Loading...
1/*
2 * net/tipc/bcast.c: TIPC broadcast code
3 *
4 * Copyright (c) 2004-2006, 2014-2017, 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 <linux/tipc_config.h>
39#include "socket.h"
40#include "msg.h"
41#include "bcast.h"
42#include "link.h"
43#include "name_table.h"
44
45#define BCLINK_WIN_DEFAULT 50 /* bcast link window size (default) */
46#define BCLINK_WIN_MIN 32 /* bcast minimum link window size */
47
48const char tipc_bclink_name[] = "broadcast-link";
49unsigned long sysctl_tipc_bc_retruni __read_mostly;
50
51/**
52 * struct tipc_bc_base - base structure for keeping broadcast send state
53 * @link: broadcast send link structure
54 * @inputq: data input queue; will only carry SOCK_WAKEUP messages
55 * @dests: array keeping number of reachable destinations per bearer
56 * @primary_bearer: a bearer having links to all broadcast destinations, if any
57 * @bcast_support: indicates if primary bearer, if any, supports broadcast
58 * @force_bcast: forces broadcast for multicast traffic
59 * @rcast_support: indicates if all peer nodes support replicast
60 * @force_rcast: forces replicast for multicast traffic
61 * @rc_ratio: dest count as percentage of cluster size where send method changes
62 * @bc_threshold: calculated from rc_ratio; if dests > threshold use broadcast
63 */
64struct tipc_bc_base {
65 struct tipc_link *link;
66 struct sk_buff_head inputq;
67 int dests[MAX_BEARERS];
68 int primary_bearer;
69 bool bcast_support;
70 bool force_bcast;
71 bool rcast_support;
72 bool force_rcast;
73 int rc_ratio;
74 int bc_threshold;
75};
76
77static struct tipc_bc_base *tipc_bc_base(struct net *net)
78{
79 return tipc_net(net)->bcbase;
80}
81
82/* tipc_bcast_get_mtu(): -get the MTU currently used by broadcast link
83 * Note: the MTU is decremented to give room for a tunnel header, in
84 * case the message needs to be sent as replicast
85 */
86int tipc_bcast_get_mtu(struct net *net)
87{
88 return tipc_link_mss(tipc_bc_sndlink(net));
89}
90
91void tipc_bcast_toggle_rcast(struct net *net, bool supp)
92{
93 tipc_bc_base(net)->rcast_support = supp;
94}
95
96static void tipc_bcbase_calc_bc_threshold(struct net *net)
97{
98 struct tipc_bc_base *bb = tipc_bc_base(net);
99 int cluster_size = tipc_link_bc_peers(tipc_bc_sndlink(net));
100
101 bb->bc_threshold = 1 + (cluster_size * bb->rc_ratio / 100);
102}
103
104/* tipc_bcbase_select_primary(): find a bearer with links to all destinations,
105 * if any, and make it primary bearer
106 */
107static void tipc_bcbase_select_primary(struct net *net)
108{
109 struct tipc_bc_base *bb = tipc_bc_base(net);
110 int all_dests = tipc_link_bc_peers(bb->link);
111 int i, mtu, prim;
112
113 bb->primary_bearer = INVALID_BEARER_ID;
114 bb->bcast_support = true;
115
116 if (!all_dests)
117 return;
118
119 for (i = 0; i < MAX_BEARERS; i++) {
120 if (!bb->dests[i])
121 continue;
122
123 mtu = tipc_bearer_mtu(net, i);
124 if (mtu < tipc_link_mtu(bb->link))
125 tipc_link_set_mtu(bb->link, mtu);
126 bb->bcast_support &= tipc_bearer_bcast_support(net, i);
127 if (bb->dests[i] < all_dests)
128 continue;
129
130 bb->primary_bearer = i;
131
132 /* Reduce risk that all nodes select same primary */
133 if ((i ^ tipc_own_addr(net)) & 1)
134 break;
135 }
136 prim = bb->primary_bearer;
137 if (prim != INVALID_BEARER_ID)
138 bb->bcast_support = tipc_bearer_bcast_support(net, prim);
139}
140
141void tipc_bcast_inc_bearer_dst_cnt(struct net *net, int bearer_id)
142{
143 struct tipc_bc_base *bb = tipc_bc_base(net);
144
145 tipc_bcast_lock(net);
146 bb->dests[bearer_id]++;
147 tipc_bcbase_select_primary(net);
148 tipc_bcast_unlock(net);
149}
150
151void tipc_bcast_dec_bearer_dst_cnt(struct net *net, int bearer_id)
152{
153 struct tipc_bc_base *bb = tipc_bc_base(net);
154
155 tipc_bcast_lock(net);
156 bb->dests[bearer_id]--;
157 tipc_bcbase_select_primary(net);
158 tipc_bcast_unlock(net);
159}
160
161/* tipc_bcbase_xmit - broadcast a packet queue across one or more bearers
162 *
163 * Note that number of reachable destinations, as indicated in the dests[]
164 * array, may transitionally differ from the number of destinations indicated
165 * in each sent buffer. We can sustain this. Excess destination nodes will
166 * drop and never acknowledge the unexpected packets, and missing destinations
167 * will either require retransmission (if they are just about to be added to
168 * the bearer), or be removed from the buffer's 'ackers' counter (if they
169 * just went down)
170 */
171static void tipc_bcbase_xmit(struct net *net, struct sk_buff_head *xmitq)
172{
173 int bearer_id;
174 struct tipc_bc_base *bb = tipc_bc_base(net);
175 struct sk_buff *skb, *_skb;
176 struct sk_buff_head _xmitq;
177
178 if (skb_queue_empty(xmitq))
179 return;
180
181 /* The typical case: at least one bearer has links to all nodes */
182 bearer_id = bb->primary_bearer;
183 if (bearer_id >= 0) {
184 tipc_bearer_bc_xmit(net, bearer_id, xmitq);
185 return;
186 }
187
188 /* We have to transmit across all bearers */
189 __skb_queue_head_init(&_xmitq);
190 for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
191 if (!bb->dests[bearer_id])
192 continue;
193
194 skb_queue_walk(xmitq, skb) {
195 _skb = pskb_copy_for_clone(skb, GFP_ATOMIC);
196 if (!_skb)
197 break;
198 __skb_queue_tail(&_xmitq, _skb);
199 }
200 tipc_bearer_bc_xmit(net, bearer_id, &_xmitq);
201 }
202 __skb_queue_purge(xmitq);
203 __skb_queue_purge(&_xmitq);
204}
205
206static void tipc_bcast_select_xmit_method(struct net *net, int dests,
207 struct tipc_mc_method *method)
208{
209 struct tipc_bc_base *bb = tipc_bc_base(net);
210 unsigned long exp = method->expires;
211
212 /* Broadcast supported by used bearer/bearers? */
213 if (!bb->bcast_support) {
214 method->rcast = true;
215 return;
216 }
217 /* Any destinations which don't support replicast ? */
218 if (!bb->rcast_support) {
219 method->rcast = false;
220 return;
221 }
222 /* Can current method be changed ? */
223 method->expires = jiffies + TIPC_METHOD_EXPIRE;
224 if (method->mandatory)
225 return;
226
227 if (!(tipc_net(net)->capabilities & TIPC_MCAST_RBCTL) &&
228 time_before(jiffies, exp))
229 return;
230
231 /* Configuration as force 'broadcast' method */
232 if (bb->force_bcast) {
233 method->rcast = false;
234 return;
235 }
236 /* Configuration as force 'replicast' method */
237 if (bb->force_rcast) {
238 method->rcast = true;
239 return;
240 }
241 /* Configuration as 'autoselect' or default method */
242 /* Determine method to use now */
243 method->rcast = dests <= bb->bc_threshold;
244}
245
246/* tipc_bcast_xmit - broadcast the buffer chain to all external nodes
247 * @net: the applicable net namespace
248 * @pkts: chain of buffers containing message
249 * @cong_link_cnt: set to 1 if broadcast link is congested, otherwise 0
250 * Consumes the buffer chain.
251 * Returns 0 if success, otherwise errno: -EHOSTUNREACH,-EMSGSIZE
252 */
253int tipc_bcast_xmit(struct net *net, struct sk_buff_head *pkts,
254 u16 *cong_link_cnt)
255{
256 struct tipc_link *l = tipc_bc_sndlink(net);
257 struct sk_buff_head xmitq;
258 int rc = 0;
259
260 __skb_queue_head_init(&xmitq);
261 tipc_bcast_lock(net);
262 if (tipc_link_bc_peers(l))
263 rc = tipc_link_xmit(l, pkts, &xmitq);
264 tipc_bcast_unlock(net);
265 tipc_bcbase_xmit(net, &xmitq);
266 __skb_queue_purge(pkts);
267 if (rc == -ELINKCONG) {
268 *cong_link_cnt = 1;
269 rc = 0;
270 }
271 return rc;
272}
273
274/* tipc_rcast_xmit - replicate and send a message to given destination nodes
275 * @net: the applicable net namespace
276 * @pkts: chain of buffers containing message
277 * @dests: list of destination nodes
278 * @cong_link_cnt: returns number of congested links
279 * @cong_links: returns identities of congested links
280 * Returns 0 if success, otherwise errno
281 */
282static int tipc_rcast_xmit(struct net *net, struct sk_buff_head *pkts,
283 struct tipc_nlist *dests, u16 *cong_link_cnt)
284{
285 struct tipc_dest *dst, *tmp;
286 struct sk_buff_head _pkts;
287 u32 dnode, selector;
288
289 selector = msg_link_selector(buf_msg(skb_peek(pkts)));
290 __skb_queue_head_init(&_pkts);
291
292 list_for_each_entry_safe(dst, tmp, &dests->list, list) {
293 dnode = dst->node;
294 if (!tipc_msg_pskb_copy(dnode, pkts, &_pkts))
295 return -ENOMEM;
296
297 /* Any other return value than -ELINKCONG is ignored */
298 if (tipc_node_xmit(net, &_pkts, dnode, selector) == -ELINKCONG)
299 (*cong_link_cnt)++;
300 }
301 return 0;
302}
303
304/* tipc_mcast_send_sync - deliver a dummy message with SYN bit
305 * @net: the applicable net namespace
306 * @skb: socket buffer to copy
307 * @method: send method to be used
308 * @dests: destination nodes for message.
309 * Returns 0 if success, otherwise errno
310 */
311static int tipc_mcast_send_sync(struct net *net, struct sk_buff *skb,
312 struct tipc_mc_method *method,
313 struct tipc_nlist *dests)
314{
315 struct tipc_msg *hdr, *_hdr;
316 struct sk_buff_head tmpq;
317 struct sk_buff *_skb;
318 u16 cong_link_cnt;
319 int rc = 0;
320
321 /* Is a cluster supporting with new capabilities ? */
322 if (!(tipc_net(net)->capabilities & TIPC_MCAST_RBCTL))
323 return 0;
324
325 hdr = buf_msg(skb);
326 if (msg_user(hdr) == MSG_FRAGMENTER)
327 hdr = msg_inner_hdr(hdr);
328 if (msg_type(hdr) != TIPC_MCAST_MSG)
329 return 0;
330
331 /* Allocate dummy message */
332 _skb = tipc_buf_acquire(MCAST_H_SIZE, GFP_KERNEL);
333 if (!_skb)
334 return -ENOMEM;
335
336 /* Preparing for 'synching' header */
337 msg_set_syn(hdr, 1);
338
339 /* Copy skb's header into a dummy header */
340 skb_copy_to_linear_data(_skb, hdr, MCAST_H_SIZE);
341 skb_orphan(_skb);
342
343 /* Reverse method for dummy message */
344 _hdr = buf_msg(_skb);
345 msg_set_size(_hdr, MCAST_H_SIZE);
346 msg_set_is_rcast(_hdr, !msg_is_rcast(hdr));
347 msg_set_errcode(_hdr, TIPC_ERR_NO_PORT);
348
349 __skb_queue_head_init(&tmpq);
350 __skb_queue_tail(&tmpq, _skb);
351 if (method->rcast)
352 rc = tipc_bcast_xmit(net, &tmpq, &cong_link_cnt);
353 else
354 rc = tipc_rcast_xmit(net, &tmpq, dests, &cong_link_cnt);
355
356 /* This queue should normally be empty by now */
357 __skb_queue_purge(&tmpq);
358
359 return rc;
360}
361
362/* tipc_mcast_xmit - deliver message to indicated destination nodes
363 * and to identified node local sockets
364 * @net: the applicable net namespace
365 * @pkts: chain of buffers containing message
366 * @method: send method to be used
367 * @dests: destination nodes for message.
368 * @cong_link_cnt: returns number of encountered congested destination links
369 * Consumes buffer chain.
370 * Returns 0 if success, otherwise errno
371 */
372int tipc_mcast_xmit(struct net *net, struct sk_buff_head *pkts,
373 struct tipc_mc_method *method, struct tipc_nlist *dests,
374 u16 *cong_link_cnt)
375{
376 struct sk_buff_head inputq, localq;
377 bool rcast = method->rcast;
378 struct tipc_msg *hdr;
379 struct sk_buff *skb;
380 int rc = 0;
381
382 skb_queue_head_init(&inputq);
383 __skb_queue_head_init(&localq);
384
385 /* Clone packets before they are consumed by next call */
386 if (dests->local && !tipc_msg_reassemble(pkts, &localq)) {
387 rc = -ENOMEM;
388 goto exit;
389 }
390 /* Send according to determined transmit method */
391 if (dests->remote) {
392 tipc_bcast_select_xmit_method(net, dests->remote, method);
393
394 skb = skb_peek(pkts);
395 hdr = buf_msg(skb);
396 if (msg_user(hdr) == MSG_FRAGMENTER)
397 hdr = msg_inner_hdr(hdr);
398 msg_set_is_rcast(hdr, method->rcast);
399
400 /* Switch method ? */
401 if (rcast != method->rcast) {
402 rc = tipc_mcast_send_sync(net, skb, method, dests);
403 if (unlikely(rc)) {
404 pr_err("Unable to send SYN: method %d, rc %d\n",
405 rcast, rc);
406 goto exit;
407 }
408 }
409
410 if (method->rcast)
411 rc = tipc_rcast_xmit(net, pkts, dests, cong_link_cnt);
412 else
413 rc = tipc_bcast_xmit(net, pkts, cong_link_cnt);
414 }
415
416 if (dests->local) {
417 tipc_loopback_trace(net, &localq);
418 tipc_sk_mcast_rcv(net, &localq, &inputq);
419 }
420exit:
421 /* This queue should normally be empty by now */
422 __skb_queue_purge(pkts);
423 return rc;
424}
425
426/* tipc_bcast_rcv - receive a broadcast packet, and deliver to rcv link
427 *
428 * RCU is locked, no other locks set
429 */
430int tipc_bcast_rcv(struct net *net, struct tipc_link *l, struct sk_buff *skb)
431{
432 struct tipc_msg *hdr = buf_msg(skb);
433 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
434 struct sk_buff_head xmitq;
435 int rc;
436
437 __skb_queue_head_init(&xmitq);
438
439 if (msg_mc_netid(hdr) != tipc_netid(net) || !tipc_link_is_up(l)) {
440 kfree_skb(skb);
441 return 0;
442 }
443
444 tipc_bcast_lock(net);
445 if (msg_user(hdr) == BCAST_PROTOCOL)
446 rc = tipc_link_bc_nack_rcv(l, skb, &xmitq);
447 else
448 rc = tipc_link_rcv(l, skb, NULL);
449 tipc_bcast_unlock(net);
450
451 tipc_bcbase_xmit(net, &xmitq);
452
453 /* Any socket wakeup messages ? */
454 if (!skb_queue_empty(inputq))
455 tipc_sk_rcv(net, inputq);
456
457 return rc;
458}
459
460/* tipc_bcast_ack_rcv - receive and handle a broadcast acknowledge
461 *
462 * RCU is locked, no other locks set
463 */
464void tipc_bcast_ack_rcv(struct net *net, struct tipc_link *l,
465 struct tipc_msg *hdr)
466{
467 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
468 u16 acked = msg_bcast_ack(hdr);
469 struct sk_buff_head xmitq;
470
471 /* Ignore bc acks sent by peer before bcast synch point was received */
472 if (msg_bc_ack_invalid(hdr))
473 return;
474
475 __skb_queue_head_init(&xmitq);
476
477 tipc_bcast_lock(net);
478 tipc_link_bc_ack_rcv(l, acked, 0, NULL, &xmitq, NULL);
479 tipc_bcast_unlock(net);
480
481 tipc_bcbase_xmit(net, &xmitq);
482
483 /* Any socket wakeup messages ? */
484 if (!skb_queue_empty(inputq))
485 tipc_sk_rcv(net, inputq);
486}
487
488/* tipc_bcast_synch_rcv - check and update rcv link with peer's send state
489 *
490 * RCU is locked, no other locks set
491 */
492int tipc_bcast_sync_rcv(struct net *net, struct tipc_link *l,
493 struct tipc_msg *hdr,
494 struct sk_buff_head *retrq)
495{
496 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
497 struct tipc_gap_ack_blks *ga;
498 struct sk_buff_head xmitq;
499 int rc = 0;
500
501 __skb_queue_head_init(&xmitq);
502
503 tipc_bcast_lock(net);
504 if (msg_type(hdr) != STATE_MSG) {
505 tipc_link_bc_init_rcv(l, hdr);
506 } else if (!msg_bc_ack_invalid(hdr)) {
507 tipc_get_gap_ack_blks(&ga, l, hdr, false);
508 if (!sysctl_tipc_bc_retruni)
509 retrq = &xmitq;
510 rc = tipc_link_bc_ack_rcv(l, msg_bcast_ack(hdr),
511 msg_bc_gap(hdr), ga, &xmitq,
512 retrq);
513 rc |= tipc_link_bc_sync_rcv(l, hdr, &xmitq);
514 }
515 tipc_bcast_unlock(net);
516
517 tipc_bcbase_xmit(net, &xmitq);
518
519 /* Any socket wakeup messages ? */
520 if (!skb_queue_empty(inputq))
521 tipc_sk_rcv(net, inputq);
522 return rc;
523}
524
525/* tipc_bcast_add_peer - add a peer node to broadcast link and bearer
526 *
527 * RCU is locked, node lock is set
528 */
529void tipc_bcast_add_peer(struct net *net, struct tipc_link *uc_l,
530 struct sk_buff_head *xmitq)
531{
532 struct tipc_link *snd_l = tipc_bc_sndlink(net);
533
534 tipc_bcast_lock(net);
535 tipc_link_add_bc_peer(snd_l, uc_l, xmitq);
536 tipc_bcbase_select_primary(net);
537 tipc_bcbase_calc_bc_threshold(net);
538 tipc_bcast_unlock(net);
539}
540
541/* tipc_bcast_remove_peer - remove a peer node from broadcast link and bearer
542 *
543 * RCU is locked, node lock is set
544 */
545void tipc_bcast_remove_peer(struct net *net, struct tipc_link *rcv_l)
546{
547 struct tipc_link *snd_l = tipc_bc_sndlink(net);
548 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
549 struct sk_buff_head xmitq;
550
551 __skb_queue_head_init(&xmitq);
552
553 tipc_bcast_lock(net);
554 tipc_link_remove_bc_peer(snd_l, rcv_l, &xmitq);
555 tipc_bcbase_select_primary(net);
556 tipc_bcbase_calc_bc_threshold(net);
557 tipc_bcast_unlock(net);
558
559 tipc_bcbase_xmit(net, &xmitq);
560
561 /* Any socket wakeup messages ? */
562 if (!skb_queue_empty(inputq))
563 tipc_sk_rcv(net, inputq);
564}
565
566int tipc_bclink_reset_stats(struct net *net, struct tipc_link *l)
567{
568 if (!l)
569 return -ENOPROTOOPT;
570
571 tipc_bcast_lock(net);
572 tipc_link_reset_stats(l);
573 tipc_bcast_unlock(net);
574 return 0;
575}
576
577static int tipc_bc_link_set_queue_limits(struct net *net, u32 max_win)
578{
579 struct tipc_link *l = tipc_bc_sndlink(net);
580
581 if (!l)
582 return -ENOPROTOOPT;
583 if (max_win < BCLINK_WIN_MIN)
584 max_win = BCLINK_WIN_MIN;
585 if (max_win > TIPC_MAX_LINK_WIN)
586 return -EINVAL;
587 tipc_bcast_lock(net);
588 tipc_link_set_queue_limits(l, BCLINK_WIN_MIN, max_win);
589 tipc_bcast_unlock(net);
590 return 0;
591}
592
593static int tipc_bc_link_set_broadcast_mode(struct net *net, u32 bc_mode)
594{
595 struct tipc_bc_base *bb = tipc_bc_base(net);
596
597 switch (bc_mode) {
598 case BCLINK_MODE_BCAST:
599 if (!bb->bcast_support)
600 return -ENOPROTOOPT;
601
602 bb->force_bcast = true;
603 bb->force_rcast = false;
604 break;
605 case BCLINK_MODE_RCAST:
606 if (!bb->rcast_support)
607 return -ENOPROTOOPT;
608
609 bb->force_bcast = false;
610 bb->force_rcast = true;
611 break;
612 case BCLINK_MODE_SEL:
613 if (!bb->bcast_support || !bb->rcast_support)
614 return -ENOPROTOOPT;
615
616 bb->force_bcast = false;
617 bb->force_rcast = false;
618 break;
619 default:
620 return -EINVAL;
621 }
622
623 return 0;
624}
625
626static int tipc_bc_link_set_broadcast_ratio(struct net *net, u32 bc_ratio)
627{
628 struct tipc_bc_base *bb = tipc_bc_base(net);
629
630 if (!bb->bcast_support || !bb->rcast_support)
631 return -ENOPROTOOPT;
632
633 if (bc_ratio > 100 || bc_ratio <= 0)
634 return -EINVAL;
635
636 bb->rc_ratio = bc_ratio;
637 tipc_bcast_lock(net);
638 tipc_bcbase_calc_bc_threshold(net);
639 tipc_bcast_unlock(net);
640
641 return 0;
642}
643
644int tipc_nl_bc_link_set(struct net *net, struct nlattr *attrs[])
645{
646 int err;
647 u32 win;
648 u32 bc_mode;
649 u32 bc_ratio;
650 struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
651
652 if (!attrs[TIPC_NLA_LINK_PROP])
653 return -EINVAL;
654
655 err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP], props);
656 if (err)
657 return err;
658
659 if (!props[TIPC_NLA_PROP_WIN] &&
660 !props[TIPC_NLA_PROP_BROADCAST] &&
661 !props[TIPC_NLA_PROP_BROADCAST_RATIO]) {
662 return -EOPNOTSUPP;
663 }
664
665 if (props[TIPC_NLA_PROP_BROADCAST]) {
666 bc_mode = nla_get_u32(props[TIPC_NLA_PROP_BROADCAST]);
667 err = tipc_bc_link_set_broadcast_mode(net, bc_mode);
668 }
669
670 if (!err && props[TIPC_NLA_PROP_BROADCAST_RATIO]) {
671 bc_ratio = nla_get_u32(props[TIPC_NLA_PROP_BROADCAST_RATIO]);
672 err = tipc_bc_link_set_broadcast_ratio(net, bc_ratio);
673 }
674
675 if (!err && props[TIPC_NLA_PROP_WIN]) {
676 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
677 err = tipc_bc_link_set_queue_limits(net, win);
678 }
679
680 return err;
681}
682
683int tipc_bcast_init(struct net *net)
684{
685 struct tipc_net *tn = tipc_net(net);
686 struct tipc_bc_base *bb = NULL;
687 struct tipc_link *l = NULL;
688
689 bb = kzalloc(sizeof(*bb), GFP_KERNEL);
690 if (!bb)
691 goto enomem;
692 tn->bcbase = bb;
693 spin_lock_init(&tipc_net(net)->bclock);
694
695 if (!tipc_link_bc_create(net, 0, 0, NULL,
696 FB_MTU,
697 BCLINK_WIN_DEFAULT,
698 BCLINK_WIN_DEFAULT,
699 0,
700 &bb->inputq,
701 NULL,
702 NULL,
703 &l))
704 goto enomem;
705 bb->link = l;
706 tn->bcl = l;
707 bb->rc_ratio = 10;
708 bb->rcast_support = true;
709 return 0;
710enomem:
711 kfree(bb);
712 kfree(l);
713 return -ENOMEM;
714}
715
716void tipc_bcast_stop(struct net *net)
717{
718 struct tipc_net *tn = net_generic(net, tipc_net_id);
719
720 synchronize_net();
721 kfree(tn->bcbase);
722 kfree(tn->bcl);
723}
724
725void tipc_nlist_init(struct tipc_nlist *nl, u32 self)
726{
727 memset(nl, 0, sizeof(*nl));
728 INIT_LIST_HEAD(&nl->list);
729 nl->self = self;
730}
731
732void tipc_nlist_add(struct tipc_nlist *nl, u32 node)
733{
734 if (node == nl->self)
735 nl->local = true;
736 else if (tipc_dest_push(&nl->list, node, 0))
737 nl->remote++;
738}
739
740void tipc_nlist_del(struct tipc_nlist *nl, u32 node)
741{
742 if (node == nl->self)
743 nl->local = false;
744 else if (tipc_dest_del(&nl->list, node, 0))
745 nl->remote--;
746}
747
748void tipc_nlist_purge(struct tipc_nlist *nl)
749{
750 tipc_dest_list_purge(&nl->list);
751 nl->remote = 0;
752 nl->local = false;
753}
754
755u32 tipc_bcast_get_mode(struct net *net)
756{
757 struct tipc_bc_base *bb = tipc_bc_base(net);
758
759 if (bb->force_bcast)
760 return BCLINK_MODE_BCAST;
761
762 if (bb->force_rcast)
763 return BCLINK_MODE_RCAST;
764
765 if (bb->bcast_support && bb->rcast_support)
766 return BCLINK_MODE_SEL;
767
768 return 0;
769}
770
771u32 tipc_bcast_get_broadcast_ratio(struct net *net)
772{
773 struct tipc_bc_base *bb = tipc_bc_base(net);
774
775 return bb->rc_ratio;
776}
777
778void tipc_mcast_filter_msg(struct net *net, struct sk_buff_head *defq,
779 struct sk_buff_head *inputq)
780{
781 struct sk_buff *skb, *_skb, *tmp;
782 struct tipc_msg *hdr, *_hdr;
783 bool match = false;
784 u32 node, port;
785
786 skb = skb_peek(inputq);
787 if (!skb)
788 return;
789
790 hdr = buf_msg(skb);
791
792 if (likely(!msg_is_syn(hdr) && skb_queue_empty(defq)))
793 return;
794
795 node = msg_orignode(hdr);
796 if (node == tipc_own_addr(net))
797 return;
798
799 port = msg_origport(hdr);
800
801 /* Has the twin SYN message already arrived ? */
802 skb_queue_walk(defq, _skb) {
803 _hdr = buf_msg(_skb);
804 if (msg_orignode(_hdr) != node)
805 continue;
806 if (msg_origport(_hdr) != port)
807 continue;
808 match = true;
809 break;
810 }
811
812 if (!match) {
813 if (!msg_is_syn(hdr))
814 return;
815 __skb_dequeue(inputq);
816 __skb_queue_tail(defq, skb);
817 return;
818 }
819
820 /* Deliver non-SYN message from other link, otherwise queue it */
821 if (!msg_is_syn(hdr)) {
822 if (msg_is_rcast(hdr) != msg_is_rcast(_hdr))
823 return;
824 __skb_dequeue(inputq);
825 __skb_queue_tail(defq, skb);
826 return;
827 }
828
829 /* Queue non-SYN/SYN message from same link */
830 if (msg_is_rcast(hdr) == msg_is_rcast(_hdr)) {
831 __skb_dequeue(inputq);
832 __skb_queue_tail(defq, skb);
833 return;
834 }
835
836 /* Matching SYN messages => return the one with data, if any */
837 __skb_unlink(_skb, defq);
838 if (msg_data_sz(hdr)) {
839 kfree_skb(_skb);
840 } else {
841 __skb_dequeue(inputq);
842 kfree_skb(skb);
843 __skb_queue_tail(inputq, _skb);
844 }
845
846 /* Deliver subsequent non-SYN messages from same peer */
847 skb_queue_walk_safe(defq, _skb, tmp) {
848 _hdr = buf_msg(_skb);
849 if (msg_orignode(_hdr) != node)
850 continue;
851 if (msg_origport(_hdr) != port)
852 continue;
853 if (msg_is_syn(_hdr))
854 break;
855 __skb_unlink(_skb, defq);
856 __skb_queue_tail(inputq, _skb);
857 }
858}
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