2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, 2012-2015, Ericsson AB
5 * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
42 #include "name_distr.h"
46 #include <linux/pkt_sched.h>
49 * Error message prefixes
51 static const char *link_co_err = "Link tunneling error, ";
52 static const char *link_rst_msg = "Resetting link ";
53 static const char tipc_bclink_name[] = "broadcast-link";
55 static const struct nla_policy tipc_nl_link_policy[TIPC_NLA_LINK_MAX + 1] = {
56 [TIPC_NLA_LINK_UNSPEC] = { .type = NLA_UNSPEC },
57 [TIPC_NLA_LINK_NAME] = {
59 .len = TIPC_MAX_LINK_NAME
61 [TIPC_NLA_LINK_MTU] = { .type = NLA_U32 },
62 [TIPC_NLA_LINK_BROADCAST] = { .type = NLA_FLAG },
63 [TIPC_NLA_LINK_UP] = { .type = NLA_FLAG },
64 [TIPC_NLA_LINK_ACTIVE] = { .type = NLA_FLAG },
65 [TIPC_NLA_LINK_PROP] = { .type = NLA_NESTED },
66 [TIPC_NLA_LINK_STATS] = { .type = NLA_NESTED },
67 [TIPC_NLA_LINK_RX] = { .type = NLA_U32 },
68 [TIPC_NLA_LINK_TX] = { .type = NLA_U32 }
71 /* Properties valid for media, bearar and link */
72 static const struct nla_policy tipc_nl_prop_policy[TIPC_NLA_PROP_MAX + 1] = {
73 [TIPC_NLA_PROP_UNSPEC] = { .type = NLA_UNSPEC },
74 [TIPC_NLA_PROP_PRIO] = { .type = NLA_U32 },
75 [TIPC_NLA_PROP_TOL] = { .type = NLA_U32 },
76 [TIPC_NLA_PROP_WIN] = { .type = NLA_U32 }
79 /* Send states for broadcast NACKs
82 BC_NACK_SND_CONDITIONAL,
83 BC_NACK_SND_UNCONDITIONAL,
88 * Interval between NACKs when packets arrive out of order
90 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
92 * Out-of-range value for link session numbers
94 #define WILDCARD_SESSION 0x10000
99 LINK_ESTABLISHED = 0xe,
100 LINK_ESTABLISHING = 0xe << 4,
101 LINK_RESET = 0x1 << 8,
102 LINK_RESETTING = 0x2 << 12,
103 LINK_PEER_RESET = 0xd << 16,
104 LINK_FAILINGOVER = 0xf << 20,
105 LINK_SYNCHING = 0xc << 24
108 /* Link FSM state checking routines
110 static int link_is_up(struct tipc_link *l)
112 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
115 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
116 struct sk_buff_head *xmitq);
117 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
118 u16 rcvgap, int tolerance, int priority,
119 struct sk_buff_head *xmitq);
120 static void link_reset_statistics(struct tipc_link *l_ptr);
121 static void link_print(struct tipc_link *l_ptr, const char *str);
122 static void tipc_link_build_nack_msg(struct tipc_link *l,
123 struct sk_buff_head *xmitq);
124 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
125 struct sk_buff_head *xmitq);
126 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
129 * Simple non-static link routines (i.e. referenced outside this file)
131 bool tipc_link_is_up(struct tipc_link *l)
133 return link_is_up(l);
136 bool tipc_link_peer_is_down(struct tipc_link *l)
138 return l->state == LINK_PEER_RESET;
141 bool tipc_link_is_reset(struct tipc_link *l)
143 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
146 bool tipc_link_is_establishing(struct tipc_link *l)
148 return l->state == LINK_ESTABLISHING;
151 bool tipc_link_is_synching(struct tipc_link *l)
153 return l->state == LINK_SYNCHING;
156 bool tipc_link_is_failingover(struct tipc_link *l)
158 return l->state == LINK_FAILINGOVER;
161 bool tipc_link_is_blocked(struct tipc_link *l)
163 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
166 bool link_is_bc_sndlink(struct tipc_link *l)
168 return !l->bc_sndlink;
171 bool link_is_bc_rcvlink(struct tipc_link *l)
173 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
176 int tipc_link_is_active(struct tipc_link *l)
178 struct tipc_node *n = l->owner;
180 return (node_active_link(n, 0) == l) || (node_active_link(n, 1) == l);
183 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
184 struct tipc_link *uc_l,
185 struct sk_buff_head *xmitq)
187 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
190 rcv_l->acked = snd_l->snd_nxt - 1;
191 tipc_link_build_bc_init_msg(uc_l, xmitq);
194 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
195 struct tipc_link *rcv_l,
196 struct sk_buff_head *xmitq)
198 u16 ack = snd_l->snd_nxt - 1;
201 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
202 tipc_link_reset(rcv_l);
203 rcv_l->state = LINK_RESET;
204 if (!snd_l->ackers) {
205 tipc_link_reset(snd_l);
206 __skb_queue_purge(xmitq);
210 int tipc_link_bc_peers(struct tipc_link *l)
215 static u32 link_own_addr(struct tipc_link *l)
217 return msg_prevnode(l->pmsg);
221 * tipc_link_create - create a new link
222 * @n: pointer to associated node
223 * @if_name: associated interface name
224 * @bearer_id: id (index) of associated bearer
225 * @tolerance: link tolerance to be used by link
226 * @net_plane: network plane (A,B,c..) this link belongs to
227 * @mtu: mtu to be advertised by link
228 * @priority: priority to be used by link
229 * @window: send window to be used by link
230 * @session: session to be used by link
231 * @ownnode: identity of own node
232 * @peer: node id of peer node
233 * @peer_caps: bitmap describing peer node capabilities
234 * @maddr: media address to be used
235 * @bc_sndlink: the namespace global link used for broadcast sending
236 * @bc_rcvlink: the peer specific link used for broadcast reception
237 * @inputq: queue to put messages ready for delivery
238 * @namedq: queue to put binding table update messages ready for delivery
239 * @link: return value, pointer to put the created link
241 * Returns true if link was created, otherwise false
243 bool tipc_link_create(struct tipc_node *n, char *if_name, int bearer_id,
244 int tolerance, char net_plane, u32 mtu, int priority,
245 int window, u32 session, u32 ownnode, u32 peer,
247 struct tipc_media_addr *maddr,
248 struct tipc_link *bc_sndlink,
249 struct tipc_link *bc_rcvlink,
250 struct sk_buff_head *inputq,
251 struct sk_buff_head *namedq,
252 struct tipc_link **link)
255 struct tipc_msg *hdr;
257 l = kzalloc(sizeof(*l), GFP_ATOMIC);
261 l->pmsg = (struct tipc_msg *)&l->proto_msg;
263 tipc_msg_init(ownnode, hdr, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, peer);
264 msg_set_size(hdr, sizeof(l->proto_msg));
265 msg_set_session(hdr, session);
266 msg_set_bearer_id(hdr, l->bearer_id);
268 /* Note: peer i/f name is completed by reset/activate message */
269 sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
270 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
271 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
272 strcpy((char *)msg_data(hdr), if_name);
275 l->peer_caps = peer_caps;
276 l->media_addr = maddr;
278 l->peer_session = WILDCARD_SESSION;
279 l->bearer_id = bearer_id;
280 l->tolerance = tolerance;
281 l->net_plane = net_plane;
282 l->advertised_mtu = mtu;
284 l->priority = priority;
285 tipc_link_set_queue_limits(l, window);
287 l->bc_sndlink = bc_sndlink;
288 l->bc_rcvlink = bc_rcvlink;
291 l->state = LINK_RESETTING;
292 __skb_queue_head_init(&l->transmq);
293 __skb_queue_head_init(&l->backlogq);
294 __skb_queue_head_init(&l->deferdq);
295 skb_queue_head_init(&l->wakeupq);
296 skb_queue_head_init(l->inputq);
301 * tipc_link_bc_create - create new link to be used for broadcast
302 * @n: pointer to associated node
303 * @mtu: mtu to be used
304 * @window: send window to be used
305 * @inputq: queue to put messages ready for delivery
306 * @namedq: queue to put binding table update messages ready for delivery
307 * @link: return value, pointer to put the created link
309 * Returns true if link was created, otherwise false
311 bool tipc_link_bc_create(struct tipc_node *n, u32 ownnode, u32 peer,
312 int mtu, int window, u16 peer_caps,
313 struct sk_buff_head *inputq,
314 struct sk_buff_head *namedq,
315 struct tipc_link *bc_sndlink,
316 struct tipc_link **link)
320 if (!tipc_link_create(n, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
321 0, ownnode, peer, peer_caps, NULL, bc_sndlink,
322 NULL, inputq, namedq, link))
326 strcpy(l->name, tipc_bclink_name);
328 l->state = LINK_RESET;
332 /* Broadcast send link is always up */
333 if (link_is_bc_sndlink(l))
334 l->state = LINK_ESTABLISHED;
340 * tipc_link_fsm_evt - link finite state machine
341 * @l: pointer to link
342 * @evt: state machine event to be processed
344 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
351 case LINK_PEER_RESET_EVT:
352 l->state = LINK_PEER_RESET;
355 l->state = LINK_RESET;
357 case LINK_FAILURE_EVT:
358 case LINK_FAILOVER_BEGIN_EVT:
359 case LINK_ESTABLISH_EVT:
360 case LINK_FAILOVER_END_EVT:
361 case LINK_SYNCH_BEGIN_EVT:
362 case LINK_SYNCH_END_EVT:
369 case LINK_PEER_RESET_EVT:
370 l->state = LINK_ESTABLISHING;
372 case LINK_FAILOVER_BEGIN_EVT:
373 l->state = LINK_FAILINGOVER;
374 case LINK_FAILURE_EVT:
376 case LINK_ESTABLISH_EVT:
377 case LINK_FAILOVER_END_EVT:
379 case LINK_SYNCH_BEGIN_EVT:
380 case LINK_SYNCH_END_EVT:
385 case LINK_PEER_RESET:
388 l->state = LINK_ESTABLISHING;
390 case LINK_PEER_RESET_EVT:
391 case LINK_ESTABLISH_EVT:
392 case LINK_FAILURE_EVT:
394 case LINK_SYNCH_BEGIN_EVT:
395 case LINK_SYNCH_END_EVT:
396 case LINK_FAILOVER_BEGIN_EVT:
397 case LINK_FAILOVER_END_EVT:
402 case LINK_FAILINGOVER:
404 case LINK_FAILOVER_END_EVT:
405 l->state = LINK_RESET;
407 case LINK_PEER_RESET_EVT:
409 case LINK_ESTABLISH_EVT:
410 case LINK_FAILURE_EVT:
412 case LINK_FAILOVER_BEGIN_EVT:
413 case LINK_SYNCH_BEGIN_EVT:
414 case LINK_SYNCH_END_EVT:
419 case LINK_ESTABLISHING:
421 case LINK_ESTABLISH_EVT:
422 l->state = LINK_ESTABLISHED;
424 case LINK_FAILOVER_BEGIN_EVT:
425 l->state = LINK_FAILINGOVER;
428 l->state = LINK_RESET;
430 case LINK_FAILURE_EVT:
431 case LINK_PEER_RESET_EVT:
432 case LINK_SYNCH_BEGIN_EVT:
433 case LINK_FAILOVER_END_EVT:
435 case LINK_SYNCH_END_EVT:
440 case LINK_ESTABLISHED:
442 case LINK_PEER_RESET_EVT:
443 l->state = LINK_PEER_RESET;
444 rc |= TIPC_LINK_DOWN_EVT;
446 case LINK_FAILURE_EVT:
447 l->state = LINK_RESETTING;
448 rc |= TIPC_LINK_DOWN_EVT;
451 l->state = LINK_RESET;
453 case LINK_ESTABLISH_EVT:
454 case LINK_SYNCH_END_EVT:
456 case LINK_SYNCH_BEGIN_EVT:
457 l->state = LINK_SYNCHING;
459 case LINK_FAILOVER_BEGIN_EVT:
460 case LINK_FAILOVER_END_EVT:
467 case LINK_PEER_RESET_EVT:
468 l->state = LINK_PEER_RESET;
469 rc |= TIPC_LINK_DOWN_EVT;
471 case LINK_FAILURE_EVT:
472 l->state = LINK_RESETTING;
473 rc |= TIPC_LINK_DOWN_EVT;
476 l->state = LINK_RESET;
478 case LINK_ESTABLISH_EVT:
479 case LINK_SYNCH_BEGIN_EVT:
481 case LINK_SYNCH_END_EVT:
482 l->state = LINK_ESTABLISHED;
484 case LINK_FAILOVER_BEGIN_EVT:
485 case LINK_FAILOVER_END_EVT:
491 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
495 pr_err("Illegal FSM event %x in state %x on link %s\n",
496 evt, l->state, l->name);
500 /* link_profile_stats - update statistical profiling of traffic
502 static void link_profile_stats(struct tipc_link *l)
505 struct tipc_msg *msg;
508 /* Update counters used in statistical profiling of send traffic */
509 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
510 l->stats.queue_sz_counts++;
512 skb = skb_peek(&l->transmq);
516 length = msg_size(msg);
518 if (msg_user(msg) == MSG_FRAGMENTER) {
519 if (msg_type(msg) != FIRST_FRAGMENT)
521 length = msg_size(msg_get_wrapped(msg));
523 l->stats.msg_lengths_total += length;
524 l->stats.msg_length_counts++;
526 l->stats.msg_length_profile[0]++;
527 else if (length <= 256)
528 l->stats.msg_length_profile[1]++;
529 else if (length <= 1024)
530 l->stats.msg_length_profile[2]++;
531 else if (length <= 4096)
532 l->stats.msg_length_profile[3]++;
533 else if (length <= 16384)
534 l->stats.msg_length_profile[4]++;
535 else if (length <= 32768)
536 l->stats.msg_length_profile[5]++;
538 l->stats.msg_length_profile[6]++;
541 /* tipc_link_timeout - perform periodic task as instructed from node timeout
543 /* tipc_link_timeout - perform periodic task as instructed from node timeout
545 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
548 int mtyp = STATE_MSG;
551 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
552 u16 bc_acked = l->bc_rcvlink->acked;
553 bool bc_up = link_is_up(l->bc_rcvlink);
555 link_profile_stats(l);
558 case LINK_ESTABLISHED:
560 if (!l->silent_intv_cnt) {
561 if (bc_up && (bc_acked != bc_snt))
563 } else if (l->silent_intv_cnt <= l->abort_limit) {
567 rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
569 l->silent_intv_cnt++;
575 case LINK_ESTABLISHING:
579 case LINK_PEER_RESET:
581 case LINK_FAILINGOVER:
588 tipc_link_build_proto_msg(l, mtyp, prb, 0, 0, 0, xmitq);
594 * link_schedule_user - schedule a message sender for wakeup after congestion
595 * @link: congested link
596 * @list: message that was attempted sent
597 * Create pseudo msg to send back to user when congestion abates
598 * Does not consume buffer list
600 static int link_schedule_user(struct tipc_link *link, struct sk_buff_head *list)
602 struct tipc_msg *msg = buf_msg(skb_peek(list));
603 int imp = msg_importance(msg);
604 u32 oport = msg_origport(msg);
605 u32 addr = link_own_addr(link);
608 /* This really cannot happen... */
609 if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
610 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
613 /* Non-blocking sender: */
614 if (TIPC_SKB_CB(skb_peek(list))->wakeup_pending)
617 /* Create and schedule wakeup pseudo message */
618 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
619 addr, addr, oport, 0, 0);
622 TIPC_SKB_CB(skb)->chain_sz = skb_queue_len(list);
623 TIPC_SKB_CB(skb)->chain_imp = imp;
624 skb_queue_tail(&link->wakeupq, skb);
625 link->stats.link_congs++;
630 * link_prepare_wakeup - prepare users for wakeup after congestion
631 * @link: congested link
632 * Move a number of waiting users, as permitted by available space in
633 * the send queue, from link wait queue to node wait queue for wakeup
635 void link_prepare_wakeup(struct tipc_link *l)
637 int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
639 struct sk_buff *skb, *tmp;
641 skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
642 imp = TIPC_SKB_CB(skb)->chain_imp;
643 lim = l->window + l->backlog[imp].limit;
644 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
645 if ((pnd[imp] + l->backlog[imp].len) >= lim)
647 skb_unlink(skb, &l->wakeupq);
648 skb_queue_tail(l->inputq, skb);
653 * tipc_link_reset_fragments - purge link's inbound message fragments queue
654 * @l_ptr: pointer to link
656 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
658 kfree_skb(l_ptr->reasm_buf);
659 l_ptr->reasm_buf = NULL;
662 void tipc_link_purge_backlog(struct tipc_link *l)
664 __skb_queue_purge(&l->backlogq);
665 l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
666 l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
667 l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
668 l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
669 l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
673 * tipc_link_purge_queues - purge all pkt queues associated with link
674 * @l_ptr: pointer to link
676 void tipc_link_purge_queues(struct tipc_link *l_ptr)
678 __skb_queue_purge(&l_ptr->deferdq);
679 __skb_queue_purge(&l_ptr->transmq);
680 tipc_link_purge_backlog(l_ptr);
681 tipc_link_reset_fragments(l_ptr);
684 void tipc_link_reset(struct tipc_link *l)
686 /* Link is down, accept any session */
687 l->peer_session = WILDCARD_SESSION;
689 /* If peer is up, it only accepts an incremented session number */
690 msg_set_session(l->pmsg, msg_session(l->pmsg) + 1);
692 /* Prepare for renewed mtu size negotiation */
693 l->mtu = l->advertised_mtu;
695 /* Clean up all queues: */
696 __skb_queue_purge(&l->transmq);
697 __skb_queue_purge(&l->deferdq);
698 skb_queue_splice_init(&l->wakeupq, l->inputq);
700 tipc_link_purge_backlog(l);
701 kfree_skb(l->reasm_buf);
702 kfree_skb(l->failover_reasm_skb);
704 l->failover_reasm_skb = NULL;
709 l->silent_intv_cnt = 0;
710 l->stats.recv_info = 0;
712 l->bc_peer_is_up = false;
713 link_reset_statistics(l);
717 * __tipc_link_xmit(): same as tipc_link_xmit, but destlink is known & locked
719 * @list: chain of buffers containing message
721 * Consumes the buffer chain, except when returning an error code,
722 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
723 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
725 int __tipc_link_xmit(struct net *net, struct tipc_link *link,
726 struct sk_buff_head *list)
728 struct tipc_msg *msg = buf_msg(skb_peek(list));
729 unsigned int maxwin = link->window;
730 unsigned int i, imp = msg_importance(msg);
731 uint mtu = link->mtu;
732 u16 ack = mod(link->rcv_nxt - 1);
733 u16 seqno = link->snd_nxt;
734 u16 bc_ack = link->bc_rcvlink->rcv_nxt - 1;
735 struct tipc_media_addr *addr = link->media_addr;
736 struct sk_buff_head *transmq = &link->transmq;
737 struct sk_buff_head *backlogq = &link->backlogq;
738 struct sk_buff *skb, *bskb;
740 /* Match msg importance against this and all higher backlog limits: */
741 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
742 if (unlikely(link->backlog[i].len >= link->backlog[i].limit))
743 return link_schedule_user(link, list);
745 if (unlikely(msg_size(msg) > mtu))
748 /* Prepare each packet for sending, and add to relevant queue: */
749 while (skb_queue_len(list)) {
750 skb = skb_peek(list);
752 msg_set_seqno(msg, seqno);
753 msg_set_ack(msg, ack);
754 msg_set_bcast_ack(msg, bc_ack);
756 if (likely(skb_queue_len(transmq) < maxwin)) {
758 __skb_queue_tail(transmq, skb);
759 tipc_bearer_send(net, link->bearer_id, skb, addr);
760 link->rcv_unacked = 0;
764 if (tipc_msg_bundle(skb_peek_tail(backlogq), msg, mtu)) {
765 kfree_skb(__skb_dequeue(list));
766 link->stats.sent_bundled++;
769 if (tipc_msg_make_bundle(&bskb, msg, mtu, link->addr)) {
770 kfree_skb(__skb_dequeue(list));
771 __skb_queue_tail(backlogq, bskb);
772 link->backlog[msg_importance(buf_msg(bskb))].len++;
773 link->stats.sent_bundled++;
774 link->stats.sent_bundles++;
777 link->backlog[imp].len += skb_queue_len(list);
778 skb_queue_splice_tail_init(list, backlogq);
780 link->snd_nxt = seqno;
785 * tipc_link_xmit(): enqueue buffer list according to queue situation
787 * @list: chain of buffers containing message
788 * @xmitq: returned list of packets to be sent by caller
790 * Consumes the buffer chain, except when returning -ELINKCONG,
791 * since the caller then may want to make more send attempts.
792 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
793 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
795 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
796 struct sk_buff_head *xmitq)
798 struct tipc_msg *hdr = buf_msg(skb_peek(list));
799 unsigned int maxwin = l->window;
800 unsigned int i, imp = msg_importance(hdr);
801 unsigned int mtu = l->mtu;
802 u16 ack = l->rcv_nxt - 1;
803 u16 seqno = l->snd_nxt;
804 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
805 struct sk_buff_head *transmq = &l->transmq;
806 struct sk_buff_head *backlogq = &l->backlogq;
807 struct sk_buff *skb, *_skb, *bskb;
809 /* Match msg importance against this and all higher backlog limits: */
810 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
811 if (unlikely(l->backlog[i].len >= l->backlog[i].limit))
812 return link_schedule_user(l, list);
814 if (unlikely(msg_size(hdr) > mtu))
817 /* Prepare each packet for sending, and add to relevant queue: */
818 while (skb_queue_len(list)) {
819 skb = skb_peek(list);
821 msg_set_seqno(hdr, seqno);
822 msg_set_ack(hdr, ack);
823 msg_set_bcast_ack(hdr, bc_ack);
825 if (likely(skb_queue_len(transmq) < maxwin)) {
826 _skb = skb_clone(skb, GFP_ATOMIC);
830 __skb_queue_tail(transmq, skb);
831 __skb_queue_tail(xmitq, _skb);
832 TIPC_SKB_CB(skb)->ackers = l->ackers;
837 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
838 kfree_skb(__skb_dequeue(list));
839 l->stats.sent_bundled++;
842 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
843 kfree_skb(__skb_dequeue(list));
844 __skb_queue_tail(backlogq, bskb);
845 l->backlog[msg_importance(buf_msg(bskb))].len++;
846 l->stats.sent_bundled++;
847 l->stats.sent_bundles++;
850 l->backlog[imp].len += skb_queue_len(list);
851 skb_queue_splice_tail_init(list, backlogq);
858 * tipc_link_push_packets - push unsent packets to bearer
860 * Push out the unsent messages of a link where congestion
861 * has abated. Node is locked.
863 * Called with node locked
865 void tipc_link_push_packets(struct tipc_link *link)
868 struct tipc_msg *msg;
869 u16 seqno = link->snd_nxt;
870 u16 ack = mod(link->rcv_nxt - 1);
872 while (skb_queue_len(&link->transmq) < link->window) {
873 skb = __skb_dequeue(&link->backlogq);
876 TIPC_SKB_CB(skb)->ackers = link->ackers;
878 link->backlog[msg_importance(msg)].len--;
879 msg_set_ack(msg, ack);
880 msg_set_seqno(msg, seqno);
881 seqno = mod(seqno + 1);
882 msg_set_bcast_ack(msg, link->owner->bclink.last_in);
883 link->rcv_unacked = 0;
884 __skb_queue_tail(&link->transmq, skb);
885 tipc_bearer_send(link->owner->net, link->bearer_id,
886 skb, link->media_addr);
888 link->snd_nxt = seqno;
891 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
893 struct sk_buff *skb, *_skb;
894 struct tipc_msg *hdr;
895 u16 seqno = l->snd_nxt;
896 u16 ack = l->rcv_nxt - 1;
897 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
899 while (skb_queue_len(&l->transmq) < l->window) {
900 skb = skb_peek(&l->backlogq);
903 _skb = skb_clone(skb, GFP_ATOMIC);
906 __skb_dequeue(&l->backlogq);
908 l->backlog[msg_importance(hdr)].len--;
909 __skb_queue_tail(&l->transmq, skb);
910 __skb_queue_tail(xmitq, _skb);
911 TIPC_SKB_CB(skb)->ackers = l->ackers;
912 msg_set_seqno(hdr, seqno);
913 msg_set_ack(hdr, ack);
914 msg_set_bcast_ack(hdr, bc_ack);
921 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
923 struct tipc_msg *hdr = buf_msg(skb);
925 pr_warn("Retransmission failure on link <%s>\n", l->name);
926 link_print(l, "Resetting link ");
927 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
928 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
929 pr_info("sqno %u, prev: %x, src: %x\n",
930 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
933 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *skb,
936 struct tipc_msg *msg;
943 /* Detect repeated retransmit failures */
944 if (l_ptr->last_retransm == msg_seqno(msg)) {
945 if (++l_ptr->stale_count > 100) {
946 link_retransmit_failure(l_ptr, skb);
950 l_ptr->last_retransm = msg_seqno(msg);
951 l_ptr->stale_count = 1;
954 skb_queue_walk_from(&l_ptr->transmq, skb) {
958 msg_set_ack(msg, mod(l_ptr->rcv_nxt - 1));
959 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
960 tipc_bearer_send(l_ptr->owner->net, l_ptr->bearer_id, skb,
963 l_ptr->stats.retransmitted++;
967 int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
968 struct sk_buff_head *xmitq)
970 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
971 struct tipc_msg *hdr;
972 u16 ack = l->rcv_nxt - 1;
973 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
978 /* Detect repeated retransmit failures on same packet */
979 if (likely(l->last_retransm != buf_seqno(skb))) {
980 l->last_retransm = buf_seqno(skb);
982 } else if (++l->stale_count > 100) {
983 link_retransmit_failure(l, skb);
984 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
987 /* Move forward to where retransmission should start */
988 skb_queue_walk(&l->transmq, skb) {
989 if (!less(buf_seqno(skb), from))
993 skb_queue_walk_from(&l->transmq, skb) {
994 if (more(buf_seqno(skb), to))
997 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1000 hdr = buf_msg(_skb);
1001 msg_set_ack(hdr, ack);
1002 msg_set_bcast_ack(hdr, bc_ack);
1003 _skb->priority = TC_PRIO_CONTROL;
1004 __skb_queue_tail(xmitq, _skb);
1005 l->stats.retransmitted++;
1010 /* tipc_data_input - deliver data and name distr msgs to upper layer
1012 * Consumes buffer if message is of right type
1013 * Node lock must be held
1015 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1016 struct sk_buff_head *inputq)
1018 switch (msg_user(buf_msg(skb))) {
1019 case TIPC_LOW_IMPORTANCE:
1020 case TIPC_MEDIUM_IMPORTANCE:
1021 case TIPC_HIGH_IMPORTANCE:
1022 case TIPC_CRITICAL_IMPORTANCE:
1024 skb_queue_tail(inputq, skb);
1026 case NAME_DISTRIBUTOR:
1027 l->bc_rcvlink->state = LINK_ESTABLISHED;
1028 skb_queue_tail(l->namedq, skb);
1031 case TUNNEL_PROTOCOL:
1032 case MSG_FRAGMENTER:
1033 case BCAST_PROTOCOL:
1036 pr_warn("Dropping received illegal msg type\n");
1042 /* tipc_link_input - process packet that has passed link protocol check
1046 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1047 struct sk_buff_head *inputq)
1049 struct tipc_msg *hdr = buf_msg(skb);
1050 struct sk_buff **reasm_skb = &l->reasm_buf;
1051 struct sk_buff *iskb;
1052 struct sk_buff_head tmpq;
1053 int usr = msg_user(hdr);
1058 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1059 if (msg_type(hdr) == SYNCH_MSG) {
1060 __skb_queue_purge(&l->deferdq);
1063 if (!tipc_msg_extract(skb, &iskb, &ipos))
1068 if (less(msg_seqno(hdr), l->drop_point))
1070 if (tipc_data_input(l, skb, inputq))
1072 usr = msg_user(hdr);
1073 reasm_skb = &l->failover_reasm_skb;
1076 if (usr == MSG_BUNDLER) {
1077 skb_queue_head_init(&tmpq);
1078 l->stats.recv_bundles++;
1079 l->stats.recv_bundled += msg_msgcnt(hdr);
1080 while (tipc_msg_extract(skb, &iskb, &pos))
1081 tipc_data_input(l, iskb, &tmpq);
1082 tipc_skb_queue_splice_tail(&tmpq, inputq);
1084 } else if (usr == MSG_FRAGMENTER) {
1085 l->stats.recv_fragments++;
1086 if (tipc_buf_append(reasm_skb, &skb)) {
1087 l->stats.recv_fragmented++;
1088 tipc_data_input(l, skb, inputq);
1089 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1090 pr_warn_ratelimited("Unable to build fragment list\n");
1091 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1094 } else if (usr == BCAST_PROTOCOL) {
1095 tipc_bcast_lock(l->owner->net);
1096 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1097 tipc_bcast_unlock(l->owner->net);
1104 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1106 bool released = false;
1107 struct sk_buff *skb, *tmp;
1109 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1110 if (more(buf_seqno(skb), acked))
1112 __skb_unlink(skb, &l->transmq);
1119 /* tipc_link_build_ack_msg: prepare link acknowledge message for transmission
1121 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1122 * risk of ack storms towards the sender
1124 int tipc_link_build_ack_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1129 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1130 if (link_is_bc_rcvlink(l)) {
1131 if (((l->rcv_nxt ^ link_own_addr(l)) & 0xf) != 0xf)
1134 return TIPC_LINK_SND_BC_ACK;
1139 l->stats.sent_acks++;
1140 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1144 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1146 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1148 int mtyp = RESET_MSG;
1150 if (l->state == LINK_ESTABLISHING)
1151 mtyp = ACTIVATE_MSG;
1153 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1156 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1158 static void tipc_link_build_nack_msg(struct tipc_link *l,
1159 struct sk_buff_head *xmitq)
1161 u32 def_cnt = ++l->stats.deferred_recv;
1163 if (link_is_bc_rcvlink(l))
1166 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1167 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1170 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1171 * @l: the link that should handle the message
1173 * @xmitq: queue to place packets to be sent after this call
1175 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1176 struct sk_buff_head *xmitq)
1178 struct sk_buff_head *defq = &l->deferdq;
1179 struct tipc_msg *hdr;
1180 u16 seqno, rcv_nxt, win_lim;
1185 seqno = msg_seqno(hdr);
1186 rcv_nxt = l->rcv_nxt;
1187 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1189 /* Verify and update link state */
1190 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1191 return tipc_link_proto_rcv(l, skb, xmitq);
1193 if (unlikely(!link_is_up(l))) {
1194 if (l->state == LINK_ESTABLISHING)
1195 rc = TIPC_LINK_UP_EVT;
1199 /* Don't send probe at next timeout expiration */
1200 l->silent_intv_cnt = 0;
1202 /* Drop if outside receive window */
1203 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1204 l->stats.duplicates++;
1208 /* Forward queues and wake up waiting users */
1209 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1210 tipc_link_advance_backlog(l, xmitq);
1211 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1212 link_prepare_wakeup(l);
1215 /* Defer delivery if sequence gap */
1216 if (unlikely(seqno != rcv_nxt)) {
1217 __tipc_skb_queue_sorted(defq, seqno, skb);
1218 tipc_link_build_nack_msg(l, xmitq);
1222 /* Deliver packet */
1224 l->stats.recv_info++;
1225 if (!tipc_data_input(l, skb, l->inputq))
1226 rc |= tipc_link_input(l, skb, l->inputq);
1227 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1228 rc |= tipc_link_build_ack_msg(l, xmitq);
1229 if (unlikely(rc & ~TIPC_LINK_SND_BC_ACK))
1231 } while ((skb = __skb_dequeue(defq)));
1240 * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1242 * Returns increase in queue length (i.e. 0 or 1)
1244 u32 tipc_link_defer_pkt(struct sk_buff_head *list, struct sk_buff *skb)
1246 struct sk_buff *skb1;
1247 u16 seq_no = buf_seqno(skb);
1250 if (skb_queue_empty(list)) {
1251 __skb_queue_tail(list, skb);
1256 if (less(buf_seqno(skb_peek_tail(list)), seq_no)) {
1257 __skb_queue_tail(list, skb);
1261 /* Locate insertion point in queue, then insert; discard if duplicate */
1262 skb_queue_walk(list, skb1) {
1263 u16 curr_seqno = buf_seqno(skb1);
1265 if (seq_no == curr_seqno) {
1270 if (less(seq_no, curr_seqno))
1274 __skb_queue_before(list, skb1, skb);
1279 * Send protocol message to the other endpoint.
1281 void tipc_link_proto_xmit(struct tipc_link *l, u32 msg_typ, int probe_msg,
1282 u32 gap, u32 tolerance, u32 priority)
1284 struct sk_buff *skb = NULL;
1285 struct sk_buff_head xmitq;
1287 __skb_queue_head_init(&xmitq);
1288 tipc_link_build_proto_msg(l, msg_typ, probe_msg, gap,
1289 tolerance, priority, &xmitq);
1290 skb = __skb_dequeue(&xmitq);
1293 tipc_bearer_send(l->owner->net, l->bearer_id, skb, l->media_addr);
1298 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1299 u16 rcvgap, int tolerance, int priority,
1300 struct sk_buff_head *xmitq)
1302 struct sk_buff *skb = NULL;
1303 struct tipc_msg *hdr = l->pmsg;
1304 bool node_up = link_is_up(l->bc_rcvlink);
1306 /* Don't send protocol message during reset or link failover */
1307 if (tipc_link_is_blocked(l))
1310 msg_set_type(hdr, mtyp);
1311 msg_set_net_plane(hdr, l->net_plane);
1312 msg_set_next_sent(hdr, l->snd_nxt);
1313 msg_set_ack(hdr, l->rcv_nxt - 1);
1314 msg_set_bcast_ack(hdr, l->bc_rcvlink->rcv_nxt - 1);
1315 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1316 msg_set_link_tolerance(hdr, tolerance);
1317 msg_set_linkprio(hdr, priority);
1318 msg_set_redundant_link(hdr, node_up);
1319 msg_set_seq_gap(hdr, 0);
1321 /* Compatibility: created msg must not be in sequence with pkt flow */
1322 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1324 if (mtyp == STATE_MSG) {
1325 if (!tipc_link_is_up(l))
1328 /* Override rcvgap if there are packets in deferred queue */
1329 if (!skb_queue_empty(&l->deferdq))
1330 rcvgap = buf_seqno(skb_peek(&l->deferdq)) - l->rcv_nxt;
1332 msg_set_seq_gap(hdr, rcvgap);
1333 l->stats.sent_nacks++;
1335 msg_set_probe(hdr, probe);
1337 l->stats.sent_probes++;
1338 l->stats.sent_states++;
1341 /* RESET_MSG or ACTIVATE_MSG */
1342 msg_set_max_pkt(hdr, l->advertised_mtu);
1343 msg_set_ack(hdr, l->rcv_nxt - 1);
1344 msg_set_next_sent(hdr, 1);
1346 skb = tipc_buf_acquire(msg_size(hdr));
1349 skb_copy_to_linear_data(skb, hdr, msg_size(hdr));
1350 skb->priority = TC_PRIO_CONTROL;
1351 __skb_queue_tail(xmitq, skb);
1354 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1355 * with contents of the link's transmit and backlog queues.
1357 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1358 int mtyp, struct sk_buff_head *xmitq)
1360 struct sk_buff *skb, *tnlskb;
1361 struct tipc_msg *hdr, tnlhdr;
1362 struct sk_buff_head *queue = &l->transmq;
1363 struct sk_buff_head tmpxq, tnlq;
1364 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1369 skb_queue_head_init(&tnlq);
1370 skb_queue_head_init(&tmpxq);
1372 /* At least one packet required for safe algorithm => add dummy */
1373 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1374 BASIC_H_SIZE, 0, l->addr, link_own_addr(l),
1375 0, 0, TIPC_ERR_NO_PORT);
1377 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1380 skb_queue_tail(&tnlq, skb);
1381 tipc_link_xmit(l, &tnlq, &tmpxq);
1382 __skb_queue_purge(&tmpxq);
1384 /* Initialize reusable tunnel packet header */
1385 tipc_msg_init(link_own_addr(l), &tnlhdr, TUNNEL_PROTOCOL,
1386 mtyp, INT_H_SIZE, l->addr);
1387 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1388 msg_set_msgcnt(&tnlhdr, pktcnt);
1389 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1391 /* Wrap each packet into a tunnel packet */
1392 skb_queue_walk(queue, skb) {
1394 if (queue == &l->backlogq)
1395 msg_set_seqno(hdr, seqno++);
1396 pktlen = msg_size(hdr);
1397 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1398 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE);
1400 pr_warn("%sunable to send packet\n", link_co_err);
1403 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1404 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1405 __skb_queue_tail(&tnlq, tnlskb);
1407 if (queue != &l->backlogq) {
1408 queue = &l->backlogq;
1412 tipc_link_xmit(tnl, &tnlq, xmitq);
1414 if (mtyp == FAILOVER_MSG) {
1415 tnl->drop_point = l->rcv_nxt;
1416 tnl->failover_reasm_skb = l->reasm_buf;
1417 l->reasm_buf = NULL;
1421 /* tipc_link_proto_rcv(): receive link level protocol message :
1422 * Note that network plane id propagates through the network, and may
1423 * change at any time. The node with lowest numerical id determines
1426 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1427 struct sk_buff_head *xmitq)
1429 struct tipc_msg *hdr = buf_msg(skb);
1431 u16 ack = msg_ack(hdr);
1432 u16 gap = msg_seq_gap(hdr);
1433 u16 peers_snd_nxt = msg_next_sent(hdr);
1434 u16 peers_tol = msg_link_tolerance(hdr);
1435 u16 peers_prio = msg_linkprio(hdr);
1436 u16 rcv_nxt = l->rcv_nxt;
1437 int mtyp = msg_type(hdr);
1441 if (tipc_link_is_blocked(l) || !xmitq)
1444 if (link_own_addr(l) > msg_prevnode(hdr))
1445 l->net_plane = msg_net_plane(hdr);
1450 /* Ignore duplicate RESET with old session number */
1451 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1452 (l->peer_session != WILDCARD_SESSION))
1458 /* Complete own link name with peer's interface name */
1459 if_name = strrchr(l->name, ':') + 1;
1460 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1462 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1464 strncpy(if_name, msg_data(hdr), TIPC_MAX_IF_NAME);
1466 /* Update own tolerance if peer indicates a non-zero value */
1467 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1468 l->tolerance = peers_tol;
1470 /* Update own priority if peer's priority is higher */
1471 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1472 l->priority = peers_prio;
1474 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1475 if ((mtyp == RESET_MSG) || !link_is_up(l))
1476 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1478 /* ACTIVATE_MSG takes up link if it was already locally reset */
1479 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1480 rc = TIPC_LINK_UP_EVT;
1482 l->peer_session = msg_session(hdr);
1483 l->peer_bearer_id = msg_bearer_id(hdr);
1484 if (l->mtu > msg_max_pkt(hdr))
1485 l->mtu = msg_max_pkt(hdr);
1490 /* Update own tolerance if peer indicates a non-zero value */
1491 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1492 l->tolerance = peers_tol;
1494 l->silent_intv_cnt = 0;
1495 l->stats.recv_states++;
1497 l->stats.recv_probes++;
1499 if (!link_is_up(l)) {
1500 if (l->state == LINK_ESTABLISHING)
1501 rc = TIPC_LINK_UP_EVT;
1505 /* Send NACK if peer has sent pkts we haven't received yet */
1506 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1507 rcvgap = peers_snd_nxt - l->rcv_nxt;
1508 if (rcvgap || (msg_probe(hdr)))
1509 tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1511 tipc_link_release_pkts(l, ack);
1513 /* If NACK, retransmit will now start at right position */
1515 rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
1516 l->stats.recv_nacks++;
1519 tipc_link_advance_backlog(l, xmitq);
1520 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1521 link_prepare_wakeup(l);
1528 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1530 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1532 struct sk_buff_head *xmitq)
1534 struct sk_buff *skb;
1535 struct tipc_msg *hdr;
1536 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1537 u16 ack = l->rcv_nxt - 1;
1538 u16 gap_to = peers_snd_nxt - 1;
1540 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1541 0, l->addr, link_own_addr(l), 0, 0, 0);
1545 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1546 msg_set_bcast_ack(hdr, ack);
1547 msg_set_bcgap_after(hdr, ack);
1549 gap_to = buf_seqno(dfrd_skb) - 1;
1550 msg_set_bcgap_to(hdr, gap_to);
1551 msg_set_non_seq(hdr, bcast);
1552 __skb_queue_tail(xmitq, skb);
1556 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1558 * Give a newly added peer node the sequence number where it should
1559 * start receiving and acking broadcast packets.
1561 void tipc_link_build_bc_init_msg(struct tipc_link *l,
1562 struct sk_buff_head *xmitq)
1564 struct sk_buff_head list;
1566 __skb_queue_head_init(&list);
1567 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1569 tipc_link_xmit(l, &list, xmitq);
1572 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1574 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1576 int mtyp = msg_type(hdr);
1577 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1582 if (msg_user(hdr) == BCAST_PROTOCOL) {
1583 l->rcv_nxt = peers_snd_nxt;
1584 l->state = LINK_ESTABLISHED;
1588 if (l->peer_caps & TIPC_BCAST_SYNCH)
1591 if (msg_peer_node_is_up(hdr))
1594 /* Compatibility: accept older, less safe initial synch data */
1595 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1596 l->rcv_nxt = peers_snd_nxt;
1599 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1601 void tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1602 struct sk_buff_head *xmitq)
1604 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1609 if (!msg_peer_node_is_up(hdr))
1612 l->bc_peer_is_up = true;
1614 /* Ignore if peers_snd_nxt goes beyond receive window */
1615 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1618 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1619 l->nack_state = BC_NACK_SND_CONDITIONAL;
1623 /* Don't NACK if one was recently sent or peeked */
1624 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1625 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1629 /* Conditionally delay NACK sending until next synch rcv */
1630 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1631 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1632 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1636 /* Send NACK now but suppress next one */
1637 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1638 l->nack_state = BC_NACK_SND_SUPPRESS;
1641 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1642 struct sk_buff_head *xmitq)
1644 struct sk_buff *skb, *tmp;
1645 struct tipc_link *snd_l = l->bc_sndlink;
1647 if (!link_is_up(l) || !l->bc_peer_is_up)
1650 if (!more(acked, l->acked))
1653 /* Skip over packets peer has already acked */
1654 skb_queue_walk(&snd_l->transmq, skb) {
1655 if (more(buf_seqno(skb), l->acked))
1659 /* Update/release the packets peer is acking now */
1660 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1661 if (more(buf_seqno(skb), acked))
1663 if (!--TIPC_SKB_CB(skb)->ackers) {
1664 __skb_unlink(skb, &snd_l->transmq);
1669 tipc_link_advance_backlog(snd_l, xmitq);
1670 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1671 link_prepare_wakeup(snd_l);
1674 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1676 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1677 struct sk_buff_head *xmitq)
1679 struct tipc_msg *hdr = buf_msg(skb);
1680 u32 dnode = msg_destnode(hdr);
1681 int mtyp = msg_type(hdr);
1682 u16 acked = msg_bcast_ack(hdr);
1683 u16 from = acked + 1;
1684 u16 to = msg_bcgap_to(hdr);
1685 u16 peers_snd_nxt = to + 1;
1690 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1693 if (mtyp != STATE_MSG)
1696 if (dnode == link_own_addr(l)) {
1697 tipc_link_bc_ack_rcv(l, acked, xmitq);
1698 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1699 l->stats.recv_nacks++;
1703 /* Msg for other node => suppress own NACK at next sync if applicable */
1704 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1705 l->nack_state = BC_NACK_SND_SUPPRESS;
1710 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1712 int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1715 l->backlog[TIPC_LOW_IMPORTANCE].limit = win / 2;
1716 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = win;
1717 l->backlog[TIPC_HIGH_IMPORTANCE].limit = win / 2 * 3;
1718 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = win * 2;
1719 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1722 /* tipc_link_find_owner - locate owner node of link by link's name
1723 * @net: the applicable net namespace
1724 * @name: pointer to link name string
1725 * @bearer_id: pointer to index in 'node->links' array where the link was found.
1727 * Returns pointer to node owning the link, or 0 if no matching link is found.
1729 static struct tipc_node *tipc_link_find_owner(struct net *net,
1730 const char *link_name,
1731 unsigned int *bearer_id)
1733 struct tipc_net *tn = net_generic(net, tipc_net_id);
1734 struct tipc_link *l_ptr;
1735 struct tipc_node *n_ptr;
1736 struct tipc_node *found_node = NULL;
1741 list_for_each_entry_rcu(n_ptr, &tn->node_list, list) {
1742 tipc_node_lock(n_ptr);
1743 for (i = 0; i < MAX_BEARERS; i++) {
1744 l_ptr = n_ptr->links[i].link;
1745 if (l_ptr && !strcmp(l_ptr->name, link_name)) {
1751 tipc_node_unlock(n_ptr);
1761 * link_reset_statistics - reset link statistics
1762 * @l_ptr: pointer to link
1764 static void link_reset_statistics(struct tipc_link *l_ptr)
1766 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
1767 l_ptr->stats.sent_info = l_ptr->snd_nxt;
1768 l_ptr->stats.recv_info = l_ptr->rcv_nxt;
1771 static void link_print(struct tipc_link *l, const char *str)
1773 struct sk_buff *hskb = skb_peek(&l->transmq);
1774 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1775 u16 tail = l->snd_nxt - 1;
1777 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1778 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1779 skb_queue_len(&l->transmq), head, tail,
1780 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1783 /* Parse and validate nested (link) properties valid for media, bearer and link
1785 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1789 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1790 tipc_nl_prop_policy);
1794 if (props[TIPC_NLA_PROP_PRIO]) {
1797 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1798 if (prio > TIPC_MAX_LINK_PRI)
1802 if (props[TIPC_NLA_PROP_TOL]) {
1805 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1806 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1810 if (props[TIPC_NLA_PROP_WIN]) {
1813 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1814 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1821 int tipc_nl_link_set(struct sk_buff *skb, struct genl_info *info)
1827 struct tipc_link *link;
1828 struct tipc_node *node;
1829 struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
1830 struct net *net = sock_net(skb->sk);
1832 if (!info->attrs[TIPC_NLA_LINK])
1835 err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
1836 info->attrs[TIPC_NLA_LINK],
1837 tipc_nl_link_policy);
1841 if (!attrs[TIPC_NLA_LINK_NAME])
1844 name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
1846 if (strcmp(name, tipc_bclink_name) == 0)
1847 return tipc_nl_bc_link_set(net, attrs);
1849 node = tipc_link_find_owner(net, name, &bearer_id);
1853 tipc_node_lock(node);
1855 link = node->links[bearer_id].link;
1861 if (attrs[TIPC_NLA_LINK_PROP]) {
1862 struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
1864 err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP],
1871 if (props[TIPC_NLA_PROP_TOL]) {
1874 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1875 link->tolerance = tol;
1876 tipc_link_proto_xmit(link, STATE_MSG, 0, 0, tol, 0);
1878 if (props[TIPC_NLA_PROP_PRIO]) {
1881 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1882 link->priority = prio;
1883 tipc_link_proto_xmit(link, STATE_MSG, 0, 0, 0, prio);
1885 if (props[TIPC_NLA_PROP_WIN]) {
1888 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1889 tipc_link_set_queue_limits(link, win);
1894 tipc_node_unlock(node);
1899 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1902 struct nlattr *stats;
1909 struct nla_map map[] = {
1910 {TIPC_NLA_STATS_RX_INFO, s->recv_info},
1911 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1912 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1913 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1914 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1915 {TIPC_NLA_STATS_TX_INFO, s->sent_info},
1916 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1917 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1918 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1919 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1920 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1921 s->msg_length_counts : 1},
1922 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1923 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1924 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1925 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1926 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1927 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1928 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1929 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1930 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1931 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1932 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1933 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1934 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1935 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1936 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1937 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1938 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1939 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1940 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1941 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1942 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1943 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1944 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1947 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1951 for (i = 0; i < ARRAY_SIZE(map); i++)
1952 if (nla_put_u32(skb, map[i].key, map[i].val))
1955 nla_nest_end(skb, stats);
1959 nla_nest_cancel(skb, stats);
1964 /* Caller should hold appropriate locks to protect the link */
1965 static int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1966 struct tipc_link *link, int nlflags)
1970 struct nlattr *attrs;
1971 struct nlattr *prop;
1972 struct tipc_net *tn = net_generic(net, tipc_net_id);
1974 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1975 nlflags, TIPC_NL_LINK_GET);
1979 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1983 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1985 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1986 tipc_cluster_mask(tn->own_addr)))
1988 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1990 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->rcv_nxt))
1992 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->snd_nxt))
1995 if (tipc_link_is_up(link))
1996 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1998 if (tipc_link_is_active(link))
1999 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2002 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2005 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2007 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2009 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2012 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2014 nla_nest_end(msg->skb, prop);
2016 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2020 nla_nest_end(msg->skb, attrs);
2021 genlmsg_end(msg->skb, hdr);
2026 nla_nest_cancel(msg->skb, prop);
2028 nla_nest_cancel(msg->skb, attrs);
2030 genlmsg_cancel(msg->skb, hdr);
2035 /* Caller should hold node lock */
2036 static int __tipc_nl_add_node_links(struct net *net, struct tipc_nl_msg *msg,
2037 struct tipc_node *node, u32 *prev_link)
2042 for (i = *prev_link; i < MAX_BEARERS; i++) {
2045 if (!node->links[i].link)
2048 err = __tipc_nl_add_link(net, msg,
2049 node->links[i].link, NLM_F_MULTI);
2058 int tipc_nl_link_dump(struct sk_buff *skb, struct netlink_callback *cb)
2060 struct net *net = sock_net(skb->sk);
2061 struct tipc_net *tn = net_generic(net, tipc_net_id);
2062 struct tipc_node *node;
2063 struct tipc_nl_msg msg;
2064 u32 prev_node = cb->args[0];
2065 u32 prev_link = cb->args[1];
2066 int done = cb->args[2];
2073 msg.portid = NETLINK_CB(cb->skb).portid;
2074 msg.seq = cb->nlh->nlmsg_seq;
2078 node = tipc_node_find(net, prev_node);
2080 /* We never set seq or call nl_dump_check_consistent()
2081 * this means that setting prev_seq here will cause the
2082 * consistence check to fail in the netlink callback
2083 * handler. Resulting in the last NLMSG_DONE message
2084 * having the NLM_F_DUMP_INTR flag set.
2089 tipc_node_put(node);
2091 list_for_each_entry_continue_rcu(node, &tn->node_list,
2093 tipc_node_lock(node);
2094 err = __tipc_nl_add_node_links(net, &msg, node,
2096 tipc_node_unlock(node);
2100 prev_node = node->addr;
2103 err = tipc_nl_add_bc_link(net, &msg);
2107 list_for_each_entry_rcu(node, &tn->node_list, list) {
2108 tipc_node_lock(node);
2109 err = __tipc_nl_add_node_links(net, &msg, node,
2111 tipc_node_unlock(node);
2115 prev_node = node->addr;
2122 cb->args[0] = prev_node;
2123 cb->args[1] = prev_link;
2129 int tipc_nl_link_get(struct sk_buff *skb, struct genl_info *info)
2131 struct net *net = genl_info_net(info);
2132 struct tipc_nl_msg msg;
2136 msg.portid = info->snd_portid;
2137 msg.seq = info->snd_seq;
2139 if (!info->attrs[TIPC_NLA_LINK_NAME])
2141 name = nla_data(info->attrs[TIPC_NLA_LINK_NAME]);
2143 msg.skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
2147 if (strcmp(name, tipc_bclink_name) == 0) {
2148 err = tipc_nl_add_bc_link(net, &msg);
2150 nlmsg_free(msg.skb);
2155 struct tipc_node *node;
2156 struct tipc_link *link;
2158 node = tipc_link_find_owner(net, name, &bearer_id);
2162 tipc_node_lock(node);
2163 link = node->links[bearer_id].link;
2165 tipc_node_unlock(node);
2166 nlmsg_free(msg.skb);
2170 err = __tipc_nl_add_link(net, &msg, link, 0);
2171 tipc_node_unlock(node);
2173 nlmsg_free(msg.skb);
2178 return genlmsg_reply(msg.skb, info);
2181 int tipc_nl_link_reset_stats(struct sk_buff *skb, struct genl_info *info)
2185 unsigned int bearer_id;
2186 struct tipc_link *link;
2187 struct tipc_node *node;
2188 struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
2189 struct net *net = sock_net(skb->sk);
2191 if (!info->attrs[TIPC_NLA_LINK])
2194 err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
2195 info->attrs[TIPC_NLA_LINK],
2196 tipc_nl_link_policy);
2200 if (!attrs[TIPC_NLA_LINK_NAME])
2203 link_name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
2205 if (strcmp(link_name, tipc_bclink_name) == 0) {
2206 err = tipc_bclink_reset_stats(net);
2212 node = tipc_link_find_owner(net, link_name, &bearer_id);
2216 tipc_node_lock(node);
2218 link = node->links[bearer_id].link;
2220 tipc_node_unlock(node);
2224 link_reset_statistics(link);
2226 tipc_node_unlock(node);