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#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}