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 int tipc_bcast_get_mtu(struct net *net)
153 return tipc_link_mtu(tipc_bc_sndlink(net));
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++) {
189 mtu = tipc_bearer_mtu(net, i);
190 if (mtu < tipc_link_mtu(bb->link))
191 tipc_link_set_mtu(bb->link, mtu);
193 if (bb->dests[i] < all_dests)
196 bb->primary_bearer = i;
198 /* Reduce risk that all nodes select same primary */
199 if ((i ^ tipc_own_addr(net)) & 1)
204 void tipc_bcast_inc_bearer_dst_cnt(struct net *net, int bearer_id)
206 struct tipc_bc_base *bb = tipc_bc_base(net);
208 tipc_bcast_lock(net);
209 bb->dests[bearer_id]++;
210 tipc_bcbase_select_primary(net);
211 tipc_bcast_unlock(net);
214 void tipc_bcast_dec_bearer_dst_cnt(struct net *net, int bearer_id)
216 struct tipc_bc_base *bb = tipc_bc_base(net);
218 tipc_bcast_lock(net);
219 bb->dests[bearer_id]--;
220 tipc_bcbase_select_primary(net);
221 tipc_bcast_unlock(net);
224 static void bclink_set_last_sent(struct net *net)
226 struct tipc_net *tn = net_generic(net, tipc_net_id);
227 struct tipc_link *bcl = tn->bcl;
229 bcl->silent_intv_cnt = mod(bcl->snd_nxt - 1);
232 u32 tipc_bclink_get_last_sent(struct net *net)
234 struct tipc_net *tn = net_generic(net, tipc_net_id);
236 return tn->bcl->silent_intv_cnt;
239 static void bclink_update_last_sent(struct tipc_node *node, u32 seqno)
241 node->bclink.last_sent = less_eq(node->bclink.last_sent, seqno) ?
242 seqno : node->bclink.last_sent;
246 * tipc_bclink_retransmit_to - get most recent node to request retransmission
248 * Called with bclink_lock locked
250 struct tipc_node *tipc_bclink_retransmit_to(struct net *net)
252 struct tipc_net *tn = net_generic(net, tipc_net_id);
254 return tn->bcbase->retransmit_to;
258 * bclink_retransmit_pkt - retransmit broadcast packets
259 * @after: sequence number of last packet to *not* retransmit
260 * @to: sequence number of last packet to retransmit
262 * Called with bclink_lock locked
264 static void bclink_retransmit_pkt(struct tipc_net *tn, u32 after, u32 to)
267 struct tipc_link *bcl = tn->bcl;
269 skb_queue_walk(&bcl->transmq, skb) {
270 if (more(buf_seqno(skb), after)) {
271 tipc_link_retransmit(bcl, skb, mod(to - after));
278 * bclink_prepare_wakeup - prepare users for wakeup after congestion
279 * @bcl: broadcast link
280 * @resultq: queue for users which can be woken up
281 * Move a number of waiting users, as permitted by available space in
282 * the send queue, from link wait queue to specified queue for wakeup
284 static void bclink_prepare_wakeup(struct tipc_link *bcl, struct sk_buff_head *resultq)
286 int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
288 struct sk_buff *skb, *tmp;
290 skb_queue_walk_safe(&bcl->wakeupq, skb, tmp) {
291 imp = TIPC_SKB_CB(skb)->chain_imp;
292 lim = bcl->window + bcl->backlog[imp].limit;
293 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
294 if ((pnd[imp] + bcl->backlog[imp].len) >= lim)
296 skb_unlink(skb, &bcl->wakeupq);
297 skb_queue_tail(resultq, skb);
302 * tipc_bclink_wakeup_users - wake up pending users
304 * Called with no locks taken
306 void tipc_bclink_wakeup_users(struct net *net)
308 struct tipc_net *tn = net_generic(net, tipc_net_id);
309 struct tipc_link *bcl = tn->bcl;
310 struct sk_buff_head resultq;
312 skb_queue_head_init(&resultq);
313 bclink_prepare_wakeup(bcl, &resultq);
314 tipc_sk_rcv(net, &resultq);
318 * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
319 * @n_ptr: node that sent acknowledgement info
320 * @acked: broadcast sequence # that has been acknowledged
322 * Node is locked, bclink_lock unlocked.
324 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
326 struct sk_buff *skb, *tmp;
327 unsigned int released = 0;
328 struct net *net = n_ptr->net;
329 struct tipc_net *tn = net_generic(net, tipc_net_id);
331 if (unlikely(!n_ptr->bclink.recv_permitted))
333 tipc_bclink_lock(net);
335 /* Bail out if tx queue is empty (no clean up is required) */
336 skb = skb_peek(&tn->bcl->transmq);
340 /* Determine which messages need to be acknowledged */
341 if (acked == INVALID_LINK_SEQ) {
343 * Contact with specified node has been lost, so need to
344 * acknowledge sent messages only (if other nodes still exist)
345 * or both sent and unsent messages (otherwise)
347 if (tn->bcbase->bcast_nodes.count)
348 acked = tn->bcl->silent_intv_cnt;
350 acked = tn->bcl->snd_nxt;
353 * Bail out if specified sequence number does not correspond
354 * to a message that has been sent and not yet acknowledged
356 if (less(acked, buf_seqno(skb)) ||
357 less(tn->bcl->silent_intv_cnt, acked) ||
358 less_eq(acked, n_ptr->bclink.acked))
361 /* Skip over packets that node has previously acknowledged */
362 skb_queue_walk(&tn->bcl->transmq, skb) {
363 if (more(buf_seqno(skb), n_ptr->bclink.acked))
366 /* Update packets that node is now acknowledging */
367 skb_queue_walk_from_safe(&tn->bcl->transmq, skb, tmp) {
368 if (more(buf_seqno(skb), acked))
370 bcbuf_decr_acks(skb);
371 bclink_set_last_sent(net);
372 if (bcbuf_acks(skb) == 0) {
373 __skb_unlink(skb, &tn->bcl->transmq);
378 n_ptr->bclink.acked = acked;
380 /* Try resolving broadcast link congestion, if necessary */
381 if (unlikely(skb_peek(&tn->bcl->backlogq))) {
382 tipc_link_push_packets(tn->bcl);
383 bclink_set_last_sent(net);
385 if (unlikely(released && !skb_queue_empty(&tn->bcl->wakeupq)))
386 n_ptr->action_flags |= TIPC_WAKEUP_BCAST_USERS;
388 tipc_bclink_unlock(net);
392 * tipc_bclink_update_link_state - update broadcast link state
394 * RCU and node lock set
396 void tipc_bclink_update_link_state(struct tipc_node *n_ptr,
400 struct net *net = n_ptr->net;
401 struct tipc_net *tn = net_generic(net, tipc_net_id);
402 struct tipc_link *bcl = tn->bcl;
404 /* Ignore "stale" link state info */
405 if (less_eq(last_sent, n_ptr->bclink.last_in))
408 /* Update link synchronization state; quit if in sync */
409 bclink_update_last_sent(n_ptr, last_sent);
411 /* This is a good location for statistical profiling */
412 bcl->stats.queue_sz_counts++;
413 bcl->stats.accu_queue_sz += skb_queue_len(&bcl->transmq);
415 if (n_ptr->bclink.last_sent == n_ptr->bclink.last_in)
418 /* Update out-of-sync state; quit if loss is still unconfirmed */
419 if ((++n_ptr->bclink.oos_state) == 1) {
420 if (n_ptr->bclink.deferred_size < (TIPC_MIN_LINK_WIN / 2))
422 n_ptr->bclink.oos_state++;
425 /* Don't NACK if one has been recently sent (or seen) */
426 if (n_ptr->bclink.oos_state & 0x1)
430 buf = tipc_buf_acquire(INT_H_SIZE);
432 struct tipc_msg *msg = buf_msg(buf);
433 struct sk_buff *skb = skb_peek(&n_ptr->bclink.deferdq);
434 u32 to = skb ? buf_seqno(skb) - 1 : n_ptr->bclink.last_sent;
436 tipc_msg_init(tn->own_addr, msg, BCAST_PROTOCOL, STATE_MSG,
437 INT_H_SIZE, n_ptr->addr);
438 msg_set_non_seq(msg, 1);
439 msg_set_mc_netid(msg, tn->net_id);
440 msg_set_bcast_ack(msg, n_ptr->bclink.last_in);
441 msg_set_bcgap_after(msg, n_ptr->bclink.last_in);
442 msg_set_bcgap_to(msg, to);
444 tipc_bclink_lock(net);
445 tipc_bearer_send(net, MAX_BEARERS, buf, NULL);
446 tn->bcl->stats.sent_nacks++;
447 tipc_bclink_unlock(net);
450 n_ptr->bclink.oos_state++;
454 void tipc_bclink_sync_state(struct tipc_node *n, struct tipc_msg *hdr)
456 u16 last = msg_last_bcast(hdr);
457 int mtyp = msg_type(hdr);
459 if (unlikely(msg_user(hdr) != LINK_PROTOCOL))
461 if (mtyp == STATE_MSG) {
462 tipc_bclink_update_link_state(n, last);
465 /* Compatibility: older nodes don't know BCAST_PROTOCOL synchronization,
466 * and transfer synch info in LINK_PROTOCOL messages.
468 if (tipc_node_is_up(n))
470 if ((mtyp != RESET_MSG) && (mtyp != ACTIVATE_MSG))
472 n->bclink.last_sent = last;
473 n->bclink.last_in = last;
474 n->bclink.oos_state = 0;
478 * bclink_peek_nack - monitor retransmission requests sent by other nodes
480 * Delay any upcoming NACK by this node if another node has already
481 * requested the first message this node is going to ask for.
483 static void bclink_peek_nack(struct net *net, struct tipc_msg *msg)
485 struct tipc_node *n_ptr = tipc_node_find(net, msg_destnode(msg));
487 if (unlikely(!n_ptr))
490 tipc_node_lock(n_ptr);
491 if (n_ptr->bclink.recv_permitted &&
492 (n_ptr->bclink.last_in != n_ptr->bclink.last_sent) &&
493 (n_ptr->bclink.last_in == msg_bcgap_after(msg)))
494 n_ptr->bclink.oos_state = 2;
495 tipc_node_unlock(n_ptr);
496 tipc_node_put(n_ptr);
499 /* tipc_bcbase_xmit - broadcast a packet queue across one or more bearers
501 * Note that number of reachable destinations, as indicated in the dests[]
502 * array, may transitionally differ from the number of destinations indicated
503 * in each sent buffer. We can sustain this. Excess destination nodes will
504 * drop and never acknowledge the unexpected packets, and missing destinations
505 * will either require retransmission (if they are just about to be added to
506 * the bearer), or be removed from the buffer's 'ackers' counter (if they
509 static void tipc_bcbase_xmit(struct net *net, struct sk_buff_head *xmitq)
512 struct tipc_bc_base *bb = tipc_bc_base(net);
513 struct sk_buff *skb, *_skb;
514 struct sk_buff_head _xmitq;
516 if (skb_queue_empty(xmitq))
519 /* The typical case: at least one bearer has links to all nodes */
520 bearer_id = bb->primary_bearer;
521 if (bearer_id >= 0) {
522 tipc_bearer_bc_xmit(net, bearer_id, xmitq);
526 /* We have to transmit across all bearers */
527 skb_queue_head_init(&_xmitq);
528 for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
529 if (!bb->dests[bearer_id])
532 skb_queue_walk(xmitq, skb) {
533 _skb = pskb_copy_for_clone(skb, GFP_ATOMIC);
536 __skb_queue_tail(&_xmitq, _skb);
538 tipc_bearer_bc_xmit(net, bearer_id, &_xmitq);
540 __skb_queue_purge(xmitq);
541 __skb_queue_purge(&_xmitq);
544 /* tipc_bcast_xmit - deliver buffer chain to all nodes in cluster
545 * and to identified node local sockets
546 * @net: the applicable net namespace
547 * @list: chain of buffers containing message
548 * Consumes the buffer chain, except when returning -ELINKCONG
549 * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
551 int tipc_bcast_xmit(struct net *net, struct sk_buff_head *list)
553 struct tipc_link *l = tipc_bc_sndlink(net);
554 struct sk_buff_head xmitq, inputq, rcvq;
557 __skb_queue_head_init(&rcvq);
558 __skb_queue_head_init(&xmitq);
559 skb_queue_head_init(&inputq);
561 /* Prepare message clone for local node */
562 if (unlikely(!tipc_msg_reassemble(list, &rcvq)))
563 return -EHOSTUNREACH;
565 tipc_bcast_lock(net);
566 if (tipc_link_bc_peers(l))
567 rc = tipc_link_xmit(l, list, &xmitq);
568 tipc_bcast_unlock(net);
570 /* Don't send to local node if adding to link failed */
572 __skb_queue_purge(&rcvq);
576 /* Broadcast to all nodes, inluding local node */
577 tipc_bcbase_xmit(net, &xmitq);
578 tipc_sk_mcast_rcv(net, &rcvq, &inputq);
579 __skb_queue_purge(list);
583 /* tipc_bcast_rcv - receive a broadcast packet, and deliver to rcv link
585 * RCU is locked, no other locks set
587 int tipc_bcast_rcv(struct net *net, struct tipc_link *l, struct sk_buff *skb)
589 struct tipc_msg *hdr = buf_msg(skb);
590 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
591 struct sk_buff_head xmitq;
594 __skb_queue_head_init(&xmitq);
596 if (msg_mc_netid(hdr) != tipc_netid(net) || !tipc_link_is_up(l)) {
601 tipc_bcast_lock(net);
602 if (msg_user(hdr) == BCAST_PROTOCOL)
603 rc = tipc_link_bc_nack_rcv(l, skb, &xmitq);
605 rc = tipc_link_rcv(l, skb, NULL);
606 tipc_bcast_unlock(net);
608 tipc_bcbase_xmit(net, &xmitq);
610 /* Any socket wakeup messages ? */
611 if (!skb_queue_empty(inputq))
612 tipc_sk_rcv(net, inputq);
617 /* tipc_bcast_ack_rcv - receive and handle a broadcast acknowledge
619 * RCU is locked, no other locks set
621 void tipc_bcast_ack_rcv(struct net *net, struct tipc_link *l, u32 acked)
623 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
624 struct sk_buff_head xmitq;
626 __skb_queue_head_init(&xmitq);
628 tipc_bcast_lock(net);
629 tipc_link_bc_ack_rcv(l, acked, &xmitq);
630 tipc_bcast_unlock(net);
632 tipc_bcbase_xmit(net, &xmitq);
634 /* Any socket wakeup messages ? */
635 if (!skb_queue_empty(inputq))
636 tipc_sk_rcv(net, inputq);
639 /* tipc_bcast_synch_rcv - check and update rcv link with peer's send state
641 * RCU is locked, no other locks set
643 void tipc_bcast_sync_rcv(struct net *net, struct tipc_link *l,
644 struct tipc_msg *hdr)
646 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
647 struct sk_buff_head xmitq;
649 __skb_queue_head_init(&xmitq);
651 tipc_bcast_lock(net);
652 if (msg_type(hdr) == STATE_MSG) {
653 tipc_link_bc_ack_rcv(l, msg_bcast_ack(hdr), &xmitq);
654 tipc_link_bc_sync_rcv(l, hdr, &xmitq);
656 tipc_link_bc_init_rcv(l, hdr);
658 tipc_bcast_unlock(net);
660 tipc_bcbase_xmit(net, &xmitq);
662 /* Any socket wakeup messages ? */
663 if (!skb_queue_empty(inputq))
664 tipc_sk_rcv(net, inputq);
667 /* tipc_bcast_add_peer - add a peer node to broadcast link and bearer
669 * RCU is locked, node lock is set
671 void tipc_bcast_add_peer(struct net *net, struct tipc_link *uc_l,
672 struct sk_buff_head *xmitq)
674 struct tipc_link *snd_l = tipc_bc_sndlink(net);
676 tipc_bcast_lock(net);
677 tipc_link_add_bc_peer(snd_l, uc_l, xmitq);
678 tipc_bcbase_select_primary(net);
679 tipc_bcast_unlock(net);
682 /* tipc_bcast_remove_peer - remove a peer node from broadcast link and bearer
684 * RCU is locked, node lock is set
686 void tipc_bcast_remove_peer(struct net *net, struct tipc_link *rcv_l)
688 struct tipc_link *snd_l = tipc_bc_sndlink(net);
689 struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
690 struct sk_buff_head xmitq;
692 __skb_queue_head_init(&xmitq);
694 tipc_bcast_lock(net);
695 tipc_link_remove_bc_peer(snd_l, rcv_l, &xmitq);
696 tipc_bcbase_select_primary(net);
697 tipc_bcast_unlock(net);
699 tipc_bcbase_xmit(net, &xmitq);
701 /* Any socket wakeup messages ? */
702 if (!skb_queue_empty(inputq))
703 tipc_sk_rcv(net, inputq);
707 * bclink_accept_pkt - accept an incoming, in-sequence broadcast packet
709 * Called with both sending node's lock and bclink_lock taken.
711 static void bclink_accept_pkt(struct tipc_node *node, u32 seqno)
713 struct tipc_net *tn = net_generic(node->net, tipc_net_id);
715 bclink_update_last_sent(node, seqno);
716 node->bclink.last_in = seqno;
717 node->bclink.oos_state = 0;
718 tn->bcl->stats.recv_info++;
721 * Unicast an ACK periodically, ensuring that
722 * all nodes in the cluster don't ACK at the same time
724 if (((seqno - tn->own_addr) % TIPC_MIN_LINK_WIN) == 0) {
725 tipc_link_proto_xmit(node_active_link(node, node->addr),
726 STATE_MSG, 0, 0, 0, 0);
727 tn->bcl->stats.sent_acks++;
732 * tipc_bclink_rcv - receive a broadcast packet, and deliver upwards
734 * RCU is locked, no other locks set
736 void tipc_bclink_rcv(struct net *net, struct sk_buff *buf)
738 struct tipc_net *tn = net_generic(net, tipc_net_id);
739 struct tipc_link *bcl = tn->bcl;
740 struct tipc_msg *msg = buf_msg(buf);
741 struct tipc_node *node;
746 struct sk_buff *iskb;
747 struct sk_buff_head *arrvq, *inputq;
749 /* Screen out unwanted broadcast messages */
750 if (msg_mc_netid(msg) != tn->net_id)
753 node = tipc_node_find(net, msg_prevnode(msg));
756 tipc_node_lock(node);
757 if (unlikely(!node->bclink.recv_permitted))
760 /* Handle broadcast protocol message */
761 if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
762 if (msg_type(msg) != STATE_MSG)
764 if (msg_destnode(msg) == tn->own_addr) {
765 tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
766 tipc_bclink_lock(net);
767 bcl->stats.recv_nacks++;
768 tn->bcbase->retransmit_to = node;
769 bclink_retransmit_pkt(tn, msg_bcgap_after(msg),
771 tipc_bclink_unlock(net);
772 tipc_node_unlock(node);
774 tipc_node_unlock(node);
775 bclink_peek_nack(net, msg);
780 /* Handle in-sequence broadcast message */
781 seqno = msg_seqno(msg);
782 next_in = mod(node->bclink.last_in + 1);
783 arrvq = &tn->bcbase->arrvq;
784 inputq = &tn->bcbase->inputq;
786 if (likely(seqno == next_in)) {
788 /* Deliver message to destination */
789 if (likely(msg_isdata(msg))) {
790 tipc_bclink_lock(net);
791 bclink_accept_pkt(node, seqno);
792 spin_lock_bh(&inputq->lock);
793 __skb_queue_tail(arrvq, buf);
794 spin_unlock_bh(&inputq->lock);
795 node->action_flags |= TIPC_BCAST_MSG_EVT;
796 tipc_bclink_unlock(net);
797 tipc_node_unlock(node);
798 } else if (msg_user(msg) == MSG_BUNDLER) {
799 tipc_bclink_lock(net);
800 bclink_accept_pkt(node, seqno);
801 bcl->stats.recv_bundles++;
802 bcl->stats.recv_bundled += msg_msgcnt(msg);
804 while (tipc_msg_extract(buf, &iskb, &pos)) {
805 spin_lock_bh(&inputq->lock);
806 __skb_queue_tail(arrvq, iskb);
807 spin_unlock_bh(&inputq->lock);
809 node->action_flags |= TIPC_BCAST_MSG_EVT;
810 tipc_bclink_unlock(net);
811 tipc_node_unlock(node);
812 } else if (msg_user(msg) == MSG_FRAGMENTER) {
813 tipc_bclink_lock(net);
814 bclink_accept_pkt(node, seqno);
815 tipc_buf_append(&node->bclink.reasm_buf, &buf);
816 if (unlikely(!buf && !node->bclink.reasm_buf)) {
817 tipc_bclink_unlock(net);
820 bcl->stats.recv_fragments++;
822 bcl->stats.recv_fragmented++;
824 tipc_bclink_unlock(net);
827 tipc_bclink_unlock(net);
828 tipc_node_unlock(node);
830 tipc_bclink_lock(net);
831 bclink_accept_pkt(node, seqno);
832 tipc_bclink_unlock(net);
833 tipc_node_unlock(node);
838 /* Determine new synchronization state */
839 tipc_node_lock(node);
840 if (unlikely(!tipc_node_is_up(node)))
843 if (node->bclink.last_in == node->bclink.last_sent)
846 if (skb_queue_empty(&node->bclink.deferdq)) {
847 node->bclink.oos_state = 1;
851 msg = buf_msg(skb_peek(&node->bclink.deferdq));
852 seqno = msg_seqno(msg);
853 next_in = mod(next_in + 1);
854 if (seqno != next_in)
857 /* Take in-sequence message from deferred queue & deliver it */
858 buf = __skb_dequeue(&node->bclink.deferdq);
862 /* Handle out-of-sequence broadcast message */
863 if (less(next_in, seqno)) {
864 deferred = tipc_link_defer_pkt(&node->bclink.deferdq,
866 bclink_update_last_sent(node, seqno);
870 tipc_bclink_lock(net);
873 bcl->stats.deferred_recv++;
875 bcl->stats.duplicates++;
877 tipc_bclink_unlock(net);
880 tipc_node_unlock(node);
886 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
888 return (n_ptr->bclink.recv_permitted &&
889 (tipc_bclink_get_last_sent(n_ptr->net) != n_ptr->bclink.acked));
894 * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
896 * Send packet over as many bearers as necessary to reach all nodes
897 * that have joined the broadcast link.
899 * Returns 0 (packet sent successfully) under all circumstances,
900 * since the broadcast link's pseudo-bearer never blocks
902 static int tipc_bcbearer_send(struct net *net, struct sk_buff *buf,
903 struct tipc_bearer *unused1,
904 struct tipc_media_addr *unused2)
907 struct tipc_msg *msg = buf_msg(buf);
908 struct tipc_net *tn = net_generic(net, tipc_net_id);
909 struct tipc_bcbearer *bcbearer = tn->bcbearer;
910 struct tipc_bc_base *bclink = tn->bcbase;
912 /* Prepare broadcast link message for reliable transmission,
913 * if first time trying to send it;
914 * preparation is skipped for broadcast link protocol messages
915 * since they are sent in an unreliable manner and don't need it
917 if (likely(!msg_non_seq(buf_msg(buf)))) {
918 bcbuf_set_acks(buf, bclink->bcast_nodes.count);
919 msg_set_non_seq(msg, 1);
920 msg_set_mc_netid(msg, tn->net_id);
921 tn->bcl->stats.sent_info++;
922 if (WARN_ON(!bclink->bcast_nodes.count)) {
927 msg_set_mc_netid(msg, tn->net_id);
929 /* Send buffer over bearers until all targets reached */
930 bcbearer->remains = bclink->bcast_nodes;
932 for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
933 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
934 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
935 struct tipc_bearer *bp[2] = {p, s};
936 struct tipc_bearer *b = bp[msg_link_selector(msg)];
937 struct sk_buff *tbuf;
940 break; /* No more bearers to try */
943 tipc_nmap_diff(&bcbearer->remains, &b->nodes,
944 &bcbearer->remains_new);
945 if (bcbearer->remains_new.count == bcbearer->remains.count)
946 continue; /* Nothing added by bearer pair */
949 /* Use original buffer for first bearer */
950 tipc_bearer_send(net, b->identity, buf, &b->bcast_addr);
952 /* Avoid concurrent buffer access */
953 tbuf = pskb_copy_for_clone(buf, GFP_ATOMIC);
956 tipc_bearer_send(net, b->identity, tbuf,
958 kfree_skb(tbuf); /* Bearer keeps a clone */
960 if (bcbearer->remains_new.count == 0)
961 break; /* All targets reached */
963 bcbearer->remains = bcbearer->remains_new;
970 * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
972 void tipc_bcbearer_sort(struct net *net, struct tipc_node_map *nm_ptr,
973 u32 node, bool action)
975 struct tipc_net *tn = net_generic(net, tipc_net_id);
976 struct tipc_bcbearer *bcbearer = tn->bcbearer;
977 struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
978 struct tipc_bcbearer_pair *bp_curr;
979 struct tipc_bearer *b;
983 tipc_bclink_lock(net);
986 tipc_nmap_add(nm_ptr, node);
988 tipc_nmap_remove(nm_ptr, node);
990 /* Group bearers by priority (can assume max of two per priority) */
991 memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
994 for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
995 b = rcu_dereference_rtnl(tn->bearer_list[b_index]);
996 if (!b || !b->nodes.count)
999 if (!bp_temp[b->priority].primary)
1000 bp_temp[b->priority].primary = b;
1002 bp_temp[b->priority].secondary = b;
1006 /* Create array of bearer pairs for broadcasting */
1007 bp_curr = bcbearer->bpairs;
1008 memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
1010 for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
1012 if (!bp_temp[pri].primary)
1015 bp_curr->primary = bp_temp[pri].primary;
1017 if (bp_temp[pri].secondary) {
1018 if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
1019 &bp_temp[pri].secondary->nodes)) {
1020 bp_curr->secondary = bp_temp[pri].secondary;
1023 bp_curr->primary = bp_temp[pri].secondary;
1030 tipc_bclink_unlock(net);
1033 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
1034 struct tipc_stats *stats)
1037 struct nlattr *nest;
1044 struct nla_map map[] = {
1045 {TIPC_NLA_STATS_RX_INFO, stats->recv_info},
1046 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
1047 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
1048 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
1049 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
1050 {TIPC_NLA_STATS_TX_INFO, stats->sent_info},
1051 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
1052 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
1053 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
1054 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
1055 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
1056 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
1057 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
1058 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
1059 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
1060 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
1061 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
1062 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
1063 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
1064 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
1067 nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1071 for (i = 0; i < ARRAY_SIZE(map); i++)
1072 if (nla_put_u32(skb, map[i].key, map[i].val))
1075 nla_nest_end(skb, nest);
1079 nla_nest_cancel(skb, nest);
1084 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
1088 struct nlattr *attrs;
1089 struct nlattr *prop;
1090 struct tipc_net *tn = net_generic(net, tipc_net_id);
1091 struct tipc_link *bcl = tn->bcl;
1096 tipc_bclink_lock(net);
1098 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1099 NLM_F_MULTI, TIPC_NL_LINK_GET);
1103 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1107 /* The broadcast link is always up */
1108 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1111 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
1113 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
1115 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, bcl->rcv_nxt))
1117 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, bcl->snd_nxt))
1120 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1123 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
1125 nla_nest_end(msg->skb, prop);
1127 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
1131 tipc_bclink_unlock(net);
1132 nla_nest_end(msg->skb, attrs);
1133 genlmsg_end(msg->skb, hdr);
1138 nla_nest_cancel(msg->skb, prop);
1140 nla_nest_cancel(msg->skb, attrs);
1142 tipc_bclink_unlock(net);
1143 genlmsg_cancel(msg->skb, hdr);
1148 int tipc_bclink_reset_stats(struct net *net)
1150 struct tipc_net *tn = net_generic(net, tipc_net_id);
1151 struct tipc_link *bcl = tn->bcl;
1154 return -ENOPROTOOPT;
1156 tipc_bclink_lock(net);
1157 memset(&bcl->stats, 0, sizeof(bcl->stats));
1158 tipc_bclink_unlock(net);
1162 int tipc_bclink_set_queue_limits(struct net *net, u32 limit)
1164 struct tipc_net *tn = net_generic(net, tipc_net_id);
1165 struct tipc_link *bcl = tn->bcl;
1168 return -ENOPROTOOPT;
1169 if (limit < BCLINK_WIN_MIN)
1170 limit = BCLINK_WIN_MIN;
1171 if (limit > TIPC_MAX_LINK_WIN)
1173 tipc_bclink_lock(net);
1174 tipc_link_set_queue_limits(bcl, limit);
1175 tipc_bclink_unlock(net);
1179 int tipc_nl_bc_link_set(struct net *net, struct nlattr *attrs[])
1183 struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
1185 if (!attrs[TIPC_NLA_LINK_PROP])
1188 err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP], props);
1192 if (!props[TIPC_NLA_PROP_WIN])
1195 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1197 return tipc_bclink_set_queue_limits(net, win);
1200 int tipc_bcast_init(struct net *net)
1202 struct tipc_net *tn = tipc_net(net);
1203 struct tipc_bcbearer *bcb = NULL;
1204 struct tipc_bc_base *bb = NULL;
1205 struct tipc_link *l = NULL;
1207 bcb = kzalloc(sizeof(*bcb), GFP_ATOMIC);
1212 bcb->bearer.window = BCLINK_WIN_DEFAULT;
1213 bcb->bearer.mtu = MAX_PKT_DEFAULT_MCAST;
1214 bcb->bearer.identity = MAX_BEARERS;
1216 bcb->bearer.media = &bcb->media;
1217 bcb->media.send_msg = tipc_bcbearer_send;
1218 sprintf(bcb->media.name, "tipc-broadcast");
1219 strcpy(bcb->bearer.name, bcb->media.name);
1221 bb = kzalloc(sizeof(*bb), GFP_ATOMIC);
1225 __skb_queue_head_init(&bb->arrvq);
1226 spin_lock_init(&tipc_net(net)->bclock);
1229 if (!tipc_link_bc_create(&bb->node, 0, 0,
1240 rcu_assign_pointer(tn->bearer_list[MAX_BEARERS], &bcb->bearer);
1249 void tipc_bcast_reinit(struct net *net)
1251 struct tipc_bc_base *b = tipc_bc_base(net);
1253 msg_set_prevnode(b->link->pmsg, tipc_own_addr(net));
1256 void tipc_bcast_stop(struct net *net)
1258 struct tipc_net *tn = net_generic(net, tipc_net_id);
1260 tipc_bclink_lock(net);
1261 tipc_link_purge_queues(tn->bcl);
1262 tipc_bclink_unlock(net);
1263 RCU_INIT_POINTER(tn->bearer_list[BCBEARER], NULL);
1265 kfree(tn->bcbearer);
1271 * tipc_nmap_add - add a node to a node map
1273 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
1275 int n = tipc_node(node);
1277 u32 mask = (1 << (n % WSIZE));
1279 if ((nm_ptr->map[w] & mask) == 0) {
1281 nm_ptr->map[w] |= mask;
1286 * tipc_nmap_remove - remove a node from a node map
1288 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
1290 int n = tipc_node(node);
1292 u32 mask = (1 << (n % WSIZE));
1294 if ((nm_ptr->map[w] & mask) != 0) {
1295 nm_ptr->map[w] &= ~mask;
1301 * tipc_nmap_diff - find differences between node maps
1302 * @nm_a: input node map A
1303 * @nm_b: input node map B
1304 * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
1306 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
1307 struct tipc_node_map *nm_b,
1308 struct tipc_node_map *nm_diff)
1310 int stop = ARRAY_SIZE(nm_a->map);
1315 memset(nm_diff, 0, sizeof(*nm_diff));
1316 for (w = 0; w < stop; w++) {
1317 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
1318 nm_diff->map[w] = map;
1320 for (b = 0 ; b < WSIZE; b++) {