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 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
220 int tipc_link_mtu(struct tipc_link *l)
225 static u32 link_own_addr(struct tipc_link *l)
227 return msg_prevnode(l->pmsg);
231 * tipc_link_create - create a new link
232 * @n: pointer to associated node
233 * @if_name: associated interface name
234 * @bearer_id: id (index) of associated bearer
235 * @tolerance: link tolerance to be used by link
236 * @net_plane: network plane (A,B,c..) this link belongs to
237 * @mtu: mtu to be advertised by link
238 * @priority: priority to be used by link
239 * @window: send window to be used by link
240 * @session: session to be used by link
241 * @ownnode: identity of own node
242 * @peer: node id of peer node
243 * @peer_caps: bitmap describing peer node capabilities
244 * @maddr: media address to be used
245 * @bc_sndlink: the namespace global link used for broadcast sending
246 * @bc_rcvlink: the peer specific link used for broadcast reception
247 * @inputq: queue to put messages ready for delivery
248 * @namedq: queue to put binding table update messages ready for delivery
249 * @link: return value, pointer to put the created link
251 * Returns true if link was created, otherwise false
253 bool tipc_link_create(struct tipc_node *n, char *if_name, int bearer_id,
254 int tolerance, char net_plane, u32 mtu, int priority,
255 int window, u32 session, u32 ownnode, u32 peer,
257 struct tipc_media_addr *maddr,
258 struct tipc_link *bc_sndlink,
259 struct tipc_link *bc_rcvlink,
260 struct sk_buff_head *inputq,
261 struct sk_buff_head *namedq,
262 struct tipc_link **link)
265 struct tipc_msg *hdr;
267 l = kzalloc(sizeof(*l), GFP_ATOMIC);
271 l->pmsg = (struct tipc_msg *)&l->proto_msg;
273 tipc_msg_init(ownnode, hdr, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, peer);
274 msg_set_size(hdr, sizeof(l->proto_msg));
275 msg_set_session(hdr, session);
276 msg_set_bearer_id(hdr, l->bearer_id);
278 /* Note: peer i/f name is completed by reset/activate message */
279 sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
280 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
281 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
282 strcpy((char *)msg_data(hdr), if_name);
285 l->peer_caps = peer_caps;
286 l->media_addr = maddr;
288 l->peer_session = WILDCARD_SESSION;
289 l->bearer_id = bearer_id;
290 l->tolerance = tolerance;
291 l->net_plane = net_plane;
292 l->advertised_mtu = mtu;
294 l->priority = priority;
295 tipc_link_set_queue_limits(l, window);
297 l->bc_sndlink = bc_sndlink;
298 l->bc_rcvlink = bc_rcvlink;
301 l->state = LINK_RESETTING;
302 __skb_queue_head_init(&l->transmq);
303 __skb_queue_head_init(&l->backlogq);
304 __skb_queue_head_init(&l->deferdq);
305 skb_queue_head_init(&l->wakeupq);
306 skb_queue_head_init(l->inputq);
311 * tipc_link_bc_create - create new link to be used for broadcast
312 * @n: pointer to associated node
313 * @mtu: mtu to be used
314 * @window: send window to be used
315 * @inputq: queue to put messages ready for delivery
316 * @namedq: queue to put binding table update messages ready for delivery
317 * @link: return value, pointer to put the created link
319 * Returns true if link was created, otherwise false
321 bool tipc_link_bc_create(struct tipc_node *n, u32 ownnode, u32 peer,
322 int mtu, int window, u16 peer_caps,
323 struct sk_buff_head *inputq,
324 struct sk_buff_head *namedq,
325 struct tipc_link *bc_sndlink,
326 struct tipc_link **link)
330 if (!tipc_link_create(n, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
331 0, ownnode, peer, peer_caps, NULL, bc_sndlink,
332 NULL, inputq, namedq, link))
336 strcpy(l->name, tipc_bclink_name);
338 l->state = LINK_RESET;
342 /* Broadcast send link is always up */
343 if (link_is_bc_sndlink(l))
344 l->state = LINK_ESTABLISHED;
350 * tipc_link_fsm_evt - link finite state machine
351 * @l: pointer to link
352 * @evt: state machine event to be processed
354 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
361 case LINK_PEER_RESET_EVT:
362 l->state = LINK_PEER_RESET;
365 l->state = LINK_RESET;
367 case LINK_FAILURE_EVT:
368 case LINK_FAILOVER_BEGIN_EVT:
369 case LINK_ESTABLISH_EVT:
370 case LINK_FAILOVER_END_EVT:
371 case LINK_SYNCH_BEGIN_EVT:
372 case LINK_SYNCH_END_EVT:
379 case LINK_PEER_RESET_EVT:
380 l->state = LINK_ESTABLISHING;
382 case LINK_FAILOVER_BEGIN_EVT:
383 l->state = LINK_FAILINGOVER;
384 case LINK_FAILURE_EVT:
386 case LINK_ESTABLISH_EVT:
387 case LINK_FAILOVER_END_EVT:
389 case LINK_SYNCH_BEGIN_EVT:
390 case LINK_SYNCH_END_EVT:
395 case LINK_PEER_RESET:
398 l->state = LINK_ESTABLISHING;
400 case LINK_PEER_RESET_EVT:
401 case LINK_ESTABLISH_EVT:
402 case LINK_FAILURE_EVT:
404 case LINK_SYNCH_BEGIN_EVT:
405 case LINK_SYNCH_END_EVT:
406 case LINK_FAILOVER_BEGIN_EVT:
407 case LINK_FAILOVER_END_EVT:
412 case LINK_FAILINGOVER:
414 case LINK_FAILOVER_END_EVT:
415 l->state = LINK_RESET;
417 case LINK_PEER_RESET_EVT:
419 case LINK_ESTABLISH_EVT:
420 case LINK_FAILURE_EVT:
422 case LINK_FAILOVER_BEGIN_EVT:
423 case LINK_SYNCH_BEGIN_EVT:
424 case LINK_SYNCH_END_EVT:
429 case LINK_ESTABLISHING:
431 case LINK_ESTABLISH_EVT:
432 l->state = LINK_ESTABLISHED;
434 case LINK_FAILOVER_BEGIN_EVT:
435 l->state = LINK_FAILINGOVER;
438 l->state = LINK_RESET;
440 case LINK_FAILURE_EVT:
441 case LINK_PEER_RESET_EVT:
442 case LINK_SYNCH_BEGIN_EVT:
443 case LINK_FAILOVER_END_EVT:
445 case LINK_SYNCH_END_EVT:
450 case LINK_ESTABLISHED:
452 case LINK_PEER_RESET_EVT:
453 l->state = LINK_PEER_RESET;
454 rc |= TIPC_LINK_DOWN_EVT;
456 case LINK_FAILURE_EVT:
457 l->state = LINK_RESETTING;
458 rc |= TIPC_LINK_DOWN_EVT;
461 l->state = LINK_RESET;
463 case LINK_ESTABLISH_EVT:
464 case LINK_SYNCH_END_EVT:
466 case LINK_SYNCH_BEGIN_EVT:
467 l->state = LINK_SYNCHING;
469 case LINK_FAILOVER_BEGIN_EVT:
470 case LINK_FAILOVER_END_EVT:
477 case LINK_PEER_RESET_EVT:
478 l->state = LINK_PEER_RESET;
479 rc |= TIPC_LINK_DOWN_EVT;
481 case LINK_FAILURE_EVT:
482 l->state = LINK_RESETTING;
483 rc |= TIPC_LINK_DOWN_EVT;
486 l->state = LINK_RESET;
488 case LINK_ESTABLISH_EVT:
489 case LINK_SYNCH_BEGIN_EVT:
491 case LINK_SYNCH_END_EVT:
492 l->state = LINK_ESTABLISHED;
494 case LINK_FAILOVER_BEGIN_EVT:
495 case LINK_FAILOVER_END_EVT:
501 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
505 pr_err("Illegal FSM event %x in state %x on link %s\n",
506 evt, l->state, l->name);
510 /* link_profile_stats - update statistical profiling of traffic
512 static void link_profile_stats(struct tipc_link *l)
515 struct tipc_msg *msg;
518 /* Update counters used in statistical profiling of send traffic */
519 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
520 l->stats.queue_sz_counts++;
522 skb = skb_peek(&l->transmq);
526 length = msg_size(msg);
528 if (msg_user(msg) == MSG_FRAGMENTER) {
529 if (msg_type(msg) != FIRST_FRAGMENT)
531 length = msg_size(msg_get_wrapped(msg));
533 l->stats.msg_lengths_total += length;
534 l->stats.msg_length_counts++;
536 l->stats.msg_length_profile[0]++;
537 else if (length <= 256)
538 l->stats.msg_length_profile[1]++;
539 else if (length <= 1024)
540 l->stats.msg_length_profile[2]++;
541 else if (length <= 4096)
542 l->stats.msg_length_profile[3]++;
543 else if (length <= 16384)
544 l->stats.msg_length_profile[4]++;
545 else if (length <= 32768)
546 l->stats.msg_length_profile[5]++;
548 l->stats.msg_length_profile[6]++;
551 /* tipc_link_timeout - perform periodic task as instructed from node timeout
553 /* tipc_link_timeout - perform periodic task as instructed from node timeout
555 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
558 int mtyp = STATE_MSG;
561 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
562 u16 bc_acked = l->bc_rcvlink->acked;
563 bool bc_up = link_is_up(l->bc_rcvlink);
565 link_profile_stats(l);
568 case LINK_ESTABLISHED:
570 if (!l->silent_intv_cnt) {
571 if (bc_up && (bc_acked != bc_snt))
573 } else if (l->silent_intv_cnt <= l->abort_limit) {
577 rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
579 l->silent_intv_cnt++;
585 case LINK_ESTABLISHING:
589 case LINK_PEER_RESET:
591 case LINK_FAILINGOVER:
598 tipc_link_build_proto_msg(l, mtyp, prb, 0, 0, 0, xmitq);
604 * link_schedule_user - schedule a message sender for wakeup after congestion
605 * @link: congested link
606 * @list: message that was attempted sent
607 * Create pseudo msg to send back to user when congestion abates
608 * Does not consume buffer list
610 static int link_schedule_user(struct tipc_link *link, struct sk_buff_head *list)
612 struct tipc_msg *msg = buf_msg(skb_peek(list));
613 int imp = msg_importance(msg);
614 u32 oport = msg_origport(msg);
615 u32 addr = link_own_addr(link);
618 /* This really cannot happen... */
619 if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
620 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
623 /* Non-blocking sender: */
624 if (TIPC_SKB_CB(skb_peek(list))->wakeup_pending)
627 /* Create and schedule wakeup pseudo message */
628 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
629 addr, addr, oport, 0, 0);
632 TIPC_SKB_CB(skb)->chain_sz = skb_queue_len(list);
633 TIPC_SKB_CB(skb)->chain_imp = imp;
634 skb_queue_tail(&link->wakeupq, skb);
635 link->stats.link_congs++;
640 * link_prepare_wakeup - prepare users for wakeup after congestion
641 * @link: congested link
642 * Move a number of waiting users, as permitted by available space in
643 * the send queue, from link wait queue to node wait queue for wakeup
645 void link_prepare_wakeup(struct tipc_link *l)
647 int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
649 struct sk_buff *skb, *tmp;
651 skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
652 imp = TIPC_SKB_CB(skb)->chain_imp;
653 lim = l->window + l->backlog[imp].limit;
654 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
655 if ((pnd[imp] + l->backlog[imp].len) >= lim)
657 skb_unlink(skb, &l->wakeupq);
658 skb_queue_tail(l->inputq, skb);
663 * tipc_link_reset_fragments - purge link's inbound message fragments queue
664 * @l_ptr: pointer to link
666 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
668 kfree_skb(l_ptr->reasm_buf);
669 l_ptr->reasm_buf = NULL;
672 void tipc_link_purge_backlog(struct tipc_link *l)
674 __skb_queue_purge(&l->backlogq);
675 l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
676 l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
677 l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
678 l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
679 l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
683 * tipc_link_purge_queues - purge all pkt queues associated with link
684 * @l_ptr: pointer to link
686 void tipc_link_purge_queues(struct tipc_link *l_ptr)
688 __skb_queue_purge(&l_ptr->deferdq);
689 __skb_queue_purge(&l_ptr->transmq);
690 tipc_link_purge_backlog(l_ptr);
691 tipc_link_reset_fragments(l_ptr);
694 void tipc_link_reset(struct tipc_link *l)
696 /* Link is down, accept any session */
697 l->peer_session = WILDCARD_SESSION;
699 /* If peer is up, it only accepts an incremented session number */
700 msg_set_session(l->pmsg, msg_session(l->pmsg) + 1);
702 /* Prepare for renewed mtu size negotiation */
703 l->mtu = l->advertised_mtu;
705 /* Clean up all queues: */
706 __skb_queue_purge(&l->transmq);
707 __skb_queue_purge(&l->deferdq);
708 skb_queue_splice_init(&l->wakeupq, l->inputq);
710 tipc_link_purge_backlog(l);
711 kfree_skb(l->reasm_buf);
712 kfree_skb(l->failover_reasm_skb);
714 l->failover_reasm_skb = NULL;
719 l->silent_intv_cnt = 0;
720 l->stats.recv_info = 0;
722 l->bc_peer_is_up = false;
723 link_reset_statistics(l);
727 * __tipc_link_xmit(): same as tipc_link_xmit, but destlink is known & locked
729 * @list: chain of buffers containing message
731 * Consumes the buffer chain, except when returning an error code,
732 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
733 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
735 int __tipc_link_xmit(struct net *net, struct tipc_link *link,
736 struct sk_buff_head *list)
738 struct tipc_msg *msg = buf_msg(skb_peek(list));
739 unsigned int maxwin = link->window;
740 unsigned int i, imp = msg_importance(msg);
741 uint mtu = link->mtu;
742 u16 ack = mod(link->rcv_nxt - 1);
743 u16 seqno = link->snd_nxt;
744 u16 bc_ack = link->bc_rcvlink->rcv_nxt - 1;
745 struct tipc_media_addr *addr = link->media_addr;
746 struct sk_buff_head *transmq = &link->transmq;
747 struct sk_buff_head *backlogq = &link->backlogq;
748 struct sk_buff *skb, *bskb;
750 /* Match msg importance against this and all higher backlog limits: */
751 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
752 if (unlikely(link->backlog[i].len >= link->backlog[i].limit))
753 return link_schedule_user(link, list);
755 if (unlikely(msg_size(msg) > mtu))
758 /* Prepare each packet for sending, and add to relevant queue: */
759 while (skb_queue_len(list)) {
760 skb = skb_peek(list);
762 msg_set_seqno(msg, seqno);
763 msg_set_ack(msg, ack);
764 msg_set_bcast_ack(msg, bc_ack);
766 if (likely(skb_queue_len(transmq) < maxwin)) {
768 __skb_queue_tail(transmq, skb);
769 tipc_bearer_send(net, link->bearer_id, skb, addr);
770 link->rcv_unacked = 0;
774 if (tipc_msg_bundle(skb_peek_tail(backlogq), msg, mtu)) {
775 kfree_skb(__skb_dequeue(list));
776 link->stats.sent_bundled++;
779 if (tipc_msg_make_bundle(&bskb, msg, mtu, link->addr)) {
780 kfree_skb(__skb_dequeue(list));
781 __skb_queue_tail(backlogq, bskb);
782 link->backlog[msg_importance(buf_msg(bskb))].len++;
783 link->stats.sent_bundled++;
784 link->stats.sent_bundles++;
787 link->backlog[imp].len += skb_queue_len(list);
788 skb_queue_splice_tail_init(list, backlogq);
790 link->snd_nxt = seqno;
795 * tipc_link_xmit(): enqueue buffer list according to queue situation
797 * @list: chain of buffers containing message
798 * @xmitq: returned list of packets to be sent by caller
800 * Consumes the buffer chain, except when returning -ELINKCONG,
801 * since the caller then may want to make more send attempts.
802 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
803 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
805 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
806 struct sk_buff_head *xmitq)
808 struct tipc_msg *hdr = buf_msg(skb_peek(list));
809 unsigned int maxwin = l->window;
810 unsigned int i, imp = msg_importance(hdr);
811 unsigned int mtu = l->mtu;
812 u16 ack = l->rcv_nxt - 1;
813 u16 seqno = l->snd_nxt;
814 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
815 struct sk_buff_head *transmq = &l->transmq;
816 struct sk_buff_head *backlogq = &l->backlogq;
817 struct sk_buff *skb, *_skb, *bskb;
819 /* Match msg importance against this and all higher backlog limits: */
820 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
821 if (unlikely(l->backlog[i].len >= l->backlog[i].limit))
822 return link_schedule_user(l, list);
824 if (unlikely(msg_size(hdr) > mtu))
827 /* Prepare each packet for sending, and add to relevant queue: */
828 while (skb_queue_len(list)) {
829 skb = skb_peek(list);
831 msg_set_seqno(hdr, seqno);
832 msg_set_ack(hdr, ack);
833 msg_set_bcast_ack(hdr, bc_ack);
835 if (likely(skb_queue_len(transmq) < maxwin)) {
836 _skb = skb_clone(skb, GFP_ATOMIC);
840 __skb_queue_tail(transmq, skb);
841 __skb_queue_tail(xmitq, _skb);
842 TIPC_SKB_CB(skb)->ackers = l->ackers;
847 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
848 kfree_skb(__skb_dequeue(list));
849 l->stats.sent_bundled++;
852 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
853 kfree_skb(__skb_dequeue(list));
854 __skb_queue_tail(backlogq, bskb);
855 l->backlog[msg_importance(buf_msg(bskb))].len++;
856 l->stats.sent_bundled++;
857 l->stats.sent_bundles++;
860 l->backlog[imp].len += skb_queue_len(list);
861 skb_queue_splice_tail_init(list, backlogq);
868 * tipc_link_push_packets - push unsent packets to bearer
870 * Push out the unsent messages of a link where congestion
871 * has abated. Node is locked.
873 * Called with node locked
875 void tipc_link_push_packets(struct tipc_link *link)
878 struct tipc_msg *msg;
879 u16 seqno = link->snd_nxt;
880 u16 ack = mod(link->rcv_nxt - 1);
882 while (skb_queue_len(&link->transmq) < link->window) {
883 skb = __skb_dequeue(&link->backlogq);
886 TIPC_SKB_CB(skb)->ackers = link->ackers;
888 link->backlog[msg_importance(msg)].len--;
889 msg_set_ack(msg, ack);
890 msg_set_seqno(msg, seqno);
891 seqno = mod(seqno + 1);
892 msg_set_bcast_ack(msg, link->owner->bclink.last_in);
893 link->rcv_unacked = 0;
894 __skb_queue_tail(&link->transmq, skb);
895 tipc_bearer_send(link->owner->net, link->bearer_id,
896 skb, link->media_addr);
898 link->snd_nxt = seqno;
901 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
903 struct sk_buff *skb, *_skb;
904 struct tipc_msg *hdr;
905 u16 seqno = l->snd_nxt;
906 u16 ack = l->rcv_nxt - 1;
907 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
909 while (skb_queue_len(&l->transmq) < l->window) {
910 skb = skb_peek(&l->backlogq);
913 _skb = skb_clone(skb, GFP_ATOMIC);
916 __skb_dequeue(&l->backlogq);
918 l->backlog[msg_importance(hdr)].len--;
919 __skb_queue_tail(&l->transmq, skb);
920 __skb_queue_tail(xmitq, _skb);
921 TIPC_SKB_CB(skb)->ackers = l->ackers;
922 msg_set_seqno(hdr, seqno);
923 msg_set_ack(hdr, ack);
924 msg_set_bcast_ack(hdr, bc_ack);
931 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
933 struct tipc_msg *hdr = buf_msg(skb);
935 pr_warn("Retransmission failure on link <%s>\n", l->name);
936 link_print(l, "Resetting link ");
937 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
938 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
939 pr_info("sqno %u, prev: %x, src: %x\n",
940 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
943 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *skb,
946 struct tipc_msg *msg;
953 /* Detect repeated retransmit failures */
954 if (l_ptr->last_retransm == msg_seqno(msg)) {
955 if (++l_ptr->stale_count > 100) {
956 link_retransmit_failure(l_ptr, skb);
960 l_ptr->last_retransm = msg_seqno(msg);
961 l_ptr->stale_count = 1;
964 skb_queue_walk_from(&l_ptr->transmq, skb) {
968 msg_set_ack(msg, mod(l_ptr->rcv_nxt - 1));
969 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
970 tipc_bearer_send(l_ptr->owner->net, l_ptr->bearer_id, skb,
973 l_ptr->stats.retransmitted++;
977 int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
978 struct sk_buff_head *xmitq)
980 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
981 struct tipc_msg *hdr;
982 u16 ack = l->rcv_nxt - 1;
983 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
988 /* Detect repeated retransmit failures on same packet */
989 if (likely(l->last_retransm != buf_seqno(skb))) {
990 l->last_retransm = buf_seqno(skb);
992 } else if (++l->stale_count > 100) {
993 link_retransmit_failure(l, skb);
994 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
997 /* Move forward to where retransmission should start */
998 skb_queue_walk(&l->transmq, skb) {
999 if (!less(buf_seqno(skb), from))
1003 skb_queue_walk_from(&l->transmq, skb) {
1004 if (more(buf_seqno(skb), to))
1007 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1010 hdr = buf_msg(_skb);
1011 msg_set_ack(hdr, ack);
1012 msg_set_bcast_ack(hdr, bc_ack);
1013 _skb->priority = TC_PRIO_CONTROL;
1014 __skb_queue_tail(xmitq, _skb);
1015 l->stats.retransmitted++;
1020 /* tipc_data_input - deliver data and name distr msgs to upper layer
1022 * Consumes buffer if message is of right type
1023 * Node lock must be held
1025 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1026 struct sk_buff_head *inputq)
1028 switch (msg_user(buf_msg(skb))) {
1029 case TIPC_LOW_IMPORTANCE:
1030 case TIPC_MEDIUM_IMPORTANCE:
1031 case TIPC_HIGH_IMPORTANCE:
1032 case TIPC_CRITICAL_IMPORTANCE:
1034 skb_queue_tail(inputq, skb);
1036 case NAME_DISTRIBUTOR:
1037 l->bc_rcvlink->state = LINK_ESTABLISHED;
1038 skb_queue_tail(l->namedq, skb);
1041 case TUNNEL_PROTOCOL:
1042 case MSG_FRAGMENTER:
1043 case BCAST_PROTOCOL:
1046 pr_warn("Dropping received illegal msg type\n");
1052 /* tipc_link_input - process packet that has passed link protocol check
1056 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1057 struct sk_buff_head *inputq)
1059 struct tipc_msg *hdr = buf_msg(skb);
1060 struct sk_buff **reasm_skb = &l->reasm_buf;
1061 struct sk_buff *iskb;
1062 struct sk_buff_head tmpq;
1063 int usr = msg_user(hdr);
1068 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1069 if (msg_type(hdr) == SYNCH_MSG) {
1070 __skb_queue_purge(&l->deferdq);
1073 if (!tipc_msg_extract(skb, &iskb, &ipos))
1078 if (less(msg_seqno(hdr), l->drop_point))
1080 if (tipc_data_input(l, skb, inputq))
1082 usr = msg_user(hdr);
1083 reasm_skb = &l->failover_reasm_skb;
1086 if (usr == MSG_BUNDLER) {
1087 skb_queue_head_init(&tmpq);
1088 l->stats.recv_bundles++;
1089 l->stats.recv_bundled += msg_msgcnt(hdr);
1090 while (tipc_msg_extract(skb, &iskb, &pos))
1091 tipc_data_input(l, iskb, &tmpq);
1092 tipc_skb_queue_splice_tail(&tmpq, inputq);
1094 } else if (usr == MSG_FRAGMENTER) {
1095 l->stats.recv_fragments++;
1096 if (tipc_buf_append(reasm_skb, &skb)) {
1097 l->stats.recv_fragmented++;
1098 tipc_data_input(l, skb, inputq);
1099 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1100 pr_warn_ratelimited("Unable to build fragment list\n");
1101 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1104 } else if (usr == BCAST_PROTOCOL) {
1105 tipc_bcast_lock(l->owner->net);
1106 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1107 tipc_bcast_unlock(l->owner->net);
1114 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1116 bool released = false;
1117 struct sk_buff *skb, *tmp;
1119 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1120 if (more(buf_seqno(skb), acked))
1122 __skb_unlink(skb, &l->transmq);
1129 /* tipc_link_build_ack_msg: prepare link acknowledge message for transmission
1131 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1132 * risk of ack storms towards the sender
1134 int tipc_link_build_ack_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1139 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1140 if (link_is_bc_rcvlink(l)) {
1141 if (((l->rcv_nxt ^ link_own_addr(l)) & 0xf) != 0xf)
1144 return TIPC_LINK_SND_BC_ACK;
1149 l->stats.sent_acks++;
1150 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1154 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1156 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1158 int mtyp = RESET_MSG;
1160 if (l->state == LINK_ESTABLISHING)
1161 mtyp = ACTIVATE_MSG;
1163 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1166 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1168 static void tipc_link_build_nack_msg(struct tipc_link *l,
1169 struct sk_buff_head *xmitq)
1171 u32 def_cnt = ++l->stats.deferred_recv;
1173 if (link_is_bc_rcvlink(l))
1176 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1177 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1180 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1181 * @l: the link that should handle the message
1183 * @xmitq: queue to place packets to be sent after this call
1185 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1186 struct sk_buff_head *xmitq)
1188 struct sk_buff_head *defq = &l->deferdq;
1189 struct tipc_msg *hdr;
1190 u16 seqno, rcv_nxt, win_lim;
1195 seqno = msg_seqno(hdr);
1196 rcv_nxt = l->rcv_nxt;
1197 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1199 /* Verify and update link state */
1200 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1201 return tipc_link_proto_rcv(l, skb, xmitq);
1203 if (unlikely(!link_is_up(l))) {
1204 if (l->state == LINK_ESTABLISHING)
1205 rc = TIPC_LINK_UP_EVT;
1209 /* Don't send probe at next timeout expiration */
1210 l->silent_intv_cnt = 0;
1212 /* Drop if outside receive window */
1213 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1214 l->stats.duplicates++;
1218 /* Forward queues and wake up waiting users */
1219 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1220 tipc_link_advance_backlog(l, xmitq);
1221 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1222 link_prepare_wakeup(l);
1225 /* Defer delivery if sequence gap */
1226 if (unlikely(seqno != rcv_nxt)) {
1227 __tipc_skb_queue_sorted(defq, seqno, skb);
1228 tipc_link_build_nack_msg(l, xmitq);
1232 /* Deliver packet */
1234 l->stats.recv_info++;
1235 if (!tipc_data_input(l, skb, l->inputq))
1236 rc |= tipc_link_input(l, skb, l->inputq);
1237 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1238 rc |= tipc_link_build_ack_msg(l, xmitq);
1239 if (unlikely(rc & ~TIPC_LINK_SND_BC_ACK))
1241 } while ((skb = __skb_dequeue(defq)));
1250 * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1252 * Returns increase in queue length (i.e. 0 or 1)
1254 u32 tipc_link_defer_pkt(struct sk_buff_head *list, struct sk_buff *skb)
1256 struct sk_buff *skb1;
1257 u16 seq_no = buf_seqno(skb);
1260 if (skb_queue_empty(list)) {
1261 __skb_queue_tail(list, skb);
1266 if (less(buf_seqno(skb_peek_tail(list)), seq_no)) {
1267 __skb_queue_tail(list, skb);
1271 /* Locate insertion point in queue, then insert; discard if duplicate */
1272 skb_queue_walk(list, skb1) {
1273 u16 curr_seqno = buf_seqno(skb1);
1275 if (seq_no == curr_seqno) {
1280 if (less(seq_no, curr_seqno))
1284 __skb_queue_before(list, skb1, skb);
1289 * Send protocol message to the other endpoint.
1291 void tipc_link_proto_xmit(struct tipc_link *l, u32 msg_typ, int probe_msg,
1292 u32 gap, u32 tolerance, u32 priority)
1294 struct sk_buff *skb = NULL;
1295 struct sk_buff_head xmitq;
1297 __skb_queue_head_init(&xmitq);
1298 tipc_link_build_proto_msg(l, msg_typ, probe_msg, gap,
1299 tolerance, priority, &xmitq);
1300 skb = __skb_dequeue(&xmitq);
1303 tipc_bearer_xmit_skb(l->owner->net, l->bearer_id, skb, l->media_addr);
1307 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1308 u16 rcvgap, int tolerance, int priority,
1309 struct sk_buff_head *xmitq)
1311 struct sk_buff *skb = NULL;
1312 struct tipc_msg *hdr = l->pmsg;
1313 bool node_up = link_is_up(l->bc_rcvlink);
1315 /* Don't send protocol message during reset or link failover */
1316 if (tipc_link_is_blocked(l))
1319 msg_set_type(hdr, mtyp);
1320 msg_set_net_plane(hdr, l->net_plane);
1321 msg_set_next_sent(hdr, l->snd_nxt);
1322 msg_set_ack(hdr, l->rcv_nxt - 1);
1323 msg_set_bcast_ack(hdr, l->bc_rcvlink->rcv_nxt - 1);
1324 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1325 msg_set_link_tolerance(hdr, tolerance);
1326 msg_set_linkprio(hdr, priority);
1327 msg_set_redundant_link(hdr, node_up);
1328 msg_set_seq_gap(hdr, 0);
1330 /* Compatibility: created msg must not be in sequence with pkt flow */
1331 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1333 if (mtyp == STATE_MSG) {
1334 if (!tipc_link_is_up(l))
1337 /* Override rcvgap if there are packets in deferred queue */
1338 if (!skb_queue_empty(&l->deferdq))
1339 rcvgap = buf_seqno(skb_peek(&l->deferdq)) - l->rcv_nxt;
1341 msg_set_seq_gap(hdr, rcvgap);
1342 l->stats.sent_nacks++;
1344 msg_set_probe(hdr, probe);
1346 l->stats.sent_probes++;
1347 l->stats.sent_states++;
1350 /* RESET_MSG or ACTIVATE_MSG */
1351 msg_set_max_pkt(hdr, l->advertised_mtu);
1352 msg_set_ack(hdr, l->rcv_nxt - 1);
1353 msg_set_next_sent(hdr, 1);
1355 skb = tipc_buf_acquire(msg_size(hdr));
1358 skb_copy_to_linear_data(skb, hdr, msg_size(hdr));
1359 skb->priority = TC_PRIO_CONTROL;
1360 __skb_queue_tail(xmitq, skb);
1363 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1364 * with contents of the link's transmit and backlog queues.
1366 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1367 int mtyp, struct sk_buff_head *xmitq)
1369 struct sk_buff *skb, *tnlskb;
1370 struct tipc_msg *hdr, tnlhdr;
1371 struct sk_buff_head *queue = &l->transmq;
1372 struct sk_buff_head tmpxq, tnlq;
1373 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1378 skb_queue_head_init(&tnlq);
1379 skb_queue_head_init(&tmpxq);
1381 /* At least one packet required for safe algorithm => add dummy */
1382 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1383 BASIC_H_SIZE, 0, l->addr, link_own_addr(l),
1384 0, 0, TIPC_ERR_NO_PORT);
1386 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1389 skb_queue_tail(&tnlq, skb);
1390 tipc_link_xmit(l, &tnlq, &tmpxq);
1391 __skb_queue_purge(&tmpxq);
1393 /* Initialize reusable tunnel packet header */
1394 tipc_msg_init(link_own_addr(l), &tnlhdr, TUNNEL_PROTOCOL,
1395 mtyp, INT_H_SIZE, l->addr);
1396 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1397 msg_set_msgcnt(&tnlhdr, pktcnt);
1398 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1400 /* Wrap each packet into a tunnel packet */
1401 skb_queue_walk(queue, skb) {
1403 if (queue == &l->backlogq)
1404 msg_set_seqno(hdr, seqno++);
1405 pktlen = msg_size(hdr);
1406 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1407 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE);
1409 pr_warn("%sunable to send packet\n", link_co_err);
1412 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1413 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1414 __skb_queue_tail(&tnlq, tnlskb);
1416 if (queue != &l->backlogq) {
1417 queue = &l->backlogq;
1421 tipc_link_xmit(tnl, &tnlq, xmitq);
1423 if (mtyp == FAILOVER_MSG) {
1424 tnl->drop_point = l->rcv_nxt;
1425 tnl->failover_reasm_skb = l->reasm_buf;
1426 l->reasm_buf = NULL;
1430 /* tipc_link_proto_rcv(): receive link level protocol message :
1431 * Note that network plane id propagates through the network, and may
1432 * change at any time. The node with lowest numerical id determines
1435 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1436 struct sk_buff_head *xmitq)
1438 struct tipc_msg *hdr = buf_msg(skb);
1440 u16 ack = msg_ack(hdr);
1441 u16 gap = msg_seq_gap(hdr);
1442 u16 peers_snd_nxt = msg_next_sent(hdr);
1443 u16 peers_tol = msg_link_tolerance(hdr);
1444 u16 peers_prio = msg_linkprio(hdr);
1445 u16 rcv_nxt = l->rcv_nxt;
1446 int mtyp = msg_type(hdr);
1450 if (tipc_link_is_blocked(l) || !xmitq)
1453 if (link_own_addr(l) > msg_prevnode(hdr))
1454 l->net_plane = msg_net_plane(hdr);
1459 /* Ignore duplicate RESET with old session number */
1460 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1461 (l->peer_session != WILDCARD_SESSION))
1467 /* Complete own link name with peer's interface name */
1468 if_name = strrchr(l->name, ':') + 1;
1469 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1471 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1473 strncpy(if_name, msg_data(hdr), TIPC_MAX_IF_NAME);
1475 /* Update own tolerance if peer indicates a non-zero value */
1476 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1477 l->tolerance = peers_tol;
1479 /* Update own priority if peer's priority is higher */
1480 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1481 l->priority = peers_prio;
1483 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1484 if ((mtyp == RESET_MSG) || !link_is_up(l))
1485 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1487 /* ACTIVATE_MSG takes up link if it was already locally reset */
1488 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1489 rc = TIPC_LINK_UP_EVT;
1491 l->peer_session = msg_session(hdr);
1492 l->peer_bearer_id = msg_bearer_id(hdr);
1493 if (l->mtu > msg_max_pkt(hdr))
1494 l->mtu = msg_max_pkt(hdr);
1499 /* Update own tolerance if peer indicates a non-zero value */
1500 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1501 l->tolerance = peers_tol;
1503 l->silent_intv_cnt = 0;
1504 l->stats.recv_states++;
1506 l->stats.recv_probes++;
1508 if (!link_is_up(l)) {
1509 if (l->state == LINK_ESTABLISHING)
1510 rc = TIPC_LINK_UP_EVT;
1514 /* Send NACK if peer has sent pkts we haven't received yet */
1515 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1516 rcvgap = peers_snd_nxt - l->rcv_nxt;
1517 if (rcvgap || (msg_probe(hdr)))
1518 tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1520 tipc_link_release_pkts(l, ack);
1522 /* If NACK, retransmit will now start at right position */
1524 rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
1525 l->stats.recv_nacks++;
1528 tipc_link_advance_backlog(l, xmitq);
1529 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1530 link_prepare_wakeup(l);
1537 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1539 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1541 struct sk_buff_head *xmitq)
1543 struct sk_buff *skb;
1544 struct tipc_msg *hdr;
1545 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1546 u16 ack = l->rcv_nxt - 1;
1547 u16 gap_to = peers_snd_nxt - 1;
1549 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1550 0, l->addr, link_own_addr(l), 0, 0, 0);
1554 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1555 msg_set_bcast_ack(hdr, ack);
1556 msg_set_bcgap_after(hdr, ack);
1558 gap_to = buf_seqno(dfrd_skb) - 1;
1559 msg_set_bcgap_to(hdr, gap_to);
1560 msg_set_non_seq(hdr, bcast);
1561 __skb_queue_tail(xmitq, skb);
1565 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1567 * Give a newly added peer node the sequence number where it should
1568 * start receiving and acking broadcast packets.
1570 void tipc_link_build_bc_init_msg(struct tipc_link *l,
1571 struct sk_buff_head *xmitq)
1573 struct sk_buff_head list;
1575 __skb_queue_head_init(&list);
1576 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1578 tipc_link_xmit(l, &list, xmitq);
1581 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1583 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1585 int mtyp = msg_type(hdr);
1586 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1591 if (msg_user(hdr) == BCAST_PROTOCOL) {
1592 l->rcv_nxt = peers_snd_nxt;
1593 l->state = LINK_ESTABLISHED;
1597 if (l->peer_caps & TIPC_BCAST_SYNCH)
1600 if (msg_peer_node_is_up(hdr))
1603 /* Compatibility: accept older, less safe initial synch data */
1604 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1605 l->rcv_nxt = peers_snd_nxt;
1608 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1610 void tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1611 struct sk_buff_head *xmitq)
1613 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1618 if (!msg_peer_node_is_up(hdr))
1621 l->bc_peer_is_up = true;
1623 /* Ignore if peers_snd_nxt goes beyond receive window */
1624 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1627 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1628 l->nack_state = BC_NACK_SND_CONDITIONAL;
1632 /* Don't NACK if one was recently sent or peeked */
1633 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1634 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1638 /* Conditionally delay NACK sending until next synch rcv */
1639 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1640 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1641 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1645 /* Send NACK now but suppress next one */
1646 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1647 l->nack_state = BC_NACK_SND_SUPPRESS;
1650 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1651 struct sk_buff_head *xmitq)
1653 struct sk_buff *skb, *tmp;
1654 struct tipc_link *snd_l = l->bc_sndlink;
1656 if (!link_is_up(l) || !l->bc_peer_is_up)
1659 if (!more(acked, l->acked))
1662 /* Skip over packets peer has already acked */
1663 skb_queue_walk(&snd_l->transmq, skb) {
1664 if (more(buf_seqno(skb), l->acked))
1668 /* Update/release the packets peer is acking now */
1669 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1670 if (more(buf_seqno(skb), acked))
1672 if (!--TIPC_SKB_CB(skb)->ackers) {
1673 __skb_unlink(skb, &snd_l->transmq);
1678 tipc_link_advance_backlog(snd_l, xmitq);
1679 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1680 link_prepare_wakeup(snd_l);
1683 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1685 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1686 struct sk_buff_head *xmitq)
1688 struct tipc_msg *hdr = buf_msg(skb);
1689 u32 dnode = msg_destnode(hdr);
1690 int mtyp = msg_type(hdr);
1691 u16 acked = msg_bcast_ack(hdr);
1692 u16 from = acked + 1;
1693 u16 to = msg_bcgap_to(hdr);
1694 u16 peers_snd_nxt = to + 1;
1699 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1702 if (mtyp != STATE_MSG)
1705 if (dnode == link_own_addr(l)) {
1706 tipc_link_bc_ack_rcv(l, acked, xmitq);
1707 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1708 l->stats.recv_nacks++;
1712 /* Msg for other node => suppress own NACK at next sync if applicable */
1713 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1714 l->nack_state = BC_NACK_SND_SUPPRESS;
1719 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1721 int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1724 l->backlog[TIPC_LOW_IMPORTANCE].limit = win / 2;
1725 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = win;
1726 l->backlog[TIPC_HIGH_IMPORTANCE].limit = win / 2 * 3;
1727 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = win * 2;
1728 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1731 /* tipc_link_find_owner - locate owner node of link by link's name
1732 * @net: the applicable net namespace
1733 * @name: pointer to link name string
1734 * @bearer_id: pointer to index in 'node->links' array where the link was found.
1736 * Returns pointer to node owning the link, or 0 if no matching link is found.
1738 static struct tipc_node *tipc_link_find_owner(struct net *net,
1739 const char *link_name,
1740 unsigned int *bearer_id)
1742 struct tipc_net *tn = net_generic(net, tipc_net_id);
1743 struct tipc_link *l_ptr;
1744 struct tipc_node *n_ptr;
1745 struct tipc_node *found_node = NULL;
1750 list_for_each_entry_rcu(n_ptr, &tn->node_list, list) {
1751 tipc_node_lock(n_ptr);
1752 for (i = 0; i < MAX_BEARERS; i++) {
1753 l_ptr = n_ptr->links[i].link;
1754 if (l_ptr && !strcmp(l_ptr->name, link_name)) {
1760 tipc_node_unlock(n_ptr);
1770 * link_reset_statistics - reset link statistics
1771 * @l_ptr: pointer to link
1773 static void link_reset_statistics(struct tipc_link *l_ptr)
1775 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
1776 l_ptr->stats.sent_info = l_ptr->snd_nxt;
1777 l_ptr->stats.recv_info = l_ptr->rcv_nxt;
1780 static void link_print(struct tipc_link *l, const char *str)
1782 struct sk_buff *hskb = skb_peek(&l->transmq);
1783 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1784 u16 tail = l->snd_nxt - 1;
1786 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1787 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1788 skb_queue_len(&l->transmq), head, tail,
1789 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1792 /* Parse and validate nested (link) properties valid for media, bearer and link
1794 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1798 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1799 tipc_nl_prop_policy);
1803 if (props[TIPC_NLA_PROP_PRIO]) {
1806 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1807 if (prio > TIPC_MAX_LINK_PRI)
1811 if (props[TIPC_NLA_PROP_TOL]) {
1814 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1815 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1819 if (props[TIPC_NLA_PROP_WIN]) {
1822 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1823 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1830 int tipc_nl_link_set(struct sk_buff *skb, struct genl_info *info)
1836 struct tipc_link *link;
1837 struct tipc_node *node;
1838 struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
1839 struct net *net = sock_net(skb->sk);
1841 if (!info->attrs[TIPC_NLA_LINK])
1844 err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
1845 info->attrs[TIPC_NLA_LINK],
1846 tipc_nl_link_policy);
1850 if (!attrs[TIPC_NLA_LINK_NAME])
1853 name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
1855 if (strcmp(name, tipc_bclink_name) == 0)
1856 return tipc_nl_bc_link_set(net, attrs);
1858 node = tipc_link_find_owner(net, name, &bearer_id);
1862 tipc_node_lock(node);
1864 link = node->links[bearer_id].link;
1870 if (attrs[TIPC_NLA_LINK_PROP]) {
1871 struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
1873 err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP],
1880 if (props[TIPC_NLA_PROP_TOL]) {
1883 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1884 link->tolerance = tol;
1885 tipc_link_proto_xmit(link, STATE_MSG, 0, 0, tol, 0);
1887 if (props[TIPC_NLA_PROP_PRIO]) {
1890 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1891 link->priority = prio;
1892 tipc_link_proto_xmit(link, STATE_MSG, 0, 0, 0, prio);
1894 if (props[TIPC_NLA_PROP_WIN]) {
1897 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1898 tipc_link_set_queue_limits(link, win);
1903 tipc_node_unlock(node);
1908 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1911 struct nlattr *stats;
1918 struct nla_map map[] = {
1919 {TIPC_NLA_STATS_RX_INFO, s->recv_info},
1920 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1921 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1922 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1923 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1924 {TIPC_NLA_STATS_TX_INFO, s->sent_info},
1925 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1926 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1927 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1928 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1929 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1930 s->msg_length_counts : 1},
1931 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1932 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1933 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1934 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1935 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1936 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1937 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1938 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1939 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1940 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1941 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1942 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1943 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1944 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1945 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1946 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1947 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1948 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1949 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1950 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1951 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1952 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1953 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1956 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1960 for (i = 0; i < ARRAY_SIZE(map); i++)
1961 if (nla_put_u32(skb, map[i].key, map[i].val))
1964 nla_nest_end(skb, stats);
1968 nla_nest_cancel(skb, stats);
1973 /* Caller should hold appropriate locks to protect the link */
1974 static int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1975 struct tipc_link *link, int nlflags)
1979 struct nlattr *attrs;
1980 struct nlattr *prop;
1981 struct tipc_net *tn = net_generic(net, tipc_net_id);
1983 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1984 nlflags, TIPC_NL_LINK_GET);
1988 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1992 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1994 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1995 tipc_cluster_mask(tn->own_addr)))
1997 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1999 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->rcv_nxt))
2001 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->snd_nxt))
2004 if (tipc_link_is_up(link))
2005 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2007 if (tipc_link_is_active(link))
2008 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2011 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2014 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2016 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2018 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2021 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2023 nla_nest_end(msg->skb, prop);
2025 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2029 nla_nest_end(msg->skb, attrs);
2030 genlmsg_end(msg->skb, hdr);
2035 nla_nest_cancel(msg->skb, prop);
2037 nla_nest_cancel(msg->skb, attrs);
2039 genlmsg_cancel(msg->skb, hdr);
2044 /* Caller should hold node lock */
2045 static int __tipc_nl_add_node_links(struct net *net, struct tipc_nl_msg *msg,
2046 struct tipc_node *node, u32 *prev_link)
2051 for (i = *prev_link; i < MAX_BEARERS; i++) {
2054 if (!node->links[i].link)
2057 err = __tipc_nl_add_link(net, msg,
2058 node->links[i].link, NLM_F_MULTI);
2067 int tipc_nl_link_dump(struct sk_buff *skb, struct netlink_callback *cb)
2069 struct net *net = sock_net(skb->sk);
2070 struct tipc_net *tn = net_generic(net, tipc_net_id);
2071 struct tipc_node *node;
2072 struct tipc_nl_msg msg;
2073 u32 prev_node = cb->args[0];
2074 u32 prev_link = cb->args[1];
2075 int done = cb->args[2];
2082 msg.portid = NETLINK_CB(cb->skb).portid;
2083 msg.seq = cb->nlh->nlmsg_seq;
2087 node = tipc_node_find(net, prev_node);
2089 /* We never set seq or call nl_dump_check_consistent()
2090 * this means that setting prev_seq here will cause the
2091 * consistence check to fail in the netlink callback
2092 * handler. Resulting in the last NLMSG_DONE message
2093 * having the NLM_F_DUMP_INTR flag set.
2098 tipc_node_put(node);
2100 list_for_each_entry_continue_rcu(node, &tn->node_list,
2102 tipc_node_lock(node);
2103 err = __tipc_nl_add_node_links(net, &msg, node,
2105 tipc_node_unlock(node);
2109 prev_node = node->addr;
2112 err = tipc_nl_add_bc_link(net, &msg);
2116 list_for_each_entry_rcu(node, &tn->node_list, list) {
2117 tipc_node_lock(node);
2118 err = __tipc_nl_add_node_links(net, &msg, node,
2120 tipc_node_unlock(node);
2124 prev_node = node->addr;
2131 cb->args[0] = prev_node;
2132 cb->args[1] = prev_link;
2138 int tipc_nl_link_get(struct sk_buff *skb, struct genl_info *info)
2140 struct net *net = genl_info_net(info);
2141 struct tipc_nl_msg msg;
2145 msg.portid = info->snd_portid;
2146 msg.seq = info->snd_seq;
2148 if (!info->attrs[TIPC_NLA_LINK_NAME])
2150 name = nla_data(info->attrs[TIPC_NLA_LINK_NAME]);
2152 msg.skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
2156 if (strcmp(name, tipc_bclink_name) == 0) {
2157 err = tipc_nl_add_bc_link(net, &msg);
2159 nlmsg_free(msg.skb);
2164 struct tipc_node *node;
2165 struct tipc_link *link;
2167 node = tipc_link_find_owner(net, name, &bearer_id);
2171 tipc_node_lock(node);
2172 link = node->links[bearer_id].link;
2174 tipc_node_unlock(node);
2175 nlmsg_free(msg.skb);
2179 err = __tipc_nl_add_link(net, &msg, link, 0);
2180 tipc_node_unlock(node);
2182 nlmsg_free(msg.skb);
2187 return genlmsg_reply(msg.skb, info);
2190 int tipc_nl_link_reset_stats(struct sk_buff *skb, struct genl_info *info)
2194 unsigned int bearer_id;
2195 struct tipc_link *link;
2196 struct tipc_node *node;
2197 struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
2198 struct net *net = sock_net(skb->sk);
2200 if (!info->attrs[TIPC_NLA_LINK])
2203 err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
2204 info->attrs[TIPC_NLA_LINK],
2205 tipc_nl_link_policy);
2209 if (!attrs[TIPC_NLA_LINK_NAME])
2212 link_name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
2214 if (strcmp(link_name, tipc_bclink_name) == 0) {
2215 err = tipc_bclink_reset_stats(net);
2221 node = tipc_link_find_owner(net, link_name, &bearer_id);
2225 tipc_node_lock(node);
2227 link = node->links[bearer_id].link;
2229 tipc_node_unlock(node);
2233 link_reset_statistics(link);
2235 tipc_node_unlock(node);