2 * net/tipc/bcast.c: TIPC broadcast code
4 * Copyright (c) 2004-2006, 2014-2015, Ericsson AB
5 * Copyright (c) 2004, Intel Corporation.
6 * Copyright (c) 2005, 2010-2011, Wind River Systems
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions are met:
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.
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.
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.
38 #include <linux/tipc_config.h>
42 #include "name_distr.h"
46 #define MAX_PKT_DEFAULT_MCAST 1500 /* bcast link max packet size (fixed) */
47 #define BCLINK_WIN_DEFAULT 50 /* bcast link window size (default) */
48 #define BCLINK_WIN_MIN 32 /* bcast minimum link window size */
50 const char tipc_bclink_name[] = "broadcast-link";
53 * struct tipc_bcbearer_pair - a pair of bearers used by broadcast link
54 * @primary: pointer to primary bearer
55 * @secondary: pointer to secondary bearer
57 * Bearers must have same priority and same set of reachable destinations
61 struct tipc_bcbearer_pair {
62 struct tipc_bearer *primary;
63 struct tipc_bearer *secondary;
66 #define BCBEARER MAX_BEARERS
69 * struct tipc_bcbearer - bearer used by broadcast link
70 * @bearer: (non-standard) broadcast bearer structure
71 * @media: (non-standard) broadcast media structure
72 * @bpairs: array of bearer pairs
73 * @bpairs_temp: temporary array of bearer pairs used by tipc_bcbearer_sort()
74 * @remains: temporary node map used by tipc_bcbearer_send()
75 * @remains_new: temporary node map used tipc_bcbearer_send()
77 * Note: The fields labelled "temporary" are incorporated into the bearer
78 * to avoid consuming potentially limited stack space through the use of
79 * large local variables within multicast routines. Concurrent access is
80 * prevented through use of the spinlock "bcast_lock".
82 struct tipc_bcbearer {
83 struct tipc_bearer bearer;
84 struct tipc_media media;
85 struct tipc_bcbearer_pair bpairs[MAX_BEARERS];
86 struct tipc_bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1];
87 struct tipc_node_map remains;
88 struct tipc_node_map remains_new;
92 * struct tipc_bc_base - link used for broadcast messages
93 * @link: broadcast send link structure
94 * @node: (non-standard) node structure representing b'cast link's peer node
95 * @bcast_nodes: map of broadcast-capable nodes
96 * @retransmit_to: node that most recently requested a retransmit
97 * @dest_nnt: array indicating number of reachable destinations per bearer
98 * @bearers: array of bearers, sorted by number of reachable destinations
100 * Handles sequence numbering, fragmentation, bundling, etc.
102 struct tipc_bc_base {
103 struct tipc_link *link;
104 struct tipc_node node;
105 struct sk_buff_head arrvq;
106 struct sk_buff_head inputq;
107 struct sk_buff_head namedq;
108 int dests[MAX_BEARERS];
110 struct tipc_node_map bcast_nodes;
111 struct tipc_node *retransmit_to;
114 static struct tipc_bc_base *tipc_bc_base(struct net *net)
116 return tipc_net(net)->bcbase;
120 * tipc_nmap_equal - test for equality of node maps
122 static int tipc_nmap_equal(struct tipc_node_map *nm_a,
123 struct tipc_node_map *nm_b)
125 return !memcmp(nm_a, nm_b, sizeof(*nm_a));
128 static void tipc_bcbearer_xmit(struct net *net, struct sk_buff_head *xmitq);
129 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
130 struct tipc_node_map *nm_b,
131 struct tipc_node_map *nm_diff);
132 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node);
133 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node);
134 static void tipc_bclink_lock(struct net *net)
136 tipc_bcast_lock(net);
139 static void tipc_bclink_unlock(struct net *net)
141 tipc_bcast_unlock(net);
144 void tipc_bclink_input(struct net *net)
146 struct tipc_net *tn = net_generic(net, tipc_net_id);
148 tipc_sk_mcast_rcv(net, &tn->bcbase->arrvq, &tn->bcbase->inputq);
151 uint tipc_bcast_get_mtu(void)
153 return MAX_PKT_DEFAULT_MCAST;
156 static u16 bcbuf_acks(struct sk_buff *skb)
158 return TIPC_SKB_CB(skb)->ackers;
161 static void bcbuf_set_acks(struct sk_buff *buf, u16 ackers)
163 TIPC_SKB_CB(buf)->ackers = ackers;
166 static void bcbuf_decr_acks(struct sk_buff *buf)
168 bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
171 /* tipc_bcbase_select_primary(): find a bearer with links to all destinations,
172 * if any, and make it primary bearer
174 static void tipc_bcbase_select_primary(struct net *net)
176 struct tipc_bc_base *bb = tipc_bc_base(net);
177 int all_dests = tipc_link_bc_peers(bb->link);
180 bb->primary_bearer = INVALID_BEARER_ID;
185 for (i = 0; i < MAX_BEARERS; i++) {
186 if (bb->dests[i] < all_dests)
189 bb->primary_bearer = i;
191 /* Reduce risk that all nodes select same primary */
192 if ((i ^ tipc_own_addr(net)) & 1)
197 void tipc_bcast_inc_bearer_dst_cnt(struct net *net, int bearer_id)
199 struct tipc_bc_base *bb = tipc_bc_base(net);
201 tipc_bcast_lock(net);
202 bb->dests[bearer_id]++;
203 tipc_bcbase_select_primary(net);
204 tipc_bcast_unlock(net);
207 void tipc_bcast_dec_bearer_dst_cnt(struct net *net, int bearer_id)
209 struct tipc_bc_base *bb = tipc_bc_base(net);
211 tipc_bcast_lock(net);
212 bb->dests[bearer_id]--;
213 tipc_bcbase_select_primary(net);
214 tipc_bcast_unlock(net);
217 static void bclink_set_last_sent(struct net *net)
219 struct tipc_net *tn = net_generic(net, tipc_net_id);
220 struct tipc_link *bcl = tn->bcl;
222 bcl->silent_intv_cnt = mod(bcl->snd_nxt - 1);
225 u32 tipc_bclink_get_last_sent(struct net *net)
227 struct tipc_net *tn = net_generic(net, tipc_net_id);
229 return tn->bcl->silent_intv_cnt;
232 static void bclink_update_last_sent(struct tipc_node *node, u32 seqno)
234 node->bclink.last_sent = less_eq(node->bclink.last_sent, seqno) ?
235 seqno : node->bclink.last_sent;
239 * tipc_bclink_retransmit_to - get most recent node to request retransmission
241 * Called with bclink_lock locked
243 struct tipc_node *tipc_bclink_retransmit_to(struct net *net)
245 struct tipc_net *tn = net_generic(net, tipc_net_id);
247 return tn->bcbase->retransmit_to;
251 * bclink_retransmit_pkt - retransmit broadcast packets
252 * @after: sequence number of last packet to *not* retransmit
253 * @to: sequence number of last packet to retransmit
255 * Called with bclink_lock locked
257 static void bclink_retransmit_pkt(struct tipc_net *tn, u32 after, u32 to)
260 struct tipc_link *bcl = tn->bcl;
262 skb_queue_walk(&bcl->transmq, skb) {
263 if (more(buf_seqno(skb), after)) {
264 tipc_link_retransmit(bcl, skb, mod(to - after));
271 * bclink_prepare_wakeup - prepare users for wakeup after congestion
272 * @bcl: broadcast link
273 * @resultq: queue for users which can be woken up
274 * Move a number of waiting users, as permitted by available space in
275 * the send queue, from link wait queue to specified queue for wakeup
277 static void bclink_prepare_wakeup(struct tipc_link *bcl, struct sk_buff_head *resultq)
279 int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
281 struct sk_buff *skb, *tmp;
283 skb_queue_walk_safe(&bcl->wakeupq, skb, tmp) {
284 imp = TIPC_SKB_CB(skb)->chain_imp;
285 lim = bcl->window + bcl->backlog[imp].limit;
286 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
287 if ((pnd[imp] + bcl->backlog[imp].len) >= lim)
289 skb_unlink(skb, &bcl->wakeupq);
290 skb_queue_tail(resultq, skb);
295 * tipc_bclink_wakeup_users - wake up pending users
297 * Called with no locks taken
299 void tipc_bclink_wakeup_users(struct net *net)
301 struct tipc_net *tn = net_generic(net, tipc_net_id);
302 struct tipc_link *bcl = tn->bcl;
303 struct sk_buff_head resultq;
305 skb_queue_head_init(&resultq);
306 bclink_prepare_wakeup(bcl, &resultq);
307 tipc_sk_rcv(net, &resultq);
311 * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
312 * @n_ptr: node that sent acknowledgement info
313 * @acked: broadcast sequence # that has been acknowledged
315 * Node is locked, bclink_lock unlocked.
317 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
319 struct sk_buff *skb, *tmp;
320 unsigned int released = 0;
321 struct net *net = n_ptr->net;
322 struct tipc_net *tn = net_generic(net, tipc_net_id);
324 if (unlikely(!n_ptr->bclink.recv_permitted))
326 tipc_bclink_lock(net);
328 /* Bail out if tx queue is empty (no clean up is required) */
329 skb = skb_peek(&tn->bcl->transmq);
333 /* Determine which messages need to be acknowledged */
334 if (acked == INVALID_LINK_SEQ) {
336 * Contact with specified node has been lost, so need to
337 * acknowledge sent messages only (if other nodes still exist)
338 * or both sent and unsent messages (otherwise)
340 if (tn->bcbase->bcast_nodes.count)
341 acked = tn->bcl->silent_intv_cnt;
343 acked = tn->bcl->snd_nxt;
346 * Bail out if specified sequence number does not correspond
347 * to a message that has been sent and not yet acknowledged
349 if (less(acked, buf_seqno(skb)) ||
350 less(tn->bcl->silent_intv_cnt, acked) ||
351 less_eq(acked, n_ptr->bclink.acked))
354 /* Skip over packets that node has previously acknowledged */
355 skb_queue_walk(&tn->bcl->transmq, skb) {
356 if (more(buf_seqno(skb), n_ptr->bclink.acked))
359 /* Update packets that node is now acknowledging */
360 skb_queue_walk_from_safe(&tn->bcl->transmq, skb, tmp) {
361 if (more(buf_seqno(skb), acked))
363 bcbuf_decr_acks(skb);
364 bclink_set_last_sent(net);
365 if (bcbuf_acks(skb) == 0) {
366 __skb_unlink(skb, &tn->bcl->transmq);
371 n_ptr->bclink.acked = acked;
373 /* Try resolving broadcast link congestion, if necessary */
374 if (unlikely(skb_peek(&tn->bcl->backlogq))) {
375 tipc_link_push_packets(tn->bcl);
376 bclink_set_last_sent(net);
378 if (unlikely(released && !skb_queue_empty(&tn->bcl->wakeupq)))
379 n_ptr->action_flags |= TIPC_WAKEUP_BCAST_USERS;
381 tipc_bclink_unlock(net);
385 * tipc_bclink_update_link_state - update broadcast link state
387 * RCU and node lock set
389 void tipc_bclink_update_link_state(struct tipc_node *n_ptr,
393 struct net *net = n_ptr->net;
394 struct tipc_net *tn = net_generic(net, tipc_net_id);
395 struct tipc_link *bcl = tn->bcl;
397 /* Ignore "stale" link state info */
398 if (less_eq(last_sent, n_ptr->bclink.last_in))
401 /* Update link synchronization state; quit if in sync */
402 bclink_update_last_sent(n_ptr, last_sent);
404 /* This is a good location for statistical profiling */
405 bcl->stats.queue_sz_counts++;
406 bcl->stats.accu_queue_sz += skb_queue_len(&bcl->transmq);
408 if (n_ptr->bclink.last_sent == n_ptr->bclink.last_in)
411 /* Update out-of-sync state; quit if loss is still unconfirmed */
412 if ((++n_ptr->bclink.oos_state) == 1) {
413 if (n_ptr->bclink.deferred_size < (TIPC_MIN_LINK_WIN / 2))
415 n_ptr->bclink.oos_state++;
418 /* Don't NACK if one has been recently sent (or seen) */
419 if (n_ptr->bclink.oos_state & 0x1)
423 buf = tipc_buf_acquire(INT_H_SIZE);
425 struct tipc_msg *msg = buf_msg(buf);
426 struct sk_buff *skb = skb_peek(&n_ptr->bclink.deferdq);
427 u32 to = skb ? buf_seqno(skb) - 1 : n_ptr->bclink.last_sent;
429 tipc_msg_init(tn->own_addr, msg, BCAST_PROTOCOL, STATE_MSG,
430 INT_H_SIZE, n_ptr->addr);
431 msg_set_non_seq(msg, 1);
432 msg_set_mc_netid(msg, tn->net_id);
433 msg_set_bcast_ack(msg, n_ptr->bclink.last_in);
434 msg_set_bcgap_after(msg, n_ptr->bclink.last_in);
435 msg_set_bcgap_to(msg, to);
437 tipc_bclink_lock(net);
438 tipc_bearer_send(net, MAX_BEARERS, buf, NULL);
439 tn->bcl->stats.sent_nacks++;
440 tipc_bclink_unlock(net);
443 n_ptr->bclink.oos_state++;
447 void tipc_bclink_sync_state(struct tipc_node *n, struct tipc_msg *hdr)
449 u16 last = msg_last_bcast(hdr);
450 int mtyp = msg_type(hdr);
452 if (unlikely(msg_user(hdr) != LINK_PROTOCOL))
454 if (mtyp == STATE_MSG) {
455 tipc_bclink_update_link_state(n, last);
458 /* Compatibility: older nodes don't know BCAST_PROTOCOL synchronization,
459 * and transfer synch info in LINK_PROTOCOL messages.
461 if (tipc_node_is_up(n))
463 if ((mtyp != RESET_MSG) && (mtyp != ACTIVATE_MSG))
465 n->bclink.last_sent = last;
466 n->bclink.last_in = last;
467 n->bclink.oos_state = 0;
471 * bclink_peek_nack - monitor retransmission requests sent by other nodes
473 * Delay any upcoming NACK by this node if another node has already
474 * requested the first message this node is going to ask for.
476 static void bclink_peek_nack(struct net *net, struct tipc_msg *msg)
478 struct tipc_node *n_ptr = tipc_node_find(net, msg_destnode(msg));
480 if (unlikely(!n_ptr))
483 tipc_node_lock(n_ptr);
484 if (n_ptr->bclink.recv_permitted &&
485 (n_ptr->bclink.last_in != n_ptr->bclink.last_sent) &&
486 (n_ptr->bclink.last_in == msg_bcgap_after(msg)))
487 n_ptr->bclink.oos_state = 2;
488 tipc_node_unlock(n_ptr);
489 tipc_node_put(n_ptr);
492 /* tipc_bcbase_xmit - broadcast a packet queue across one or more bearers
494 * Note that number of reachable destinations, as indicated in the dests[]
495 * array, may transitionally differ from the number of destinations indicated
496 * in each sent buffer. We can sustain this. Excess destination nodes will
497 * drop and never acknowledge the unexpected packets, and missing destinations
498 * will either require retransmission (if they are just about to be added to
499 * the bearer), or be removed from the buffer's 'ackers' counter (if they
502 static void tipc_bcbase_xmit(struct net *net, struct sk_buff_head *xmitq)
505 struct tipc_bc_base *bb = tipc_bc_base(net);
506 struct sk_buff *skb, *_skb;
507 struct sk_buff_head _xmitq;
509 if (skb_queue_empty(xmitq))
512 /* The typical case: at least one bearer has links to all nodes */
513 bearer_id = bb->primary_bearer;
514 if (bearer_id >= 0) {
515 tipc_bearer_bc_xmit(net, bearer_id, xmitq);
519 /* We have to transmit across all bearers */
520 skb_queue_head_init(&_xmitq);
521 for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
522 if (!bb->dests[bearer_id])
525 skb_queue_walk(xmitq, skb) {
526 _skb = pskb_copy_for_clone(skb, GFP_ATOMIC);
529 __skb_queue_tail(&_xmitq, _skb);
531 tipc_bearer_bc_xmit(net, bearer_id, &_xmitq);
533 __skb_queue_purge(xmitq);
534 __skb_queue_purge(&_xmitq);
537 /* tipc_bcast_xmit - deliver buffer chain to all nodes in cluster
538 * and to identified node local sockets
539 * @net: the applicable net namespace
540 * @list: chain of buffers containing message
541 * Consumes the buffer chain, except when returning -ELINKCONG
542 * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
544 int tipc_bcast_xmit(struct net *net, struct sk_buff_head *list)
546 struct tipc_link *l = tipc_bc_sndlink(net);
547 struct sk_buff_head xmitq, inputq, rcvq;
550 __skb_queue_head_init(&rcvq);
551 __skb_queue_head_init(&xmitq);
552 skb_queue_head_init(&inputq);
554 /* Prepare message clone for local node */
555 if (unlikely(!tipc_msg_reassemble(list, &rcvq)))
556 return -EHOSTUNREACH;
558 tipc_bcast_lock(net);
559 if (tipc_link_bc_peers(l))
560 rc = tipc_link_xmit(l, list, &xmitq);
561 tipc_bcast_unlock(net);
563 /* Don't send to local node if adding to link failed */
565 __skb_queue_purge(&rcvq);
569 /* Broadcast to all nodes, inluding local node */
570 tipc_bcbase_xmit(net, &xmitq);
571 tipc_sk_mcast_rcv(net, &rcvq, &inputq);
572 __skb_queue_purge(list);
576 /* tipc_bcast_rcv - receive a broadcast packet, and deliver to rcv link
578 * RCU is locked, no other locks set
580 int tipc_bcast_rcv(struct net *net, struct tipc_link *l, struct sk_buff *skb)
582 struct tipc_msg *hdr = buf_msg(skb);
583 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
584 struct sk_buff_head xmitq;
587 __skb_queue_head_init(&xmitq);
589 if (msg_mc_netid(hdr) != tipc_netid(net) || !tipc_link_is_up(l)) {
594 tipc_bcast_lock(net);
595 if (msg_user(hdr) == BCAST_PROTOCOL)
596 rc = tipc_link_bc_nack_rcv(l, skb, &xmitq);
598 rc = tipc_link_rcv(l, skb, NULL);
599 tipc_bcast_unlock(net);
601 tipc_bcbase_xmit(net, &xmitq);
603 /* Any socket wakeup messages ? */
604 if (!skb_queue_empty(inputq))
605 tipc_sk_rcv(net, inputq);
610 /* tipc_bcast_ack_rcv - receive and handle a broadcast acknowledge
612 * RCU is locked, no other locks set
614 void tipc_bcast_ack_rcv(struct net *net, struct tipc_link *l, u32 acked)
616 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
617 struct sk_buff_head xmitq;
619 __skb_queue_head_init(&xmitq);
621 tipc_bcast_lock(net);
622 tipc_link_bc_ack_rcv(l, acked, &xmitq);
623 tipc_bcast_unlock(net);
625 tipc_bcbase_xmit(net, &xmitq);
627 /* Any socket wakeup messages ? */
628 if (!skb_queue_empty(inputq))
629 tipc_sk_rcv(net, inputq);
632 /* tipc_bcast_synch_rcv - check and update rcv link with peer's send state
634 * RCU is locked, no other locks set
636 void tipc_bcast_sync_rcv(struct net *net, struct tipc_link *l,
637 struct tipc_msg *hdr)
639 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
640 struct sk_buff_head xmitq;
642 __skb_queue_head_init(&xmitq);
644 tipc_bcast_lock(net);
645 if (msg_type(hdr) == STATE_MSG) {
646 tipc_link_bc_ack_rcv(l, msg_bcast_ack(hdr), &xmitq);
647 tipc_link_bc_sync_rcv(l, hdr, &xmitq);
649 tipc_link_bc_init_rcv(l, hdr);
651 tipc_bcast_unlock(net);
653 tipc_bcbase_xmit(net, &xmitq);
655 /* Any socket wakeup messages ? */
656 if (!skb_queue_empty(inputq))
657 tipc_sk_rcv(net, inputq);
660 /* tipc_bcast_add_peer - add a peer node to broadcast link and bearer
662 * RCU is locked, node lock is set
664 void tipc_bcast_add_peer(struct net *net, struct tipc_link *uc_l,
665 struct sk_buff_head *xmitq)
667 struct tipc_link *snd_l = tipc_bc_sndlink(net);
669 tipc_bcast_lock(net);
670 tipc_link_add_bc_peer(snd_l, uc_l, xmitq);
671 tipc_bcbase_select_primary(net);
672 tipc_bcast_unlock(net);
675 /* tipc_bcast_remove_peer - remove a peer node from broadcast link and bearer
677 * RCU is locked, node lock is set
679 void tipc_bcast_remove_peer(struct net *net, struct tipc_link *rcv_l)
681 struct tipc_link *snd_l = tipc_bc_sndlink(net);
682 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
683 struct sk_buff_head xmitq;
685 __skb_queue_head_init(&xmitq);
687 tipc_bcast_lock(net);
688 tipc_link_remove_bc_peer(snd_l, rcv_l, &xmitq);
689 tipc_bcbase_select_primary(net);
690 tipc_bcast_unlock(net);
692 tipc_bcbase_xmit(net, &xmitq);
694 /* Any socket wakeup messages ? */
695 if (!skb_queue_empty(inputq))
696 tipc_sk_rcv(net, inputq);
700 * bclink_accept_pkt - accept an incoming, in-sequence broadcast packet
702 * Called with both sending node's lock and bclink_lock taken.
704 static void bclink_accept_pkt(struct tipc_node *node, u32 seqno)
706 struct tipc_net *tn = net_generic(node->net, tipc_net_id);
708 bclink_update_last_sent(node, seqno);
709 node->bclink.last_in = seqno;
710 node->bclink.oos_state = 0;
711 tn->bcl->stats.recv_info++;
714 * Unicast an ACK periodically, ensuring that
715 * all nodes in the cluster don't ACK at the same time
717 if (((seqno - tn->own_addr) % TIPC_MIN_LINK_WIN) == 0) {
718 tipc_link_proto_xmit(node_active_link(node, node->addr),
719 STATE_MSG, 0, 0, 0, 0);
720 tn->bcl->stats.sent_acks++;
725 * tipc_bclink_rcv - receive a broadcast packet, and deliver upwards
727 * RCU is locked, no other locks set
729 void tipc_bclink_rcv(struct net *net, struct sk_buff *buf)
731 struct tipc_net *tn = net_generic(net, tipc_net_id);
732 struct tipc_link *bcl = tn->bcl;
733 struct tipc_msg *msg = buf_msg(buf);
734 struct tipc_node *node;
739 struct sk_buff *iskb;
740 struct sk_buff_head *arrvq, *inputq;
742 /* Screen out unwanted broadcast messages */
743 if (msg_mc_netid(msg) != tn->net_id)
746 node = tipc_node_find(net, msg_prevnode(msg));
749 tipc_node_lock(node);
750 if (unlikely(!node->bclink.recv_permitted))
753 /* Handle broadcast protocol message */
754 if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
755 if (msg_type(msg) != STATE_MSG)
757 if (msg_destnode(msg) == tn->own_addr) {
758 tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
759 tipc_bclink_lock(net);
760 bcl->stats.recv_nacks++;
761 tn->bcbase->retransmit_to = node;
762 bclink_retransmit_pkt(tn, msg_bcgap_after(msg),
764 tipc_bclink_unlock(net);
765 tipc_node_unlock(node);
767 tipc_node_unlock(node);
768 bclink_peek_nack(net, msg);
773 /* Handle in-sequence broadcast message */
774 seqno = msg_seqno(msg);
775 next_in = mod(node->bclink.last_in + 1);
776 arrvq = &tn->bcbase->arrvq;
777 inputq = &tn->bcbase->inputq;
779 if (likely(seqno == next_in)) {
781 /* Deliver message to destination */
782 if (likely(msg_isdata(msg))) {
783 tipc_bclink_lock(net);
784 bclink_accept_pkt(node, seqno);
785 spin_lock_bh(&inputq->lock);
786 __skb_queue_tail(arrvq, buf);
787 spin_unlock_bh(&inputq->lock);
788 node->action_flags |= TIPC_BCAST_MSG_EVT;
789 tipc_bclink_unlock(net);
790 tipc_node_unlock(node);
791 } else if (msg_user(msg) == MSG_BUNDLER) {
792 tipc_bclink_lock(net);
793 bclink_accept_pkt(node, seqno);
794 bcl->stats.recv_bundles++;
795 bcl->stats.recv_bundled += msg_msgcnt(msg);
797 while (tipc_msg_extract(buf, &iskb, &pos)) {
798 spin_lock_bh(&inputq->lock);
799 __skb_queue_tail(arrvq, iskb);
800 spin_unlock_bh(&inputq->lock);
802 node->action_flags |= TIPC_BCAST_MSG_EVT;
803 tipc_bclink_unlock(net);
804 tipc_node_unlock(node);
805 } else if (msg_user(msg) == MSG_FRAGMENTER) {
806 tipc_bclink_lock(net);
807 bclink_accept_pkt(node, seqno);
808 tipc_buf_append(&node->bclink.reasm_buf, &buf);
809 if (unlikely(!buf && !node->bclink.reasm_buf)) {
810 tipc_bclink_unlock(net);
813 bcl->stats.recv_fragments++;
815 bcl->stats.recv_fragmented++;
817 tipc_bclink_unlock(net);
820 tipc_bclink_unlock(net);
821 tipc_node_unlock(node);
823 tipc_bclink_lock(net);
824 bclink_accept_pkt(node, seqno);
825 tipc_bclink_unlock(net);
826 tipc_node_unlock(node);
831 /* Determine new synchronization state */
832 tipc_node_lock(node);
833 if (unlikely(!tipc_node_is_up(node)))
836 if (node->bclink.last_in == node->bclink.last_sent)
839 if (skb_queue_empty(&node->bclink.deferdq)) {
840 node->bclink.oos_state = 1;
844 msg = buf_msg(skb_peek(&node->bclink.deferdq));
845 seqno = msg_seqno(msg);
846 next_in = mod(next_in + 1);
847 if (seqno != next_in)
850 /* Take in-sequence message from deferred queue & deliver it */
851 buf = __skb_dequeue(&node->bclink.deferdq);
855 /* Handle out-of-sequence broadcast message */
856 if (less(next_in, seqno)) {
857 deferred = tipc_link_defer_pkt(&node->bclink.deferdq,
859 bclink_update_last_sent(node, seqno);
863 tipc_bclink_lock(net);
866 bcl->stats.deferred_recv++;
868 bcl->stats.duplicates++;
870 tipc_bclink_unlock(net);
873 tipc_node_unlock(node);
879 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
881 return (n_ptr->bclink.recv_permitted &&
882 (tipc_bclink_get_last_sent(n_ptr->net) != n_ptr->bclink.acked));
887 * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
889 * Send packet over as many bearers as necessary to reach all nodes
890 * that have joined the broadcast link.
892 * Returns 0 (packet sent successfully) under all circumstances,
893 * since the broadcast link's pseudo-bearer never blocks
895 static int tipc_bcbearer_send(struct net *net, struct sk_buff *buf,
896 struct tipc_bearer *unused1,
897 struct tipc_media_addr *unused2)
900 struct tipc_msg *msg = buf_msg(buf);
901 struct tipc_net *tn = net_generic(net, tipc_net_id);
902 struct tipc_bcbearer *bcbearer = tn->bcbearer;
903 struct tipc_bc_base *bclink = tn->bcbase;
905 /* Prepare broadcast link message for reliable transmission,
906 * if first time trying to send it;
907 * preparation is skipped for broadcast link protocol messages
908 * since they are sent in an unreliable manner and don't need it
910 if (likely(!msg_non_seq(buf_msg(buf)))) {
911 bcbuf_set_acks(buf, bclink->bcast_nodes.count);
912 msg_set_non_seq(msg, 1);
913 msg_set_mc_netid(msg, tn->net_id);
914 tn->bcl->stats.sent_info++;
915 if (WARN_ON(!bclink->bcast_nodes.count)) {
920 msg_set_mc_netid(msg, tn->net_id);
922 /* Send buffer over bearers until all targets reached */
923 bcbearer->remains = bclink->bcast_nodes;
925 for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
926 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
927 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
928 struct tipc_bearer *bp[2] = {p, s};
929 struct tipc_bearer *b = bp[msg_link_selector(msg)];
930 struct sk_buff *tbuf;
933 break; /* No more bearers to try */
936 tipc_nmap_diff(&bcbearer->remains, &b->nodes,
937 &bcbearer->remains_new);
938 if (bcbearer->remains_new.count == bcbearer->remains.count)
939 continue; /* Nothing added by bearer pair */
942 /* Use original buffer for first bearer */
943 tipc_bearer_send(net, b->identity, buf, &b->bcast_addr);
945 /* Avoid concurrent buffer access */
946 tbuf = pskb_copy_for_clone(buf, GFP_ATOMIC);
949 tipc_bearer_send(net, b->identity, tbuf,
951 kfree_skb(tbuf); /* Bearer keeps a clone */
953 if (bcbearer->remains_new.count == 0)
954 break; /* All targets reached */
956 bcbearer->remains = bcbearer->remains_new;
963 * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
965 void tipc_bcbearer_sort(struct net *net, struct tipc_node_map *nm_ptr,
966 u32 node, bool action)
968 struct tipc_net *tn = net_generic(net, tipc_net_id);
969 struct tipc_bcbearer *bcbearer = tn->bcbearer;
970 struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
971 struct tipc_bcbearer_pair *bp_curr;
972 struct tipc_bearer *b;
976 tipc_bclink_lock(net);
979 tipc_nmap_add(nm_ptr, node);
981 tipc_nmap_remove(nm_ptr, node);
983 /* Group bearers by priority (can assume max of two per priority) */
984 memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
987 for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
988 b = rcu_dereference_rtnl(tn->bearer_list[b_index]);
989 if (!b || !b->nodes.count)
992 if (!bp_temp[b->priority].primary)
993 bp_temp[b->priority].primary = b;
995 bp_temp[b->priority].secondary = b;
999 /* Create array of bearer pairs for broadcasting */
1000 bp_curr = bcbearer->bpairs;
1001 memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
1003 for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
1005 if (!bp_temp[pri].primary)
1008 bp_curr->primary = bp_temp[pri].primary;
1010 if (bp_temp[pri].secondary) {
1011 if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
1012 &bp_temp[pri].secondary->nodes)) {
1013 bp_curr->secondary = bp_temp[pri].secondary;
1016 bp_curr->primary = bp_temp[pri].secondary;
1023 tipc_bclink_unlock(net);
1026 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
1027 struct tipc_stats *stats)
1030 struct nlattr *nest;
1037 struct nla_map map[] = {
1038 {TIPC_NLA_STATS_RX_INFO, stats->recv_info},
1039 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
1040 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
1041 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
1042 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
1043 {TIPC_NLA_STATS_TX_INFO, stats->sent_info},
1044 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
1045 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
1046 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
1047 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
1048 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
1049 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
1050 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
1051 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
1052 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
1053 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
1054 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
1055 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
1056 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
1057 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
1060 nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1064 for (i = 0; i < ARRAY_SIZE(map); i++)
1065 if (nla_put_u32(skb, map[i].key, map[i].val))
1068 nla_nest_end(skb, nest);
1072 nla_nest_cancel(skb, nest);
1077 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
1081 struct nlattr *attrs;
1082 struct nlattr *prop;
1083 struct tipc_net *tn = net_generic(net, tipc_net_id);
1084 struct tipc_link *bcl = tn->bcl;
1089 tipc_bclink_lock(net);
1091 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1092 NLM_F_MULTI, TIPC_NL_LINK_GET);
1096 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1100 /* The broadcast link is always up */
1101 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1104 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
1106 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
1108 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, bcl->rcv_nxt))
1110 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, bcl->snd_nxt))
1113 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1116 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
1118 nla_nest_end(msg->skb, prop);
1120 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
1124 tipc_bclink_unlock(net);
1125 nla_nest_end(msg->skb, attrs);
1126 genlmsg_end(msg->skb, hdr);
1131 nla_nest_cancel(msg->skb, prop);
1133 nla_nest_cancel(msg->skb, attrs);
1135 tipc_bclink_unlock(net);
1136 genlmsg_cancel(msg->skb, hdr);
1141 int tipc_bclink_reset_stats(struct net *net)
1143 struct tipc_net *tn = net_generic(net, tipc_net_id);
1144 struct tipc_link *bcl = tn->bcl;
1147 return -ENOPROTOOPT;
1149 tipc_bclink_lock(net);
1150 memset(&bcl->stats, 0, sizeof(bcl->stats));
1151 tipc_bclink_unlock(net);
1155 int tipc_bclink_set_queue_limits(struct net *net, u32 limit)
1157 struct tipc_net *tn = net_generic(net, tipc_net_id);
1158 struct tipc_link *bcl = tn->bcl;
1161 return -ENOPROTOOPT;
1162 if (limit < BCLINK_WIN_MIN)
1163 limit = BCLINK_WIN_MIN;
1164 if (limit > TIPC_MAX_LINK_WIN)
1166 tipc_bclink_lock(net);
1167 tipc_link_set_queue_limits(bcl, limit);
1168 tipc_bclink_unlock(net);
1172 int tipc_nl_bc_link_set(struct net *net, struct nlattr *attrs[])
1176 struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
1178 if (!attrs[TIPC_NLA_LINK_PROP])
1181 err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP], props);
1185 if (!props[TIPC_NLA_PROP_WIN])
1188 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1190 return tipc_bclink_set_queue_limits(net, win);
1193 int tipc_bcast_init(struct net *net)
1195 struct tipc_net *tn = tipc_net(net);
1196 struct tipc_bcbearer *bcb = NULL;
1197 struct tipc_bc_base *bb = NULL;
1198 struct tipc_link *l = NULL;
1200 bcb = kzalloc(sizeof(*bcb), GFP_ATOMIC);
1205 bcb->bearer.window = BCLINK_WIN_DEFAULT;
1206 bcb->bearer.mtu = MAX_PKT_DEFAULT_MCAST;
1207 bcb->bearer.identity = MAX_BEARERS;
1209 bcb->bearer.media = &bcb->media;
1210 bcb->media.send_msg = tipc_bcbearer_send;
1211 sprintf(bcb->media.name, "tipc-broadcast");
1212 strcpy(bcb->bearer.name, bcb->media.name);
1214 bb = kzalloc(sizeof(*bb), GFP_ATOMIC);
1218 __skb_queue_head_init(&bb->arrvq);
1219 spin_lock_init(&tipc_net(net)->bclock);
1222 if (!tipc_link_bc_create(&bb->node, 0, 0,
1223 MAX_PKT_DEFAULT_MCAST,
1233 rcu_assign_pointer(tn->bearer_list[MAX_BEARERS], &bcb->bearer);
1242 void tipc_bcast_reinit(struct net *net)
1244 struct tipc_bc_base *b = tipc_bc_base(net);
1246 msg_set_prevnode(b->link->pmsg, tipc_own_addr(net));
1249 void tipc_bcast_stop(struct net *net)
1251 struct tipc_net *tn = net_generic(net, tipc_net_id);
1253 tipc_bclink_lock(net);
1254 tipc_link_purge_queues(tn->bcl);
1255 tipc_bclink_unlock(net);
1256 RCU_INIT_POINTER(tn->bearer_list[BCBEARER], NULL);
1258 kfree(tn->bcbearer);
1264 * tipc_nmap_add - add a node to a node map
1266 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
1268 int n = tipc_node(node);
1270 u32 mask = (1 << (n % WSIZE));
1272 if ((nm_ptr->map[w] & mask) == 0) {
1274 nm_ptr->map[w] |= mask;
1279 * tipc_nmap_remove - remove a node from a node map
1281 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
1283 int n = tipc_node(node);
1285 u32 mask = (1 << (n % WSIZE));
1287 if ((nm_ptr->map[w] & mask) != 0) {
1288 nm_ptr->map[w] &= ~mask;
1294 * tipc_nmap_diff - find differences between node maps
1295 * @nm_a: input node map A
1296 * @nm_b: input node map B
1297 * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
1299 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
1300 struct tipc_node_map *nm_b,
1301 struct tipc_node_map *nm_diff)
1303 int stop = ARRAY_SIZE(nm_a->map);
1308 memset(nm_diff, 0, sizeof(*nm_diff));
1309 for (w = 0; w < stop; w++) {
1310 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
1311 nm_diff->map[w] = map;
1313 for (b = 0 ; b < WSIZE; b++) {