblock the
625          * pseudo-bearer used by the broadcast link)
626          */
627
628         return TIPC_OK;
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 0;
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 0;
778 }
779
780 int tipc_bclink_init(void)
781 {
782         bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
783         bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
784         if (!bcbearer || !bclink) {
785  nomem:
786                 warn("Multicast link creation failed, no memory\n");
787                 kfree(bcbearer);
788                 bcbearer = NULL;
789                 kfree(bclink);
790                 bclink = NULL;
791                 return -ENOMEM;
792         }
793
794         INIT_LIST_HEAD(&bcbearer->bearer.cong_links);
795         bcbearer->bearer.media = &bcbearer->media;
796         bcbearer->media.send_msg = tipc_bcbearer_send;
797         sprintf(bcbearer->media.name, "tipc-multicast");
798
799         bcl = &bclink->link;
800         INIT_LIST_HEAD(&bcl->waiting_ports);
801         bcl->next_out_no = 1;
802         spin_lock_init(&bclink->node.lock);
803         bcl->owner = &bclink->node;
804         bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
805         tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
806         bcl->b_ptr = &bcbearer->bearer;
807         bcl->state = WORKING_WORKING;
808         strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
809
810         if (BCLINK_LOG_BUF_SIZE) {
811                 char *pb = kmalloc(BCLINK_LOG_BUF_SIZE, GFP_ATOMIC);
812
813                 if (!pb)
814                         goto nomem;
815                 tipc_printbuf_init(&bcl->print_buf, pb, BCLINK_LOG_BUF_SIZE);
816         }
817
818         return 0;
819 }
820
821 void tipc_bclink_stop(void)
822 {
823         spin_lock_bh(&bc_lock);
824         if (bcbearer) {
825                 tipc_link_stop(bcl);
826                 if (BCLINK_LOG_BUF_SIZE)
827                         kfree(bcl->print_buf.buf);
828                 bcl = NULL;
829                 kfree(bclink);
830                 bclink = NULL;
831                 kfree(bcbearer);
832                 bcbearer = NULL;
833         }
834         spin_unlock_bh(&bc_lock);
835 }
836
837
838 /**
839  * tipc_nmap_add - add a node to a node map
840  */
841
842 void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
843 {
844         int n = tipc_node(node);
845         int w = n / WSIZE;
846         u32 mask = (1 << (n % WSIZE));
847
848         if ((nm_ptr->map[w] & mask) == 0) {
849                 nm_ptr->count++;
850                 nm_ptr->map[w] |= mask;
851         }
852 }
853
854 /**
855  * tipc_nmap_remove - remove a node from a node map
856  */
857
858 void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
859 {
860         int n = tipc_node(node);
861         int w = n / WSIZE;
862         u32 mask = (1 << (n % WSIZE));
863
864         if ((nm_ptr->map[w] & mask) != 0) {
865                 nm_ptr->map[w] &= ~mask;
866                 nm_ptr->count--;
867         }
868 }
869
870 /**
871  * tipc_nmap_diff - find differences between node maps
872  * @nm_a: input node map A
873  * @nm_b: input node map B
874  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
875  */
876
877 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
878                            struct tipc_node_map *nm_b,
879                            struct tipc_node_map *nm_diff)
880 {
881         int stop = ARRAY_SIZE(nm_a->map);
882         int w;
883         int b;
884         u32 map;
885
886         memset(nm_diff, 0, sizeof(*nm_diff));
887         for (w = 0; w < stop; w++) {
888                 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
889                 nm_diff->map[w] = map;
890                 if (map != 0) {
891                         for (b = 0 ; b < WSIZE; b++) {
892                                 if (map & (1 << b))
893                                         nm_diff->count++;
894                         }
895                 }
896         }
897 }
898
899 /**
900  * tipc_port_list_add - add a port to a port list, ensuring no duplicates
901  */
902
903 void tipc_port_list_add(struct port_list *pl_ptr, u32 port)
904 {
905         struct port_list *item = pl_ptr;
906         int i;
907         int item_sz = PLSIZE;
908         int cnt = pl_ptr->count;
909
910         for (; ; cnt -= item_sz, item = item->next) {
911                 if (cnt < PLSIZE)
912                         item_sz = cnt;
913                 for (i = 0; i < item_sz; i++)
914                         if (item->ports[i] == port)
915                                 return;
916                 if (i < PLSIZE) {
917                         item->ports[i] = port;
918                         pl_ptr->count++;
919                         return;
920                 }
921                 if (!item->next) {
922                         item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
923                         if (!item->next) {
924                                 warn("Incomplete multicast delivery, no memory\n");
925                                 return;
926                         }
927                         item->next->next = NULL;
928                 }
929         }
930 }
931
932 /**
933  * tipc_port_list_free - free dynamically created entries in port_list chain
934  *
935  */
936
937 void tipc_port_list_free(struct port_list *pl_ptr)
938 {
939         struct port_list *item;
940         struct port_list *next;
941
942         for (item = pl_ptr->next; item; item = next) {
943                 next = item->next;
944                 kfree(item);
945         }
946 }
947