Loading...
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 * struct bcbearer_pair - a pair of bearers used by broadcast link
49 * @primary: pointer to primary bearer
50 * @secondary: pointer to secondary bearer
51 *
52 * Bearers must have same priority and same set of reachable destinations
53 * to be paired.
54 */
55
56struct bcbearer_pair {
57 struct tipc_bearer *primary;
58 struct tipc_bearer *secondary;
59};
60
61/**
62 * struct bcbearer - bearer used by broadcast link
63 * @bearer: (non-standard) broadcast bearer structure
64 * @media: (non-standard) broadcast media structure
65 * @bpairs: array of bearer pairs
66 * @bpairs_temp: temporary array of bearer pairs used by tipc_bcbearer_sort()
67 * @remains: temporary node map used by tipc_bcbearer_send()
68 * @remains_new: temporary node map used tipc_bcbearer_send()
69 *
70 * Note: The fields labelled "temporary" are incorporated into the bearer
71 * to avoid consuming potentially limited stack space through the use of
72 * large local variables within multicast routines. Concurrent access is
73 * prevented through use of the spinlock "bc_lock".
74 */
75
76struct bcbearer {
77 struct tipc_bearer bearer;
78 struct media media;
79 struct bcbearer_pair bpairs[MAX_BEARERS];
80 struct 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 bclink - link used for broadcast messages
87 * @link: (non-standard) broadcast link structure
88 * @node: (non-standard) node structure representing b'cast link's peer node
89 * @retransmit_to: node that most recently requested a retransmit
90 *
91 * Handles sequence numbering, fragmentation, bundling, etc.
92 */
93
94struct bclink {
95 struct link link;
96 struct tipc_node node;
97 struct tipc_node *retransmit_to;
98};
99
100
101static struct bcbearer *bcbearer;
102static struct bclink *bclink;
103static struct link *bcl;
104static DEFINE_SPINLOCK(bc_lock);
105
106/* broadcast-capable node map */
107struct tipc_node_map tipc_bcast_nmap;
108
109const char tipc_bclink_name[] = "broadcast-link";
110
111static 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
115static u32 buf_seqno(struct sk_buff *buf)
116{
117 return msg_seqno(buf_msg(buf));
118}
119
120static u32 bcbuf_acks(struct sk_buff *buf)
121{
122 return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
123}
124
125static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
126{
127 TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
128}
129
130static void bcbuf_decr_acks(struct sk_buff *buf)
131{
132 bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
133}
134
135
136static void bclink_set_last_sent(void)
137{
138 if (bcl->next_out)
139 bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
140 else
141 bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
142}
143
144u32 tipc_bclink_get_last_sent(void)
145{
146 return bcl->fsm_msg_cnt;
147}
148
149/**
150 * bclink_set_gap - set gap according to contents of current deferred pkt queue
151 *
152 * Called with 'node' locked, bc_lock unlocked
153 */
154
155static void bclink_set_gap(struct tipc_node *n_ptr)
156{
157 struct sk_buff *buf = n_ptr->bclink.deferred_head;
158
159 n_ptr->bclink.gap_after = n_ptr->bclink.gap_to =
160 mod(n_ptr->bclink.last_in);
161 if (unlikely(buf != NULL))
162 n_ptr->bclink.gap_to = mod(buf_seqno(buf) - 1);
163}
164
165/**
166 * bclink_ack_allowed - test if ACK or NACK message can be sent at this moment
167 *
168 * This mechanism endeavours to prevent all nodes in network from trying
169 * to ACK or NACK at the same time.
170 *
171 * Note: TIPC uses a different trigger to distribute ACKs than it does to
172 * distribute NACKs, but tries to use the same spacing (divide by 16).
173 */
174
175static int bclink_ack_allowed(u32 n)
176{
177 return (n % TIPC_MIN_LINK_WIN) == tipc_own_tag;
178}
179
180
181/**
182 * tipc_bclink_retransmit_to - get most recent node to request retransmission
183 *
184 * Called with bc_lock locked
185 */
186
187struct tipc_node *tipc_bclink_retransmit_to(void)
188{
189 return bclink->retransmit_to;
190}
191
192/**
193 * bclink_retransmit_pkt - retransmit broadcast packets
194 * @after: sequence number of last packet to *not* retransmit
195 * @to: sequence number of last packet to retransmit
196 *
197 * Called with bc_lock locked
198 */
199
200static void bclink_retransmit_pkt(u32 after, u32 to)
201{
202 struct sk_buff *buf;
203
204 buf = bcl->first_out;
205 while (buf && less_eq(buf_seqno(buf), after))
206 buf = buf->next;
207 tipc_link_retransmit(bcl, buf, mod(to - after));
208}
209
210/**
211 * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
212 * @n_ptr: node that sent acknowledgement info
213 * @acked: broadcast sequence # that has been acknowledged
214 *
215 * Node is locked, bc_lock unlocked.
216 */
217
218void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
219{
220 struct sk_buff *crs;
221 struct sk_buff *next;
222 unsigned int released = 0;
223
224 if (less_eq(acked, n_ptr->bclink.acked))
225 return;
226
227 spin_lock_bh(&bc_lock);
228
229 /* Skip over packets that node has previously acknowledged */
230
231 crs = bcl->first_out;
232 while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked))
233 crs = crs->next;
234
235 /* Update packets that node is now acknowledging */
236
237 while (crs && less_eq(buf_seqno(crs), acked)) {
238 next = crs->next;
239 bcbuf_decr_acks(crs);
240 if (bcbuf_acks(crs) == 0) {
241 bcl->first_out = next;
242 bcl->out_queue_size--;
243 buf_discard(crs);
244 released = 1;
245 }
246 crs = next;
247 }
248 n_ptr->bclink.acked = acked;
249
250 /* Try resolving broadcast link congestion, if necessary */
251
252 if (unlikely(bcl->next_out)) {
253 tipc_link_push_queue(bcl);
254 bclink_set_last_sent();
255 }
256 if (unlikely(released && !list_empty(&bcl->waiting_ports)))
257 tipc_link_wakeup_ports(bcl, 0);
258 spin_unlock_bh(&bc_lock);
259}
260
261/**
262 * bclink_send_ack - unicast an ACK msg
263 *
264 * tipc_net_lock and node lock set
265 */
266
267static void bclink_send_ack(struct tipc_node *n_ptr)
268{
269 struct link *l_ptr = n_ptr->active_links[n_ptr->addr & 1];
270
271 if (l_ptr != NULL)
272 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
273}
274
275/**
276 * bclink_send_nack- broadcast a NACK msg
277 *
278 * tipc_net_lock and node lock set
279 */
280
281static void bclink_send_nack(struct tipc_node *n_ptr)
282{
283 struct sk_buff *buf;
284 struct tipc_msg *msg;
285
286 if (!less(n_ptr->bclink.gap_after, n_ptr->bclink.gap_to))
287 return;
288
289 buf = tipc_buf_acquire(INT_H_SIZE);
290 if (buf) {
291 msg = buf_msg(buf);
292 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
293 INT_H_SIZE, n_ptr->addr);
294 msg_set_non_seq(msg, 1);
295 msg_set_mc_netid(msg, tipc_net_id);
296 msg_set_bcast_ack(msg, mod(n_ptr->bclink.last_in));
297 msg_set_bcgap_after(msg, n_ptr->bclink.gap_after);
298 msg_set_bcgap_to(msg, n_ptr->bclink.gap_to);
299 msg_set_bcast_tag(msg, tipc_own_tag);
300
301 if (tipc_bearer_send(&bcbearer->bearer, buf, NULL)) {
302 bcl->stats.sent_nacks++;
303 buf_discard(buf);
304 } else {
305 tipc_bearer_schedule(bcl->b_ptr, bcl);
306 bcl->proto_msg_queue = buf;
307 bcl->stats.bearer_congs++;
308 }
309
310 /*
311 * Ensure we doesn't send another NACK msg to the node
312 * until 16 more deferred messages arrive from it
313 * (i.e. helps prevent all nodes from NACK'ing at same time)
314 */
315
316 n_ptr->bclink.nack_sync = tipc_own_tag;
317 }
318}
319
320/**
321 * tipc_bclink_check_gap - send a NACK if a sequence gap exists
322 *
323 * tipc_net_lock and node lock set
324 */
325
326void tipc_bclink_check_gap(struct tipc_node *n_ptr, u32 last_sent)
327{
328 if (!n_ptr->bclink.supported ||
329 less_eq(last_sent, mod(n_ptr->bclink.last_in)))
330 return;
331
332 bclink_set_gap(n_ptr);
333 if (n_ptr->bclink.gap_after == n_ptr->bclink.gap_to)
334 n_ptr->bclink.gap_to = last_sent;
335 bclink_send_nack(n_ptr);
336}
337
338/**
339 * tipc_bclink_peek_nack - process a NACK msg meant for another node
340 *
341 * Only tipc_net_lock set.
342 */
343
344static void tipc_bclink_peek_nack(u32 dest, u32 sender_tag, u32 gap_after, u32 gap_to)
345{
346 struct tipc_node *n_ptr = tipc_node_find(dest);
347 u32 my_after, my_to;
348
349 if (unlikely(!n_ptr || !tipc_node_is_up(n_ptr)))
350 return;
351 tipc_node_lock(n_ptr);
352 /*
353 * Modify gap to suppress unnecessary NACKs from this node
354 */
355 my_after = n_ptr->bclink.gap_after;
356 my_to = n_ptr->bclink.gap_to;
357
358 if (less_eq(gap_after, my_after)) {
359 if (less(my_after, gap_to) && less(gap_to, my_to))
360 n_ptr->bclink.gap_after = gap_to;
361 else if (less_eq(my_to, gap_to))
362 n_ptr->bclink.gap_to = n_ptr->bclink.gap_after;
363 } else if (less_eq(gap_after, my_to)) {
364 if (less_eq(my_to, gap_to))
365 n_ptr->bclink.gap_to = gap_after;
366 } else {
367 /*
368 * Expand gap if missing bufs not in deferred queue:
369 */
370 struct sk_buff *buf = n_ptr->bclink.deferred_head;
371 u32 prev = n_ptr->bclink.gap_to;
372
373 for (; buf; buf = buf->next) {
374 u32 seqno = buf_seqno(buf);
375
376 if (mod(seqno - prev) != 1) {
377 buf = NULL;
378 break;
379 }
380 if (seqno == gap_after)
381 break;
382 prev = seqno;
383 }
384 if (buf == NULL)
385 n_ptr->bclink.gap_to = gap_after;
386 }
387 /*
388 * Some nodes may send a complementary NACK now:
389 */
390 if (bclink_ack_allowed(sender_tag + 1)) {
391 if (n_ptr->bclink.gap_to != n_ptr->bclink.gap_after) {
392 bclink_send_nack(n_ptr);
393 bclink_set_gap(n_ptr);
394 }
395 }
396 tipc_node_unlock(n_ptr);
397}
398
399/**
400 * tipc_bclink_send_msg - broadcast a packet to all nodes in cluster
401 */
402
403int tipc_bclink_send_msg(struct sk_buff *buf)
404{
405 int res;
406
407 spin_lock_bh(&bc_lock);
408
409 res = tipc_link_send_buf(bcl, buf);
410 if (likely(res > 0))
411 bclink_set_last_sent();
412
413 bcl->stats.queue_sz_counts++;
414 bcl->stats.accu_queue_sz += bcl->out_queue_size;
415
416 spin_unlock_bh(&bc_lock);
417 return res;
418}
419
420/**
421 * tipc_bclink_recv_pkt - receive a broadcast packet, and deliver upwards
422 *
423 * tipc_net_lock is read_locked, no other locks set
424 */
425
426void tipc_bclink_recv_pkt(struct sk_buff *buf)
427{
428 struct tipc_msg *msg = buf_msg(buf);
429 struct tipc_node *node = tipc_node_find(msg_prevnode(msg));
430 u32 next_in;
431 u32 seqno;
432 struct sk_buff *deferred;
433
434 if (unlikely(!node || !tipc_node_is_up(node) || !node->bclink.supported ||
435 (msg_mc_netid(msg) != tipc_net_id))) {
436 buf_discard(buf);
437 return;
438 }
439
440 if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
441 if (msg_destnode(msg) == tipc_own_addr) {
442 tipc_node_lock(node);
443 tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
444 tipc_node_unlock(node);
445 spin_lock_bh(&bc_lock);
446 bcl->stats.recv_nacks++;
447 bclink->retransmit_to = node;
448 bclink_retransmit_pkt(msg_bcgap_after(msg),
449 msg_bcgap_to(msg));
450 spin_unlock_bh(&bc_lock);
451 } else {
452 tipc_bclink_peek_nack(msg_destnode(msg),
453 msg_bcast_tag(msg),
454 msg_bcgap_after(msg),
455 msg_bcgap_to(msg));
456 }
457 buf_discard(buf);
458 return;
459 }
460
461 tipc_node_lock(node);
462receive:
463 deferred = node->bclink.deferred_head;
464 next_in = mod(node->bclink.last_in + 1);
465 seqno = msg_seqno(msg);
466
467 if (likely(seqno == next_in)) {
468 bcl->stats.recv_info++;
469 node->bclink.last_in++;
470 bclink_set_gap(node);
471 if (unlikely(bclink_ack_allowed(seqno))) {
472 bclink_send_ack(node);
473 bcl->stats.sent_acks++;
474 }
475 if (likely(msg_isdata(msg))) {
476 tipc_node_unlock(node);
477 tipc_port_recv_mcast(buf, NULL);
478 } else if (msg_user(msg) == MSG_BUNDLER) {
479 bcl->stats.recv_bundles++;
480 bcl->stats.recv_bundled += msg_msgcnt(msg);
481 tipc_node_unlock(node);
482 tipc_link_recv_bundle(buf);
483 } else if (msg_user(msg) == MSG_FRAGMENTER) {
484 bcl->stats.recv_fragments++;
485 if (tipc_link_recv_fragment(&node->bclink.defragm,
486 &buf, &msg))
487 bcl->stats.recv_fragmented++;
488 tipc_node_unlock(node);
489 tipc_net_route_msg(buf);
490 } else {
491 tipc_node_unlock(node);
492 tipc_net_route_msg(buf);
493 }
494 if (deferred && (buf_seqno(deferred) == mod(next_in + 1))) {
495 tipc_node_lock(node);
496 buf = deferred;
497 msg = buf_msg(buf);
498 node->bclink.deferred_head = deferred->next;
499 goto receive;
500 }
501 return;
502 } else if (less(next_in, seqno)) {
503 u32 gap_after = node->bclink.gap_after;
504 u32 gap_to = node->bclink.gap_to;
505
506 if (tipc_link_defer_pkt(&node->bclink.deferred_head,
507 &node->bclink.deferred_tail,
508 buf)) {
509 node->bclink.nack_sync++;
510 bcl->stats.deferred_recv++;
511 if (seqno == mod(gap_after + 1))
512 node->bclink.gap_after = seqno;
513 else if (less(gap_after, seqno) && less(seqno, gap_to))
514 node->bclink.gap_to = seqno;
515 }
516 if (bclink_ack_allowed(node->bclink.nack_sync)) {
517 if (gap_to != gap_after)
518 bclink_send_nack(node);
519 bclink_set_gap(node);
520 }
521 } else {
522 bcl->stats.duplicates++;
523 buf_discard(buf);
524 }
525 tipc_node_unlock(node);
526}
527
528u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
529{
530 return (n_ptr->bclink.supported &&
531 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
532}
533
534
535/**
536 * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
537 *
538 * Send through as many bearers as necessary to reach all nodes
539 * that support TIPC multicasting.
540 *
541 * Returns 0 if packet sent successfully, non-zero if not
542 */
543
544static int tipc_bcbearer_send(struct sk_buff *buf,
545 struct tipc_bearer *unused1,
546 struct tipc_media_addr *unused2)
547{
548 int bp_index;
549
550 /* Prepare buffer for broadcasting (if first time trying to send it) */
551
552 if (likely(!msg_non_seq(buf_msg(buf)))) {
553 struct tipc_msg *msg;
554
555 bcbuf_set_acks(buf, tipc_bcast_nmap.count);
556 msg = buf_msg(buf);
557 msg_set_non_seq(msg, 1);
558 msg_set_mc_netid(msg, tipc_net_id);
559 bcl->stats.sent_info++;
560
561 if (WARN_ON(!tipc_bcast_nmap.count)) {
562 dump_stack();
563 return 0;
564 }
565 }
566
567 /* Send buffer over bearers until all targets reached */
568
569 bcbearer->remains = tipc_bcast_nmap;
570
571 for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
572 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
573 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
574
575 if (!p)
576 break; /* no more bearers to try */
577
578 tipc_nmap_diff(&bcbearer->remains, &p->nodes, &bcbearer->remains_new);
579 if (bcbearer->remains_new.count == bcbearer->remains.count)
580 continue; /* bearer pair doesn't add anything */
581
582 if (p->blocked ||
583 p->media->send_msg(buf, p, &p->media->bcast_addr)) {
584 /* unable to send on primary bearer */
585 if (!s || s->blocked ||
586 s->media->send_msg(buf, s,
587 &s->media->bcast_addr)) {
588 /* unable to send on either bearer */
589 continue;
590 }
591 }
592
593 if (s) {
594 bcbearer->bpairs[bp_index].primary = s;
595 bcbearer->bpairs[bp_index].secondary = p;
596 }
597
598 if (bcbearer->remains_new.count == 0)
599 return 0;
600
601 bcbearer->remains = bcbearer->remains_new;
602 }
603
604 /*
605 * Unable to reach all targets (indicate success, since currently
606 * there isn't code in place to properly block & unblock the
607 * pseudo-bearer used by the broadcast link)
608 */
609
610 return TIPC_OK;
611}
612
613/**
614 * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
615 */
616
617void tipc_bcbearer_sort(void)
618{
619 struct bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
620 struct bcbearer_pair *bp_curr;
621 int b_index;
622 int pri;
623
624 spin_lock_bh(&bc_lock);
625
626 /* Group bearers by priority (can assume max of two per priority) */
627
628 memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
629
630 for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
631 struct tipc_bearer *b = &tipc_bearers[b_index];
632
633 if (!b->active || !b->nodes.count)
634 continue;
635
636 if (!bp_temp[b->priority].primary)
637 bp_temp[b->priority].primary = b;
638 else
639 bp_temp[b->priority].secondary = b;
640 }
641
642 /* Create array of bearer pairs for broadcasting */
643
644 bp_curr = bcbearer->bpairs;
645 memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
646
647 for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
648
649 if (!bp_temp[pri].primary)
650 continue;
651
652 bp_curr->primary = bp_temp[pri].primary;
653
654 if (bp_temp[pri].secondary) {
655 if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
656 &bp_temp[pri].secondary->nodes)) {
657 bp_curr->secondary = bp_temp[pri].secondary;
658 } else {
659 bp_curr++;
660 bp_curr->primary = bp_temp[pri].secondary;
661 }
662 }
663
664 bp_curr++;
665 }
666
667 spin_unlock_bh(&bc_lock);
668}
669
670/**
671 * tipc_bcbearer_push - resolve bearer congestion
672 *
673 * Forces bclink to push out any unsent packets, until all packets are gone
674 * or congestion reoccurs.
675 * No locks set when function called
676 */
677
678void tipc_bcbearer_push(void)
679{
680 struct tipc_bearer *b_ptr;
681
682 spin_lock_bh(&bc_lock);
683 b_ptr = &bcbearer->bearer;
684 if (b_ptr->blocked) {
685 b_ptr->blocked = 0;
686 tipc_bearer_lock_push(b_ptr);
687 }
688 spin_unlock_bh(&bc_lock);
689}
690
691
692int tipc_bclink_stats(char *buf, const u32 buf_size)
693{
694 struct print_buf pb;
695
696 if (!bcl)
697 return 0;
698
699 tipc_printbuf_init(&pb, buf, buf_size);
700
701 spin_lock_bh(&bc_lock);
702
703 tipc_printf(&pb, "Link <%s>\n"
704 " Window:%u packets\n",
705 bcl->name, bcl->queue_limit[0]);
706 tipc_printf(&pb, " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
707 bcl->stats.recv_info,
708 bcl->stats.recv_fragments,
709 bcl->stats.recv_fragmented,
710 bcl->stats.recv_bundles,
711 bcl->stats.recv_bundled);
712 tipc_printf(&pb, " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
713 bcl->stats.sent_info,
714 bcl->stats.sent_fragments,
715 bcl->stats.sent_fragmented,
716 bcl->stats.sent_bundles,
717 bcl->stats.sent_bundled);
718 tipc_printf(&pb, " RX naks:%u defs:%u dups:%u\n",
719 bcl->stats.recv_nacks,
720 bcl->stats.deferred_recv,
721 bcl->stats.duplicates);
722 tipc_printf(&pb, " TX naks:%u acks:%u dups:%u\n",
723 bcl->stats.sent_nacks,
724 bcl->stats.sent_acks,
725 bcl->stats.retransmitted);
726 tipc_printf(&pb, " Congestion bearer:%u link:%u Send queue max:%u avg:%u\n",
727 bcl->stats.bearer_congs,
728 bcl->stats.link_congs,
729 bcl->stats.max_queue_sz,
730 bcl->stats.queue_sz_counts
731 ? (bcl->stats.accu_queue_sz / bcl->stats.queue_sz_counts)
732 : 0);
733
734 spin_unlock_bh(&bc_lock);
735 return tipc_printbuf_validate(&pb);
736}
737
738int tipc_bclink_reset_stats(void)
739{
740 if (!bcl)
741 return -ENOPROTOOPT;
742
743 spin_lock_bh(&bc_lock);
744 memset(&bcl->stats, 0, sizeof(bcl->stats));
745 spin_unlock_bh(&bc_lock);
746 return 0;
747}
748
749int tipc_bclink_set_queue_limits(u32 limit)
750{
751 if (!bcl)
752 return -ENOPROTOOPT;
753 if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
754 return -EINVAL;
755
756 spin_lock_bh(&bc_lock);
757 tipc_link_set_queue_limits(bcl, limit);
758 spin_unlock_bh(&bc_lock);
759 return 0;
760}
761
762int tipc_bclink_init(void)
763{
764 bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
765 bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
766 if (!bcbearer || !bclink) {
767 warn("Multicast link creation failed, no memory\n");
768 kfree(bcbearer);
769 bcbearer = NULL;
770 kfree(bclink);
771 bclink = NULL;
772 return -ENOMEM;
773 }
774
775 INIT_LIST_HEAD(&bcbearer->bearer.cong_links);
776 bcbearer->bearer.media = &bcbearer->media;
777 bcbearer->media.send_msg = tipc_bcbearer_send;
778 sprintf(bcbearer->media.name, "tipc-multicast");
779
780 bcl = &bclink->link;
781 INIT_LIST_HEAD(&bcl->waiting_ports);
782 bcl->next_out_no = 1;
783 spin_lock_init(&bclink->node.lock);
784 bcl->owner = &bclink->node;
785 bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
786 tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
787 bcl->b_ptr = &bcbearer->bearer;
788 bcl->state = WORKING_WORKING;
789 strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
790
791 return 0;
792}
793
794void tipc_bclink_stop(void)
795{
796 spin_lock_bh(&bc_lock);
797 if (bcbearer) {
798 tipc_link_stop(bcl);
799 bcl = NULL;
800 kfree(bclink);
801 bclink = NULL;
802 kfree(bcbearer);
803 bcbearer = NULL;
804 }
805 spin_unlock_bh(&bc_lock);
806}
807
808
809/**
810 * tipc_nmap_add - add a node to a node map
811 */
812
813void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
814{
815 int n = tipc_node(node);
816 int w = n / WSIZE;
817 u32 mask = (1 << (n % WSIZE));
818
819 if ((nm_ptr->map[w] & mask) == 0) {
820 nm_ptr->count++;
821 nm_ptr->map[w] |= mask;
822 }
823}
824
825/**
826 * tipc_nmap_remove - remove a node from a node map
827 */
828
829void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
830{
831 int n = tipc_node(node);
832 int w = n / WSIZE;
833 u32 mask = (1 << (n % WSIZE));
834
835 if ((nm_ptr->map[w] & mask) != 0) {
836 nm_ptr->map[w] &= ~mask;
837 nm_ptr->count--;
838 }
839}
840
841/**
842 * tipc_nmap_diff - find differences between node maps
843 * @nm_a: input node map A
844 * @nm_b: input node map B
845 * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
846 */
847
848static void tipc_nmap_diff(struct tipc_node_map *nm_a,
849 struct tipc_node_map *nm_b,
850 struct tipc_node_map *nm_diff)
851{
852 int stop = ARRAY_SIZE(nm_a->map);
853 int w;
854 int b;
855 u32 map;
856
857 memset(nm_diff, 0, sizeof(*nm_diff));
858 for (w = 0; w < stop; w++) {
859 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
860 nm_diff->map[w] = map;
861 if (map != 0) {
862 for (b = 0 ; b < WSIZE; b++) {
863 if (map & (1 << b))
864 nm_diff->count++;
865 }
866 }
867 }
868}
869
870/**
871 * tipc_port_list_add - add a port to a port list, ensuring no duplicates
872 */
873
874void tipc_port_list_add(struct port_list *pl_ptr, u32 port)
875{
876 struct port_list *item = pl_ptr;
877 int i;
878 int item_sz = PLSIZE;
879 int cnt = pl_ptr->count;
880
881 for (; ; cnt -= item_sz, item = item->next) {
882 if (cnt < PLSIZE)
883 item_sz = cnt;
884 for (i = 0; i < item_sz; i++)
885 if (item->ports[i] == port)
886 return;
887 if (i < PLSIZE) {
888 item->ports[i] = port;
889 pl_ptr->count++;
890 return;
891 }
892 if (!item->next) {
893 item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
894 if (!item->next) {
895 warn("Incomplete multicast delivery, no memory\n");
896 return;
897 }
898 item->next->next = NULL;
899 }
900 }
901}
902
903/**
904 * tipc_port_list_free - free dynamically created entries in port_list chain
905 *
906 */
907
908void tipc_port_list_free(struct port_list *pl_ptr)
909{
910 struct port_list *item;
911 struct port_list *next;
912
913 for (item = pl_ptr->next; item; item = next) {
914 next = item->next;
915 kfree(item);
916 }
917}
918
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
57struct 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 "bc_lock".
75 */
76struct 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 * @link: (non-standard) broadcast link structure
88 * @node: (non-standard) node structure representing b'cast link's peer node
89 * @bcast_nodes: map of broadcast-capable nodes
90 * @retransmit_to: node that most recently requested a retransmit
91 *
92 * Handles sequence numbering, fragmentation, bundling, etc.
93 */
94struct tipc_bclink {
95 struct tipc_link link;
96 struct tipc_node node;
97 struct tipc_node_map bcast_nodes;
98 struct tipc_node *retransmit_to;
99};
100
101static struct tipc_bcbearer bcast_bearer;
102static struct tipc_bclink bcast_link;
103
104static struct tipc_bcbearer *bcbearer = &bcast_bearer;
105static struct tipc_bclink *bclink = &bcast_link;
106static struct tipc_link *bcl = &bcast_link.link;
107
108static DEFINE_SPINLOCK(bc_lock);
109
110const char tipc_bclink_name[] = "broadcast-link";
111
112static void tipc_nmap_diff(struct tipc_node_map *nm_a,
113 struct tipc_node_map *nm_b,
114 struct tipc_node_map *nm_diff);
115
116static u32 bcbuf_acks(struct sk_buff *buf)
117{
118 return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
119}
120
121static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
122{
123 TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
124}
125
126static void bcbuf_decr_acks(struct sk_buff *buf)
127{
128 bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
129}
130
131void tipc_bclink_add_node(u32 addr)
132{
133 spin_lock_bh(&bc_lock);
134 tipc_nmap_add(&bclink->bcast_nodes, addr);
135 spin_unlock_bh(&bc_lock);
136}
137
138void tipc_bclink_remove_node(u32 addr)
139{
140 spin_lock_bh(&bc_lock);
141 tipc_nmap_remove(&bclink->bcast_nodes, addr);
142 spin_unlock_bh(&bc_lock);
143}
144
145static void bclink_set_last_sent(void)
146{
147 if (bcl->next_out)
148 bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
149 else
150 bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
151}
152
153u32 tipc_bclink_get_last_sent(void)
154{
155 return bcl->fsm_msg_cnt;
156}
157
158static void bclink_update_last_sent(struct tipc_node *node, u32 seqno)
159{
160 node->bclink.last_sent = less_eq(node->bclink.last_sent, seqno) ?
161 seqno : node->bclink.last_sent;
162}
163
164
165/**
166 * tipc_bclink_retransmit_to - get most recent node to request retransmission
167 *
168 * Called with bc_lock locked
169 */
170struct tipc_node *tipc_bclink_retransmit_to(void)
171{
172 return bclink->retransmit_to;
173}
174
175/**
176 * bclink_retransmit_pkt - retransmit broadcast packets
177 * @after: sequence number of last packet to *not* retransmit
178 * @to: sequence number of last packet to retransmit
179 *
180 * Called with bc_lock locked
181 */
182static void bclink_retransmit_pkt(u32 after, u32 to)
183{
184 struct sk_buff *buf;
185
186 buf = bcl->first_out;
187 while (buf && less_eq(buf_seqno(buf), after))
188 buf = buf->next;
189 tipc_link_retransmit(bcl, buf, mod(to - after));
190}
191
192/**
193 * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
194 * @n_ptr: node that sent acknowledgement info
195 * @acked: broadcast sequence # that has been acknowledged
196 *
197 * Node is locked, bc_lock unlocked.
198 */
199void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
200{
201 struct sk_buff *crs;
202 struct sk_buff *next;
203 unsigned int released = 0;
204
205 spin_lock_bh(&bc_lock);
206
207 /* Bail out if tx queue is empty (no clean up is required) */
208 crs = bcl->first_out;
209 if (!crs)
210 goto exit;
211
212 /* Determine which messages need to be acknowledged */
213 if (acked == INVALID_LINK_SEQ) {
214 /*
215 * Contact with specified node has been lost, so need to
216 * acknowledge sent messages only (if other nodes still exist)
217 * or both sent and unsent messages (otherwise)
218 */
219 if (bclink->bcast_nodes.count)
220 acked = bcl->fsm_msg_cnt;
221 else
222 acked = bcl->next_out_no;
223 } else {
224 /*
225 * Bail out if specified sequence number does not correspond
226 * to a message that has been sent and not yet acknowledged
227 */
228 if (less(acked, buf_seqno(crs)) ||
229 less(bcl->fsm_msg_cnt, acked) ||
230 less_eq(acked, n_ptr->bclink.acked))
231 goto exit;
232 }
233
234 /* Skip over packets that node has previously acknowledged */
235 while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked))
236 crs = crs->next;
237
238 /* Update packets that node is now acknowledging */
239
240 while (crs && less_eq(buf_seqno(crs), acked)) {
241 next = crs->next;
242
243 if (crs != bcl->next_out)
244 bcbuf_decr_acks(crs);
245 else {
246 bcbuf_set_acks(crs, 0);
247 bcl->next_out = next;
248 bclink_set_last_sent();
249 }
250
251 if (bcbuf_acks(crs) == 0) {
252 bcl->first_out = next;
253 bcl->out_queue_size--;
254 kfree_skb(crs);
255 released = 1;
256 }
257 crs = next;
258 }
259 n_ptr->bclink.acked = acked;
260
261 /* Try resolving broadcast link congestion, if necessary */
262
263 if (unlikely(bcl->next_out)) {
264 tipc_link_push_queue(bcl);
265 bclink_set_last_sent();
266 }
267 if (unlikely(released && !list_empty(&bcl->waiting_ports)))
268 tipc_link_wakeup_ports(bcl, 0);
269exit:
270 spin_unlock_bh(&bc_lock);
271}
272
273/**
274 * tipc_bclink_update_link_state - update broadcast link state
275 *
276 * tipc_net_lock and node lock set
277 */
278void tipc_bclink_update_link_state(struct tipc_node *n_ptr, u32 last_sent)
279{
280 struct sk_buff *buf;
281
282 /* Ignore "stale" link state info */
283
284 if (less_eq(last_sent, n_ptr->bclink.last_in))
285 return;
286
287 /* Update link synchronization state; quit if in sync */
288
289 bclink_update_last_sent(n_ptr, last_sent);
290
291 if (n_ptr->bclink.last_sent == n_ptr->bclink.last_in)
292 return;
293
294 /* Update out-of-sync state; quit if loss is still unconfirmed */
295
296 if ((++n_ptr->bclink.oos_state) == 1) {
297 if (n_ptr->bclink.deferred_size < (TIPC_MIN_LINK_WIN / 2))
298 return;
299 n_ptr->bclink.oos_state++;
300 }
301
302 /* Don't NACK if one has been recently sent (or seen) */
303
304 if (n_ptr->bclink.oos_state & 0x1)
305 return;
306
307 /* Send NACK */
308
309 buf = tipc_buf_acquire(INT_H_SIZE);
310 if (buf) {
311 struct tipc_msg *msg = buf_msg(buf);
312
313 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
314 INT_H_SIZE, n_ptr->addr);
315 msg_set_non_seq(msg, 1);
316 msg_set_mc_netid(msg, tipc_net_id);
317 msg_set_bcast_ack(msg, n_ptr->bclink.last_in);
318 msg_set_bcgap_after(msg, n_ptr->bclink.last_in);
319 msg_set_bcgap_to(msg, n_ptr->bclink.deferred_head
320 ? buf_seqno(n_ptr->bclink.deferred_head) - 1
321 : n_ptr->bclink.last_sent);
322
323 spin_lock_bh(&bc_lock);
324 tipc_bearer_send(&bcbearer->bearer, buf, NULL);
325 bcl->stats.sent_nacks++;
326 spin_unlock_bh(&bc_lock);
327 kfree_skb(buf);
328
329 n_ptr->bclink.oos_state++;
330 }
331}
332
333/**
334 * bclink_peek_nack - monitor retransmission requests sent by other nodes
335 *
336 * Delay any upcoming NACK by this node if another node has already
337 * requested the first message this node is going to ask for.
338 *
339 * Only tipc_net_lock set.
340 */
341static void bclink_peek_nack(struct tipc_msg *msg)
342{
343 struct tipc_node *n_ptr = tipc_node_find(msg_destnode(msg));
344
345 if (unlikely(!n_ptr))
346 return;
347
348 tipc_node_lock(n_ptr);
349
350 if (n_ptr->bclink.recv_permitted &&
351 (n_ptr->bclink.last_in != n_ptr->bclink.last_sent) &&
352 (n_ptr->bclink.last_in == msg_bcgap_after(msg)))
353 n_ptr->bclink.oos_state = 2;
354
355 tipc_node_unlock(n_ptr);
356}
357
358/*
359 * tipc_bclink_xmit - broadcast a packet to all nodes in cluster
360 */
361int tipc_bclink_xmit(struct sk_buff *buf)
362{
363 int res;
364
365 spin_lock_bh(&bc_lock);
366
367 if (!bclink->bcast_nodes.count) {
368 res = msg_data_sz(buf_msg(buf));
369 kfree_skb(buf);
370 goto exit;
371 }
372
373 res = __tipc_link_xmit(bcl, buf);
374 if (likely(res >= 0)) {
375 bclink_set_last_sent();
376 bcl->stats.queue_sz_counts++;
377 bcl->stats.accu_queue_sz += bcl->out_queue_size;
378 }
379exit:
380 spin_unlock_bh(&bc_lock);
381 return res;
382}
383
384/**
385 * bclink_accept_pkt - accept an incoming, in-sequence broadcast packet
386 *
387 * Called with both sending node's lock and bc_lock taken.
388 */
389static void bclink_accept_pkt(struct tipc_node *node, u32 seqno)
390{
391 bclink_update_last_sent(node, seqno);
392 node->bclink.last_in = seqno;
393 node->bclink.oos_state = 0;
394 bcl->stats.recv_info++;
395
396 /*
397 * Unicast an ACK periodically, ensuring that
398 * all nodes in the cluster don't ACK at the same time
399 */
400
401 if (((seqno - tipc_own_addr) % TIPC_MIN_LINK_WIN) == 0) {
402 tipc_link_proto_xmit(node->active_links[node->addr & 1],
403 STATE_MSG, 0, 0, 0, 0, 0);
404 bcl->stats.sent_acks++;
405 }
406}
407
408/**
409 * tipc_bclink_rcv - receive a broadcast packet, and deliver upwards
410 *
411 * tipc_net_lock is read_locked, no other locks set
412 */
413void tipc_bclink_rcv(struct sk_buff *buf)
414{
415 struct tipc_msg *msg = buf_msg(buf);
416 struct tipc_node *node;
417 u32 next_in;
418 u32 seqno;
419 int deferred;
420
421 /* Screen out unwanted broadcast messages */
422
423 if (msg_mc_netid(msg) != tipc_net_id)
424 goto exit;
425
426 node = tipc_node_find(msg_prevnode(msg));
427 if (unlikely(!node))
428 goto exit;
429
430 tipc_node_lock(node);
431 if (unlikely(!node->bclink.recv_permitted))
432 goto unlock;
433
434 /* Handle broadcast protocol message */
435
436 if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
437 if (msg_type(msg) != STATE_MSG)
438 goto unlock;
439 if (msg_destnode(msg) == tipc_own_addr) {
440 tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
441 tipc_node_unlock(node);
442 spin_lock_bh(&bc_lock);
443 bcl->stats.recv_nacks++;
444 bclink->retransmit_to = node;
445 bclink_retransmit_pkt(msg_bcgap_after(msg),
446 msg_bcgap_to(msg));
447 spin_unlock_bh(&bc_lock);
448 } else {
449 tipc_node_unlock(node);
450 bclink_peek_nack(msg);
451 }
452 goto exit;
453 }
454
455 /* Handle in-sequence broadcast message */
456
457 seqno = msg_seqno(msg);
458 next_in = mod(node->bclink.last_in + 1);
459
460 if (likely(seqno == next_in)) {
461receive:
462 /* Deliver message to destination */
463
464 if (likely(msg_isdata(msg))) {
465 spin_lock_bh(&bc_lock);
466 bclink_accept_pkt(node, seqno);
467 spin_unlock_bh(&bc_lock);
468 tipc_node_unlock(node);
469 if (likely(msg_mcast(msg)))
470 tipc_port_mcast_rcv(buf, NULL);
471 else
472 kfree_skb(buf);
473 } else if (msg_user(msg) == MSG_BUNDLER) {
474 spin_lock_bh(&bc_lock);
475 bclink_accept_pkt(node, seqno);
476 bcl->stats.recv_bundles++;
477 bcl->stats.recv_bundled += msg_msgcnt(msg);
478 spin_unlock_bh(&bc_lock);
479 tipc_node_unlock(node);
480 tipc_link_bundle_rcv(buf);
481 } else if (msg_user(msg) == MSG_FRAGMENTER) {
482 int ret;
483 ret = tipc_link_frag_rcv(&node->bclink.reasm_head,
484 &node->bclink.reasm_tail,
485 &buf);
486 if (ret == LINK_REASM_ERROR)
487 goto unlock;
488 spin_lock_bh(&bc_lock);
489 bclink_accept_pkt(node, seqno);
490 bcl->stats.recv_fragments++;
491 if (ret == LINK_REASM_COMPLETE) {
492 bcl->stats.recv_fragmented++;
493 /* Point msg to inner header */
494 msg = buf_msg(buf);
495 spin_unlock_bh(&bc_lock);
496 goto receive;
497 }
498 spin_unlock_bh(&bc_lock);
499 tipc_node_unlock(node);
500 } else if (msg_user(msg) == NAME_DISTRIBUTOR) {
501 spin_lock_bh(&bc_lock);
502 bclink_accept_pkt(node, seqno);
503 spin_unlock_bh(&bc_lock);
504 tipc_node_unlock(node);
505 tipc_named_rcv(buf);
506 } else {
507 spin_lock_bh(&bc_lock);
508 bclink_accept_pkt(node, seqno);
509 spin_unlock_bh(&bc_lock);
510 tipc_node_unlock(node);
511 kfree_skb(buf);
512 }
513 buf = NULL;
514
515 /* Determine new synchronization state */
516
517 tipc_node_lock(node);
518 if (unlikely(!tipc_node_is_up(node)))
519 goto unlock;
520
521 if (node->bclink.last_in == node->bclink.last_sent)
522 goto unlock;
523
524 if (!node->bclink.deferred_head) {
525 node->bclink.oos_state = 1;
526 goto unlock;
527 }
528
529 msg = buf_msg(node->bclink.deferred_head);
530 seqno = msg_seqno(msg);
531 next_in = mod(next_in + 1);
532 if (seqno != next_in)
533 goto unlock;
534
535 /* Take in-sequence message from deferred queue & deliver it */
536
537 buf = node->bclink.deferred_head;
538 node->bclink.deferred_head = buf->next;
539 node->bclink.deferred_size--;
540 goto receive;
541 }
542
543 /* Handle out-of-sequence broadcast message */
544
545 if (less(next_in, seqno)) {
546 deferred = tipc_link_defer_pkt(&node->bclink.deferred_head,
547 &node->bclink.deferred_tail,
548 buf);
549 node->bclink.deferred_size += deferred;
550 bclink_update_last_sent(node, seqno);
551 buf = NULL;
552 } else
553 deferred = 0;
554
555 spin_lock_bh(&bc_lock);
556
557 if (deferred)
558 bcl->stats.deferred_recv++;
559 else
560 bcl->stats.duplicates++;
561
562 spin_unlock_bh(&bc_lock);
563
564unlock:
565 tipc_node_unlock(node);
566exit:
567 kfree_skb(buf);
568}
569
570u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
571{
572 return (n_ptr->bclink.recv_permitted &&
573 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
574}
575
576
577/**
578 * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
579 *
580 * Send packet over as many bearers as necessary to reach all nodes
581 * that have joined the broadcast link.
582 *
583 * Returns 0 (packet sent successfully) under all circumstances,
584 * since the broadcast link's pseudo-bearer never blocks
585 */
586static int tipc_bcbearer_send(struct sk_buff *buf, struct tipc_bearer *unused1,
587 struct tipc_media_addr *unused2)
588{
589 int bp_index;
590
591 /* Prepare broadcast link message for reliable transmission,
592 * if first time trying to send it;
593 * preparation is skipped for broadcast link protocol messages
594 * since they are sent in an unreliable manner and don't need it
595 */
596 if (likely(!msg_non_seq(buf_msg(buf)))) {
597 struct tipc_msg *msg;
598
599 bcbuf_set_acks(buf, bclink->bcast_nodes.count);
600 msg = buf_msg(buf);
601 msg_set_non_seq(msg, 1);
602 msg_set_mc_netid(msg, tipc_net_id);
603 bcl->stats.sent_info++;
604
605 if (WARN_ON(!bclink->bcast_nodes.count)) {
606 dump_stack();
607 return 0;
608 }
609 }
610
611 /* Send buffer over bearers until all targets reached */
612 bcbearer->remains = bclink->bcast_nodes;
613
614 for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
615 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
616 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
617 struct tipc_bearer *b = p;
618 struct sk_buff *tbuf;
619
620 if (!p)
621 break; /* No more bearers to try */
622
623 tipc_nmap_diff(&bcbearer->remains, &b->nodes,
624 &bcbearer->remains_new);
625 if (bcbearer->remains_new.count == bcbearer->remains.count)
626 continue; /* Nothing added by bearer pair */
627
628 if (bp_index == 0) {
629 /* Use original buffer for first bearer */
630 tipc_bearer_send(b, buf, &b->bcast_addr);
631 } else {
632 /* Avoid concurrent buffer access */
633 tbuf = pskb_copy(buf, GFP_ATOMIC);
634 if (!tbuf)
635 break;
636 tipc_bearer_send(b, tbuf, &b->bcast_addr);
637 kfree_skb(tbuf); /* Bearer keeps a clone */
638 }
639
640 /* Swap bearers for next packet */
641 if (s) {
642 bcbearer->bpairs[bp_index].primary = s;
643 bcbearer->bpairs[bp_index].secondary = p;
644 }
645
646 if (bcbearer->remains_new.count == 0)
647 break; /* All targets reached */
648
649 bcbearer->remains = bcbearer->remains_new;
650 }
651
652 return 0;
653}
654
655/**
656 * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
657 */
658void tipc_bcbearer_sort(void)
659{
660 struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
661 struct tipc_bcbearer_pair *bp_curr;
662 int b_index;
663 int pri;
664
665 spin_lock_bh(&bc_lock);
666
667 /* Group bearers by priority (can assume max of two per priority) */
668 memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
669
670 for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
671 struct tipc_bearer *b = bearer_list[b_index];
672 if (!b || !b->nodes.count)
673 continue;
674
675 if (!bp_temp[b->priority].primary)
676 bp_temp[b->priority].primary = b;
677 else
678 bp_temp[b->priority].secondary = b;
679 }
680
681 /* Create array of bearer pairs for broadcasting */
682 bp_curr = bcbearer->bpairs;
683 memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
684
685 for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
686
687 if (!bp_temp[pri].primary)
688 continue;
689
690 bp_curr->primary = bp_temp[pri].primary;
691
692 if (bp_temp[pri].secondary) {
693 if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
694 &bp_temp[pri].secondary->nodes)) {
695 bp_curr->secondary = bp_temp[pri].secondary;
696 } else {
697 bp_curr++;
698 bp_curr->primary = bp_temp[pri].secondary;
699 }
700 }
701
702 bp_curr++;
703 }
704
705 spin_unlock_bh(&bc_lock);
706}
707
708
709int tipc_bclink_stats(char *buf, const u32 buf_size)
710{
711 int ret;
712 struct tipc_stats *s;
713
714 if (!bcl)
715 return 0;
716
717 spin_lock_bh(&bc_lock);
718
719 s = &bcl->stats;
720
721 ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
722 " Window:%u packets\n",
723 bcl->name, bcl->queue_limit[0]);
724 ret += tipc_snprintf(buf + ret, buf_size - ret,
725 " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
726 s->recv_info, s->recv_fragments,
727 s->recv_fragmented, s->recv_bundles,
728 s->recv_bundled);
729 ret += tipc_snprintf(buf + ret, buf_size - ret,
730 " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
731 s->sent_info, s->sent_fragments,
732 s->sent_fragmented, s->sent_bundles,
733 s->sent_bundled);
734 ret += tipc_snprintf(buf + ret, buf_size - ret,
735 " RX naks:%u defs:%u dups:%u\n",
736 s->recv_nacks, s->deferred_recv, s->duplicates);
737 ret += tipc_snprintf(buf + ret, buf_size - ret,
738 " TX naks:%u acks:%u dups:%u\n",
739 s->sent_nacks, s->sent_acks, s->retransmitted);
740 ret += tipc_snprintf(buf + ret, buf_size - ret,
741 " Congestion link:%u Send queue max:%u avg:%u\n",
742 s->link_congs, s->max_queue_sz,
743 s->queue_sz_counts ?
744 (s->accu_queue_sz / s->queue_sz_counts) : 0);
745
746 spin_unlock_bh(&bc_lock);
747 return ret;
748}
749
750int tipc_bclink_reset_stats(void)
751{
752 if (!bcl)
753 return -ENOPROTOOPT;
754
755 spin_lock_bh(&bc_lock);
756 memset(&bcl->stats, 0, sizeof(bcl->stats));
757 spin_unlock_bh(&bc_lock);
758 return 0;
759}
760
761int tipc_bclink_set_queue_limits(u32 limit)
762{
763 if (!bcl)
764 return -ENOPROTOOPT;
765 if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
766 return -EINVAL;
767
768 spin_lock_bh(&bc_lock);
769 tipc_link_set_queue_limits(bcl, limit);
770 spin_unlock_bh(&bc_lock);
771 return 0;
772}
773
774void tipc_bclink_init(void)
775{
776 bcbearer->bearer.media = &bcbearer->media;
777 bcbearer->media.send_msg = tipc_bcbearer_send;
778 sprintf(bcbearer->media.name, "tipc-broadcast");
779
780 INIT_LIST_HEAD(&bcl->waiting_ports);
781 bcl->next_out_no = 1;
782 spin_lock_init(&bclink->node.lock);
783 bcl->owner = &bclink->node;
784 bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
785 tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
786 bcl->b_ptr = &bcbearer->bearer;
787 bearer_list[BCBEARER] = &bcbearer->bearer;
788 bcl->state = WORKING_WORKING;
789 strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
790}
791
792void tipc_bclink_stop(void)
793{
794 spin_lock_bh(&bc_lock);
795 tipc_link_purge_queues(bcl);
796 spin_unlock_bh(&bc_lock);
797
798 bearer_list[BCBEARER] = NULL;
799 memset(bclink, 0, sizeof(*bclink));
800 memset(bcbearer, 0, sizeof(*bcbearer));
801}
802
803
804/**
805 * tipc_nmap_add - add a node to a node map
806 */
807void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
808{
809 int n = tipc_node(node);
810 int w = n / WSIZE;
811 u32 mask = (1 << (n % WSIZE));
812
813 if ((nm_ptr->map[w] & mask) == 0) {
814 nm_ptr->count++;
815 nm_ptr->map[w] |= mask;
816 }
817}
818
819/**
820 * tipc_nmap_remove - remove a node from a node map
821 */
822void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
823{
824 int n = tipc_node(node);
825 int w = n / WSIZE;
826 u32 mask = (1 << (n % WSIZE));
827
828 if ((nm_ptr->map[w] & mask) != 0) {
829 nm_ptr->map[w] &= ~mask;
830 nm_ptr->count--;
831 }
832}
833
834/**
835 * tipc_nmap_diff - find differences between node maps
836 * @nm_a: input node map A
837 * @nm_b: input node map B
838 * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
839 */
840static void tipc_nmap_diff(struct tipc_node_map *nm_a,
841 struct tipc_node_map *nm_b,
842 struct tipc_node_map *nm_diff)
843{
844 int stop = ARRAY_SIZE(nm_a->map);
845 int w;
846 int b;
847 u32 map;
848
849 memset(nm_diff, 0, sizeof(*nm_diff));
850 for (w = 0; w < stop; w++) {
851 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
852 nm_diff->map[w] = map;
853 if (map != 0) {
854 for (b = 0 ; b < WSIZE; b++) {
855 if (map & (1 << b))
856 nm_diff->count++;
857 }
858 }
859 }
860}
861
862/**
863 * tipc_port_list_add - add a port to a port list, ensuring no duplicates
864 */
865void tipc_port_list_add(struct tipc_port_list *pl_ptr, u32 port)
866{
867 struct tipc_port_list *item = pl_ptr;
868 int i;
869 int item_sz = PLSIZE;
870 int cnt = pl_ptr->count;
871
872 for (; ; cnt -= item_sz, item = item->next) {
873 if (cnt < PLSIZE)
874 item_sz = cnt;
875 for (i = 0; i < item_sz; i++)
876 if (item->ports[i] == port)
877 return;
878 if (i < PLSIZE) {
879 item->ports[i] = port;
880 pl_ptr->count++;
881 return;
882 }
883 if (!item->next) {
884 item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
885 if (!item->next) {
886 pr_warn("Incomplete multicast delivery, no memory\n");
887 return;
888 }
889 item->next->next = NULL;
890 }
891 }
892}
893
894/**
895 * tipc_port_list_free - free dynamically created entries in port_list chain
896 *
897 */
898void tipc_port_list_free(struct tipc_port_list *pl_ptr)
899{
900 struct tipc_port_list *item;
901 struct tipc_port_list *next;
902
903 for (item = pl_ptr->next; item; item = next) {
904 next = item->next;
905 kfree(item);
906 }
907}