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)
181 void tipc_link_set_active(struct tipc_link *l, bool active)
186 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
187 struct tipc_link *uc_l,
188 struct sk_buff_head *xmitq)
190 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
193 rcv_l->acked = snd_l->snd_nxt - 1;
194 tipc_link_build_bc_init_msg(uc_l, xmitq);
197 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
198 struct tipc_link *rcv_l,
199 struct sk_buff_head *xmitq)
201 u16 ack = snd_l->snd_nxt - 1;
204 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
205 tipc_link_reset(rcv_l);
206 rcv_l->state = LINK_RESET;
207 if (!snd_l->ackers) {
208 tipc_link_reset(snd_l);
209 __skb_queue_purge(xmitq);
213 int tipc_link_bc_peers(struct tipc_link *l)
218 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
223 int tipc_link_mtu(struct tipc_link *l)
228 static u32 link_own_addr(struct tipc_link *l)
230 return msg_prevnode(l->pmsg);
234 * tipc_link_create - create a new link
235 * @n: pointer to associated node
236 * @if_name: associated interface name
237 * @bearer_id: id (index) of associated bearer
238 * @tolerance: link tolerance to be used by link
239 * @net_plane: network plane (A,B,c..) this link belongs to
240 * @mtu: mtu to be advertised by link
241 * @priority: priority to be used by link
242 * @window: send window to be used by link
243 * @session: session to be used by link
244 * @ownnode: identity of own node
245 * @peer: node id of peer node
246 * @peer_caps: bitmap describing peer node capabilities
247 * @maddr: media address to be used
248 * @bc_sndlink: the namespace global link used for broadcast sending
249 * @bc_rcvlink: the peer specific link used for broadcast reception
250 * @inputq: queue to put messages ready for delivery
251 * @namedq: queue to put binding table update messages ready for delivery
252 * @link: return value, pointer to put the created link
254 * Returns true if link was created, otherwise false
256 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
257 int tolerance, char net_plane, u32 mtu, int priority,
258 int window, u32 session, u32 ownnode, u32 peer,
260 struct tipc_media_addr *maddr,
261 struct tipc_link *bc_sndlink,
262 struct tipc_link *bc_rcvlink,
263 struct sk_buff_head *inputq,
264 struct sk_buff_head *namedq,
265 struct tipc_link **link)
268 struct tipc_msg *hdr;
270 l = kzalloc(sizeof(*l), GFP_ATOMIC);
274 l->pmsg = (struct tipc_msg *)&l->proto_msg;
276 tipc_msg_init(ownnode, hdr, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, peer);
277 msg_set_size(hdr, sizeof(l->proto_msg));
278 msg_set_session(hdr, session);
279 msg_set_bearer_id(hdr, l->bearer_id);
281 /* Note: peer i/f name is completed by reset/activate message */
282 sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
283 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
284 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
285 strcpy((char *)msg_data(hdr), if_name);
288 l->peer_caps = peer_caps;
289 l->media_addr = maddr;
291 l->peer_session = WILDCARD_SESSION;
292 l->bearer_id = bearer_id;
293 l->tolerance = tolerance;
294 l->net_plane = net_plane;
295 l->advertised_mtu = mtu;
297 l->priority = priority;
298 tipc_link_set_queue_limits(l, window);
300 l->bc_sndlink = bc_sndlink;
301 l->bc_rcvlink = bc_rcvlink;
304 l->state = LINK_RESETTING;
305 __skb_queue_head_init(&l->transmq);
306 __skb_queue_head_init(&l->backlogq);
307 __skb_queue_head_init(&l->deferdq);
308 skb_queue_head_init(&l->wakeupq);
309 skb_queue_head_init(l->inputq);
314 * tipc_link_bc_create - create new link to be used for broadcast
315 * @n: pointer to associated node
316 * @mtu: mtu to be used
317 * @window: send window to be used
318 * @inputq: queue to put messages ready for delivery
319 * @namedq: queue to put binding table update messages ready for delivery
320 * @link: return value, pointer to put the created link
322 * Returns true if link was created, otherwise false
324 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
325 int mtu, int window, u16 peer_caps,
326 struct sk_buff_head *inputq,
327 struct sk_buff_head *namedq,
328 struct tipc_link *bc_sndlink,
329 struct tipc_link **link)
333 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
334 0, ownnode, peer, peer_caps, NULL, bc_sndlink,
335 NULL, inputq, namedq, link))
339 strcpy(l->name, tipc_bclink_name);
341 l->state = LINK_RESET;
345 /* Broadcast send link is always up */
346 if (link_is_bc_sndlink(l))
347 l->state = LINK_ESTABLISHED;
353 * tipc_link_fsm_evt - link finite state machine
354 * @l: pointer to link
355 * @evt: state machine event to be processed
357 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
364 case LINK_PEER_RESET_EVT:
365 l->state = LINK_PEER_RESET;
368 l->state = LINK_RESET;
370 case LINK_FAILURE_EVT:
371 case LINK_FAILOVER_BEGIN_EVT:
372 case LINK_ESTABLISH_EVT:
373 case LINK_FAILOVER_END_EVT:
374 case LINK_SYNCH_BEGIN_EVT:
375 case LINK_SYNCH_END_EVT:
382 case LINK_PEER_RESET_EVT:
383 l->state = LINK_ESTABLISHING;
385 case LINK_FAILOVER_BEGIN_EVT:
386 l->state = LINK_FAILINGOVER;
387 case LINK_FAILURE_EVT:
389 case LINK_ESTABLISH_EVT:
390 case LINK_FAILOVER_END_EVT:
392 case LINK_SYNCH_BEGIN_EVT:
393 case LINK_SYNCH_END_EVT:
398 case LINK_PEER_RESET:
401 l->state = LINK_ESTABLISHING;
403 case LINK_PEER_RESET_EVT:
404 case LINK_ESTABLISH_EVT:
405 case LINK_FAILURE_EVT:
407 case LINK_SYNCH_BEGIN_EVT:
408 case LINK_SYNCH_END_EVT:
409 case LINK_FAILOVER_BEGIN_EVT:
410 case LINK_FAILOVER_END_EVT:
415 case LINK_FAILINGOVER:
417 case LINK_FAILOVER_END_EVT:
418 l->state = LINK_RESET;
420 case LINK_PEER_RESET_EVT:
422 case LINK_ESTABLISH_EVT:
423 case LINK_FAILURE_EVT:
425 case LINK_FAILOVER_BEGIN_EVT:
426 case LINK_SYNCH_BEGIN_EVT:
427 case LINK_SYNCH_END_EVT:
432 case LINK_ESTABLISHING:
434 case LINK_ESTABLISH_EVT:
435 l->state = LINK_ESTABLISHED;
437 case LINK_FAILOVER_BEGIN_EVT:
438 l->state = LINK_FAILINGOVER;
441 l->state = LINK_RESET;
443 case LINK_FAILURE_EVT:
444 case LINK_PEER_RESET_EVT:
445 case LINK_SYNCH_BEGIN_EVT:
446 case LINK_FAILOVER_END_EVT:
448 case LINK_SYNCH_END_EVT:
453 case LINK_ESTABLISHED:
455 case LINK_PEER_RESET_EVT:
456 l->state = LINK_PEER_RESET;
457 rc |= TIPC_LINK_DOWN_EVT;
459 case LINK_FAILURE_EVT:
460 l->state = LINK_RESETTING;
461 rc |= TIPC_LINK_DOWN_EVT;
464 l->state = LINK_RESET;
466 case LINK_ESTABLISH_EVT:
467 case LINK_SYNCH_END_EVT:
469 case LINK_SYNCH_BEGIN_EVT:
470 l->state = LINK_SYNCHING;
472 case LINK_FAILOVER_BEGIN_EVT:
473 case LINK_FAILOVER_END_EVT:
480 case LINK_PEER_RESET_EVT:
481 l->state = LINK_PEER_RESET;
482 rc |= TIPC_LINK_DOWN_EVT;
484 case LINK_FAILURE_EVT:
485 l->state = LINK_RESETTING;
486 rc |= TIPC_LINK_DOWN_EVT;
489 l->state = LINK_RESET;
491 case LINK_ESTABLISH_EVT:
492 case LINK_SYNCH_BEGIN_EVT:
494 case LINK_SYNCH_END_EVT:
495 l->state = LINK_ESTABLISHED;
497 case LINK_FAILOVER_BEGIN_EVT:
498 case LINK_FAILOVER_END_EVT:
504 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
508 pr_err("Illegal FSM event %x in state %x on link %s\n",
509 evt, l->state, l->name);
513 /* link_profile_stats - update statistical profiling of traffic
515 static void link_profile_stats(struct tipc_link *l)
518 struct tipc_msg *msg;
521 /* Update counters used in statistical profiling of send traffic */
522 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
523 l->stats.queue_sz_counts++;
525 skb = skb_peek(&l->transmq);
529 length = msg_size(msg);
531 if (msg_user(msg) == MSG_FRAGMENTER) {
532 if (msg_type(msg) != FIRST_FRAGMENT)
534 length = msg_size(msg_get_wrapped(msg));
536 l->stats.msg_lengths_total += length;
537 l->stats.msg_length_counts++;
539 l->stats.msg_length_profile[0]++;
540 else if (length <= 256)
541 l->stats.msg_length_profile[1]++;
542 else if (length <= 1024)
543 l->stats.msg_length_profile[2]++;
544 else if (length <= 4096)
545 l->stats.msg_length_profile[3]++;
546 else if (length <= 16384)
547 l->stats.msg_length_profile[4]++;
548 else if (length <= 32768)
549 l->stats.msg_length_profile[5]++;
551 l->stats.msg_length_profile[6]++;
554 /* tipc_link_timeout - perform periodic task as instructed from node timeout
556 /* tipc_link_timeout - perform periodic task as instructed from node timeout
558 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
561 int mtyp = STATE_MSG;
564 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
565 u16 bc_acked = l->bc_rcvlink->acked;
566 bool bc_up = link_is_up(l->bc_rcvlink);
568 link_profile_stats(l);
571 case LINK_ESTABLISHED:
573 if (!l->silent_intv_cnt) {
574 if (bc_up && (bc_acked != bc_snt))
576 } else if (l->silent_intv_cnt <= l->abort_limit) {
580 rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
582 l->silent_intv_cnt++;
588 case LINK_ESTABLISHING:
592 case LINK_PEER_RESET:
594 case LINK_FAILINGOVER:
601 tipc_link_build_proto_msg(l, mtyp, prb, 0, 0, 0, xmitq);
607 * link_schedule_user - schedule a message sender for wakeup after congestion
608 * @link: congested link
609 * @list: message that was attempted sent
610 * Create pseudo msg to send back to user when congestion abates
611 * Does not consume buffer list
613 static int link_schedule_user(struct tipc_link *link, struct sk_buff_head *list)
615 struct tipc_msg *msg = buf_msg(skb_peek(list));
616 int imp = msg_importance(msg);
617 u32 oport = msg_origport(msg);
618 u32 addr = link_own_addr(link);
621 /* This really cannot happen... */
622 if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
623 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
626 /* Non-blocking sender: */
627 if (TIPC_SKB_CB(skb_peek(list))->wakeup_pending)
630 /* Create and schedule wakeup pseudo message */
631 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
632 addr, addr, oport, 0, 0);
635 TIPC_SKB_CB(skb)->chain_sz = skb_queue_len(list);
636 TIPC_SKB_CB(skb)->chain_imp = imp;
637 skb_queue_tail(&link->wakeupq, skb);
638 link->stats.link_congs++;
643 * link_prepare_wakeup - prepare users for wakeup after congestion
644 * @link: congested link
645 * Move a number of waiting users, as permitted by available space in
646 * the send queue, from link wait queue to node wait queue for wakeup
648 void link_prepare_wakeup(struct tipc_link *l)
650 int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
652 struct sk_buff *skb, *tmp;
654 skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
655 imp = TIPC_SKB_CB(skb)->chain_imp;
656 lim = l->window + l->backlog[imp].limit;
657 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
658 if ((pnd[imp] + l->backlog[imp].len) >= lim)
660 skb_unlink(skb, &l->wakeupq);
661 skb_queue_tail(l->inputq, skb);
666 * tipc_link_reset_fragments - purge link's inbound message fragments queue
667 * @l_ptr: pointer to link
669 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
671 kfree_skb(l_ptr->reasm_buf);
672 l_ptr->reasm_buf = NULL;
675 void tipc_link_purge_backlog(struct tipc_link *l)
677 __skb_queue_purge(&l->backlogq);
678 l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
679 l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
680 l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
681 l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
682 l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
686 * tipc_link_purge_queues - purge all pkt queues associated with link
687 * @l_ptr: pointer to link
689 void tipc_link_purge_queues(struct tipc_link *l_ptr)
691 __skb_queue_purge(&l_ptr->deferdq);
692 __skb_queue_purge(&l_ptr->transmq);
693 tipc_link_purge_backlog(l_ptr);
694 tipc_link_reset_fragments(l_ptr);
697 void tipc_link_reset(struct tipc_link *l)
699 /* Link is down, accept any session */
700 l->peer_session = WILDCARD_SESSION;
702 /* If peer is up, it only accepts an incremented session number */
703 msg_set_session(l->pmsg, msg_session(l->pmsg) + 1);
705 /* Prepare for renewed mtu size negotiation */
706 l->mtu = l->advertised_mtu;
708 /* Clean up all queues: */
709 __skb_queue_purge(&l->transmq);
710 __skb_queue_purge(&l->deferdq);
711 skb_queue_splice_init(&l->wakeupq, l->inputq);
713 tipc_link_purge_backlog(l);
714 kfree_skb(l->reasm_buf);
715 kfree_skb(l->failover_reasm_skb);
717 l->failover_reasm_skb = NULL;
722 l->silent_intv_cnt = 0;
723 l->stats.recv_info = 0;
725 l->bc_peer_is_up = false;
726 link_reset_statistics(l);
730 * __tipc_link_xmit(): same as tipc_link_xmit, but destlink is known & locked
732 * @list: chain of buffers containing message
734 * Consumes the buffer chain, except when returning an error code,
735 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
736 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
738 int __tipc_link_xmit(struct net *net, struct tipc_link *link,
739 struct sk_buff_head *list)
741 struct tipc_msg *msg = buf_msg(skb_peek(list));
742 unsigned int maxwin = link->window;
743 unsigned int i, imp = msg_importance(msg);
744 uint mtu = link->mtu;
745 u16 ack = mod(link->rcv_nxt - 1);
746 u16 seqno = link->snd_nxt;
747 u16 bc_ack = link->bc_rcvlink->rcv_nxt - 1;
748 struct tipc_media_addr *addr = link->media_addr;
749 struct sk_buff_head *transmq = &link->transmq;
750 struct sk_buff_head *backlogq = &link->backlogq;
751 struct sk_buff *skb, *bskb;
753 /* Match msg importance against this and all higher backlog limits: */
754 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
755 if (unlikely(link->backlog[i].len >= link->backlog[i].limit))
756 return link_schedule_user(link, list);
758 if (unlikely(msg_size(msg) > mtu))
761 /* Prepare each packet for sending, and add to relevant queue: */
762 while (skb_queue_len(list)) {
763 skb = skb_peek(list);
765 msg_set_seqno(msg, seqno);
766 msg_set_ack(msg, ack);
767 msg_set_bcast_ack(msg, bc_ack);
769 if (likely(skb_queue_len(transmq) < maxwin)) {
771 __skb_queue_tail(transmq, skb);
772 tipc_bearer_send(net, link->bearer_id, skb, addr);
773 link->rcv_unacked = 0;
777 if (tipc_msg_bundle(skb_peek_tail(backlogq), msg, mtu)) {
778 kfree_skb(__skb_dequeue(list));
779 link->stats.sent_bundled++;
782 if (tipc_msg_make_bundle(&bskb, msg, mtu, link->addr)) {
783 kfree_skb(__skb_dequeue(list));
784 __skb_queue_tail(backlogq, bskb);
785 link->backlog[msg_importance(buf_msg(bskb))].len++;
786 link->stats.sent_bundled++;
787 link->stats.sent_bundles++;
790 link->backlog[imp].len += skb_queue_len(list);
791 skb_queue_splice_tail_init(list, backlogq);
793 link->snd_nxt = seqno;
798 * tipc_link_xmit(): enqueue buffer list according to queue situation
800 * @list: chain of buffers containing message
801 * @xmitq: returned list of packets to be sent by caller
803 * Consumes the buffer chain, except when returning -ELINKCONG,
804 * since the caller then may want to make more send attempts.
805 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
806 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
808 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
809 struct sk_buff_head *xmitq)
811 struct tipc_msg *hdr = buf_msg(skb_peek(list));
812 unsigned int maxwin = l->window;
813 unsigned int i, imp = msg_importance(hdr);
814 unsigned int mtu = l->mtu;
815 u16 ack = l->rcv_nxt - 1;
816 u16 seqno = l->snd_nxt;
817 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
818 struct sk_buff_head *transmq = &l->transmq;
819 struct sk_buff_head *backlogq = &l->backlogq;
820 struct sk_buff *skb, *_skb, *bskb;
822 /* Match msg importance against this and all higher backlog limits: */
823 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
824 if (unlikely(l->backlog[i].len >= l->backlog[i].limit))
825 return link_schedule_user(l, list);
827 if (unlikely(msg_size(hdr) > mtu))
830 /* Prepare each packet for sending, and add to relevant queue: */
831 while (skb_queue_len(list)) {
832 skb = skb_peek(list);
834 msg_set_seqno(hdr, seqno);
835 msg_set_ack(hdr, ack);
836 msg_set_bcast_ack(hdr, bc_ack);
838 if (likely(skb_queue_len(transmq) < maxwin)) {
839 _skb = skb_clone(skb, GFP_ATOMIC);
843 __skb_queue_tail(transmq, skb);
844 __skb_queue_tail(xmitq, _skb);
845 TIPC_SKB_CB(skb)->ackers = l->ackers;
850 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
851 kfree_skb(__skb_dequeue(list));
852 l->stats.sent_bundled++;
855 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
856 kfree_skb(__skb_dequeue(list));
857 __skb_queue_tail(backlogq, bskb);
858 l->backlog[msg_importance(buf_msg(bskb))].len++;
859 l->stats.sent_bundled++;
860 l->stats.sent_bundles++;
863 l->backlog[imp].len += skb_queue_len(list);
864 skb_queue_splice_tail_init(list, backlogq);
871 * tipc_link_push_packets - push unsent packets to bearer
873 * Push out the unsent messages of a link where congestion
874 * has abated. Node is locked.
876 * Called with node locked
878 void tipc_link_push_packets(struct tipc_link *link)
881 struct tipc_msg *msg;
882 u16 seqno = link->snd_nxt;
883 u16 ack = mod(link->rcv_nxt - 1);
885 while (skb_queue_len(&link->transmq) < link->window) {
886 skb = __skb_dequeue(&link->backlogq);
889 TIPC_SKB_CB(skb)->ackers = link->ackers;
891 link->backlog[msg_importance(msg)].len--;
892 msg_set_ack(msg, ack);
893 msg_set_seqno(msg, seqno);
894 seqno = mod(seqno + 1);
895 /* msg_set_bcast_ack(msg, link->owner->bclink.last_in); */
896 link->rcv_unacked = 0;
897 __skb_queue_tail(&link->transmq, skb);
898 tipc_bearer_send(link->net, link->bearer_id,
899 skb, link->media_addr);
901 link->snd_nxt = seqno;
904 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
906 struct sk_buff *skb, *_skb;
907 struct tipc_msg *hdr;
908 u16 seqno = l->snd_nxt;
909 u16 ack = l->rcv_nxt - 1;
910 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
912 while (skb_queue_len(&l->transmq) < l->window) {
913 skb = skb_peek(&l->backlogq);
916 _skb = skb_clone(skb, GFP_ATOMIC);
919 __skb_dequeue(&l->backlogq);
921 l->backlog[msg_importance(hdr)].len--;
922 __skb_queue_tail(&l->transmq, skb);
923 __skb_queue_tail(xmitq, _skb);
924 TIPC_SKB_CB(skb)->ackers = l->ackers;
925 msg_set_seqno(hdr, seqno);
926 msg_set_ack(hdr, ack);
927 msg_set_bcast_ack(hdr, bc_ack);
934 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
936 struct tipc_msg *hdr = buf_msg(skb);
938 pr_warn("Retransmission failure on link <%s>\n", l->name);
939 link_print(l, "Resetting link ");
940 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
941 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
942 pr_info("sqno %u, prev: %x, src: %x\n",
943 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
946 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *skb,
949 struct tipc_msg *msg;
956 /* Detect repeated retransmit failures */
957 if (l_ptr->last_retransm == msg_seqno(msg)) {
958 if (++l_ptr->stale_count > 100) {
959 link_retransmit_failure(l_ptr, skb);
963 l_ptr->last_retransm = msg_seqno(msg);
964 l_ptr->stale_count = 1;
967 skb_queue_walk_from(&l_ptr->transmq, skb) {
971 msg_set_ack(msg, mod(l_ptr->rcv_nxt - 1));
972 /* msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in); */
973 tipc_bearer_send(l_ptr->net, l_ptr->bearer_id, skb,
976 l_ptr->stats.retransmitted++;
980 int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
981 struct sk_buff_head *xmitq)
983 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
984 struct tipc_msg *hdr;
985 u16 ack = l->rcv_nxt - 1;
986 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
991 /* Detect repeated retransmit failures on same packet */
992 if (likely(l->last_retransm != buf_seqno(skb))) {
993 l->last_retransm = buf_seqno(skb);
995 } else if (++l->stale_count > 100) {
996 link_retransmit_failure(l, skb);
997 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1000 /* Move forward to where retransmission should start */
1001 skb_queue_walk(&l->transmq, skb) {
1002 if (!less(buf_seqno(skb), from))
1006 skb_queue_walk_from(&l->transmq, skb) {
1007 if (more(buf_seqno(skb), to))
1010 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1013 hdr = buf_msg(_skb);
1014 msg_set_ack(hdr, ack);
1015 msg_set_bcast_ack(hdr, bc_ack);
1016 _skb->priority = TC_PRIO_CONTROL;
1017 __skb_queue_tail(xmitq, _skb);
1018 l->stats.retransmitted++;
1023 /* tipc_data_input - deliver data and name distr msgs to upper layer
1025 * Consumes buffer if message is of right type
1026 * Node lock must be held
1028 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1029 struct sk_buff_head *inputq)
1031 switch (msg_user(buf_msg(skb))) {
1032 case TIPC_LOW_IMPORTANCE:
1033 case TIPC_MEDIUM_IMPORTANCE:
1034 case TIPC_HIGH_IMPORTANCE:
1035 case TIPC_CRITICAL_IMPORTANCE:
1037 skb_queue_tail(inputq, skb);
1039 case NAME_DISTRIBUTOR:
1040 l->bc_rcvlink->state = LINK_ESTABLISHED;
1041 skb_queue_tail(l->namedq, skb);
1044 case TUNNEL_PROTOCOL:
1045 case MSG_FRAGMENTER:
1046 case BCAST_PROTOCOL:
1049 pr_warn("Dropping received illegal msg type\n");
1055 /* tipc_link_input - process packet that has passed link protocol check
1059 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1060 struct sk_buff_head *inputq)
1062 struct tipc_msg *hdr = buf_msg(skb);
1063 struct sk_buff **reasm_skb = &l->reasm_buf;
1064 struct sk_buff *iskb;
1065 struct sk_buff_head tmpq;
1066 int usr = msg_user(hdr);
1071 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1072 if (msg_type(hdr) == SYNCH_MSG) {
1073 __skb_queue_purge(&l->deferdq);
1076 if (!tipc_msg_extract(skb, &iskb, &ipos))
1081 if (less(msg_seqno(hdr), l->drop_point))
1083 if (tipc_data_input(l, skb, inputq))
1085 usr = msg_user(hdr);
1086 reasm_skb = &l->failover_reasm_skb;
1089 if (usr == MSG_BUNDLER) {
1090 skb_queue_head_init(&tmpq);
1091 l->stats.recv_bundles++;
1092 l->stats.recv_bundled += msg_msgcnt(hdr);
1093 while (tipc_msg_extract(skb, &iskb, &pos))
1094 tipc_data_input(l, iskb, &tmpq);
1095 tipc_skb_queue_splice_tail(&tmpq, inputq);
1097 } else if (usr == MSG_FRAGMENTER) {
1098 l->stats.recv_fragments++;
1099 if (tipc_buf_append(reasm_skb, &skb)) {
1100 l->stats.recv_fragmented++;
1101 tipc_data_input(l, skb, inputq);
1102 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1103 pr_warn_ratelimited("Unable to build fragment list\n");
1104 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1107 } else if (usr == BCAST_PROTOCOL) {
1108 tipc_bcast_lock(l->net);
1109 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1110 tipc_bcast_unlock(l->net);
1117 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1119 bool released = false;
1120 struct sk_buff *skb, *tmp;
1122 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1123 if (more(buf_seqno(skb), acked))
1125 __skb_unlink(skb, &l->transmq);
1132 /* tipc_link_build_ack_msg: prepare link acknowledge message for transmission
1134 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1135 * risk of ack storms towards the sender
1137 int tipc_link_build_ack_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1142 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1143 if (link_is_bc_rcvlink(l)) {
1144 if (((l->rcv_nxt ^ link_own_addr(l)) & 0xf) != 0xf)
1147 return TIPC_LINK_SND_BC_ACK;
1152 l->stats.sent_acks++;
1153 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1157 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1159 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1161 int mtyp = RESET_MSG;
1163 if (l->state == LINK_ESTABLISHING)
1164 mtyp = ACTIVATE_MSG;
1166 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1169 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1171 static void tipc_link_build_nack_msg(struct tipc_link *l,
1172 struct sk_buff_head *xmitq)
1174 u32 def_cnt = ++l->stats.deferred_recv;
1176 if (link_is_bc_rcvlink(l))
1179 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1180 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1183 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1184 * @l: the link that should handle the message
1186 * @xmitq: queue to place packets to be sent after this call
1188 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1189 struct sk_buff_head *xmitq)
1191 struct sk_buff_head *defq = &l->deferdq;
1192 struct tipc_msg *hdr;
1193 u16 seqno, rcv_nxt, win_lim;
1198 seqno = msg_seqno(hdr);
1199 rcv_nxt = l->rcv_nxt;
1200 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1202 /* Verify and update link state */
1203 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1204 return tipc_link_proto_rcv(l, skb, xmitq);
1206 if (unlikely(!link_is_up(l))) {
1207 if (l->state == LINK_ESTABLISHING)
1208 rc = TIPC_LINK_UP_EVT;
1212 /* Don't send probe at next timeout expiration */
1213 l->silent_intv_cnt = 0;
1215 /* Drop if outside receive window */
1216 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1217 l->stats.duplicates++;
1221 /* Forward queues and wake up waiting users */
1222 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1223 tipc_link_advance_backlog(l, xmitq);
1224 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1225 link_prepare_wakeup(l);
1228 /* Defer delivery if sequence gap */
1229 if (unlikely(seqno != rcv_nxt)) {
1230 __tipc_skb_queue_sorted(defq, seqno, skb);
1231 tipc_link_build_nack_msg(l, xmitq);
1235 /* Deliver packet */
1237 l->stats.recv_info++;
1238 if (!tipc_data_input(l, skb, l->inputq))
1239 rc |= tipc_link_input(l, skb, l->inputq);
1240 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1241 rc |= tipc_link_build_ack_msg(l, xmitq);
1242 if (unlikely(rc & ~TIPC_LINK_SND_BC_ACK))
1244 } while ((skb = __skb_dequeue(defq)));
1253 * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1255 * Returns increase in queue length (i.e. 0 or 1)
1257 u32 tipc_link_defer_pkt(struct sk_buff_head *list, struct sk_buff *skb)
1259 struct sk_buff *skb1;
1260 u16 seq_no = buf_seqno(skb);
1263 if (skb_queue_empty(list)) {
1264 __skb_queue_tail(list, skb);
1269 if (less(buf_seqno(skb_peek_tail(list)), seq_no)) {
1270 __skb_queue_tail(list, skb);
1274 /* Locate insertion point in queue, then insert; discard if duplicate */
1275 skb_queue_walk(list, skb1) {
1276 u16 curr_seqno = buf_seqno(skb1);
1278 if (seq_no == curr_seqno) {
1283 if (less(seq_no, curr_seqno))
1287 __skb_queue_before(list, skb1, skb);
1292 * Send protocol message to the other endpoint.
1294 void tipc_link_proto_xmit(struct tipc_link *l, u32 msg_typ, int probe_msg,
1295 u32 gap, u32 tolerance, u32 priority)
1297 struct sk_buff *skb = NULL;
1298 struct sk_buff_head xmitq;
1300 __skb_queue_head_init(&xmitq);
1301 tipc_link_build_proto_msg(l, msg_typ, probe_msg, gap,
1302 tolerance, priority, &xmitq);
1303 skb = __skb_dequeue(&xmitq);
1306 tipc_bearer_xmit_skb(l->net, l->bearer_id, skb, l->media_addr);
1310 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1311 u16 rcvgap, int tolerance, int priority,
1312 struct sk_buff_head *xmitq)
1314 struct sk_buff *skb = NULL;
1315 struct tipc_msg *hdr = l->pmsg;
1316 bool node_up = link_is_up(l->bc_rcvlink);
1318 /* Don't send protocol message during reset or link failover */
1319 if (tipc_link_is_blocked(l))
1322 msg_set_type(hdr, mtyp);
1323 msg_set_net_plane(hdr, l->net_plane);
1324 msg_set_next_sent(hdr, l->snd_nxt);
1325 msg_set_ack(hdr, l->rcv_nxt - 1);
1326 msg_set_bcast_ack(hdr, l->bc_rcvlink->rcv_nxt - 1);
1327 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1328 msg_set_link_tolerance(hdr, tolerance);
1329 msg_set_linkprio(hdr, priority);
1330 msg_set_redundant_link(hdr, node_up);
1331 msg_set_seq_gap(hdr, 0);
1333 /* Compatibility: created msg must not be in sequence with pkt flow */
1334 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1336 if (mtyp == STATE_MSG) {
1337 if (!tipc_link_is_up(l))
1340 /* Override rcvgap if there are packets in deferred queue */
1341 if (!skb_queue_empty(&l->deferdq))
1342 rcvgap = buf_seqno(skb_peek(&l->deferdq)) - l->rcv_nxt;
1344 msg_set_seq_gap(hdr, rcvgap);
1345 l->stats.sent_nacks++;
1347 msg_set_probe(hdr, probe);
1349 l->stats.sent_probes++;
1350 l->stats.sent_states++;
1353 /* RESET_MSG or ACTIVATE_MSG */
1354 msg_set_max_pkt(hdr, l->advertised_mtu);
1355 msg_set_ack(hdr, l->rcv_nxt - 1);
1356 msg_set_next_sent(hdr, 1);
1358 skb = tipc_buf_acquire(msg_size(hdr));
1361 skb_copy_to_linear_data(skb, hdr, msg_size(hdr));
1362 skb->priority = TC_PRIO_CONTROL;
1363 __skb_queue_tail(xmitq, skb);
1366 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1367 * with contents of the link's transmit and backlog queues.
1369 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1370 int mtyp, struct sk_buff_head *xmitq)
1372 struct sk_buff *skb, *tnlskb;
1373 struct tipc_msg *hdr, tnlhdr;
1374 struct sk_buff_head *queue = &l->transmq;
1375 struct sk_buff_head tmpxq, tnlq;
1376 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1381 skb_queue_head_init(&tnlq);
1382 skb_queue_head_init(&tmpxq);
1384 /* At least one packet required for safe algorithm => add dummy */
1385 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1386 BASIC_H_SIZE, 0, l->addr, link_own_addr(l),
1387 0, 0, TIPC_ERR_NO_PORT);
1389 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1392 skb_queue_tail(&tnlq, skb);
1393 tipc_link_xmit(l, &tnlq, &tmpxq);
1394 __skb_queue_purge(&tmpxq);
1396 /* Initialize reusable tunnel packet header */
1397 tipc_msg_init(link_own_addr(l), &tnlhdr, TUNNEL_PROTOCOL,
1398 mtyp, INT_H_SIZE, l->addr);
1399 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1400 msg_set_msgcnt(&tnlhdr, pktcnt);
1401 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1403 /* Wrap each packet into a tunnel packet */
1404 skb_queue_walk(queue, skb) {
1406 if (queue == &l->backlogq)
1407 msg_set_seqno(hdr, seqno++);
1408 pktlen = msg_size(hdr);
1409 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1410 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE);
1412 pr_warn("%sunable to send packet\n", link_co_err);
1415 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1416 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1417 __skb_queue_tail(&tnlq, tnlskb);
1419 if (queue != &l->backlogq) {
1420 queue = &l->backlogq;
1424 tipc_link_xmit(tnl, &tnlq, xmitq);
1426 if (mtyp == FAILOVER_MSG) {
1427 tnl->drop_point = l->rcv_nxt;
1428 tnl->failover_reasm_skb = l->reasm_buf;
1429 l->reasm_buf = NULL;
1433 /* tipc_link_proto_rcv(): receive link level protocol message :
1434 * Note that network plane id propagates through the network, and may
1435 * change at any time. The node with lowest numerical id determines
1438 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1439 struct sk_buff_head *xmitq)
1441 struct tipc_msg *hdr = buf_msg(skb);
1443 u16 ack = msg_ack(hdr);
1444 u16 gap = msg_seq_gap(hdr);
1445 u16 peers_snd_nxt = msg_next_sent(hdr);
1446 u16 peers_tol = msg_link_tolerance(hdr);
1447 u16 peers_prio = msg_linkprio(hdr);
1448 u16 rcv_nxt = l->rcv_nxt;
1449 int mtyp = msg_type(hdr);
1453 if (tipc_link_is_blocked(l) || !xmitq)
1456 if (link_own_addr(l) > msg_prevnode(hdr))
1457 l->net_plane = msg_net_plane(hdr);
1462 /* Ignore duplicate RESET with old session number */
1463 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1464 (l->peer_session != WILDCARD_SESSION))
1470 /* Complete own link name with peer's interface name */
1471 if_name = strrchr(l->name, ':') + 1;
1472 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1474 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1476 strncpy(if_name, msg_data(hdr), TIPC_MAX_IF_NAME);
1478 /* Update own tolerance if peer indicates a non-zero value */
1479 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1480 l->tolerance = peers_tol;
1482 /* Update own priority if peer's priority is higher */
1483 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1484 l->priority = peers_prio;
1486 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1487 if ((mtyp == RESET_MSG) || !link_is_up(l))
1488 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1490 /* ACTIVATE_MSG takes up link if it was already locally reset */
1491 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1492 rc = TIPC_LINK_UP_EVT;
1494 l->peer_session = msg_session(hdr);
1495 l->peer_bearer_id = msg_bearer_id(hdr);
1496 if (l->mtu > msg_max_pkt(hdr))
1497 l->mtu = msg_max_pkt(hdr);
1502 /* Update own tolerance if peer indicates a non-zero value */
1503 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1504 l->tolerance = peers_tol;
1506 l->silent_intv_cnt = 0;
1507 l->stats.recv_states++;
1509 l->stats.recv_probes++;
1511 if (!link_is_up(l)) {
1512 if (l->state == LINK_ESTABLISHING)
1513 rc = TIPC_LINK_UP_EVT;
1517 /* Send NACK if peer has sent pkts we haven't received yet */
1518 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1519 rcvgap = peers_snd_nxt - l->rcv_nxt;
1520 if (rcvgap || (msg_probe(hdr)))
1521 tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1523 tipc_link_release_pkts(l, ack);
1525 /* If NACK, retransmit will now start at right position */
1527 rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
1528 l->stats.recv_nacks++;
1531 tipc_link_advance_backlog(l, xmitq);
1532 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1533 link_prepare_wakeup(l);
1540 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1542 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1544 struct sk_buff_head *xmitq)
1546 struct sk_buff *skb;
1547 struct tipc_msg *hdr;
1548 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1549 u16 ack = l->rcv_nxt - 1;
1550 u16 gap_to = peers_snd_nxt - 1;
1552 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1553 0, l->addr, link_own_addr(l), 0, 0, 0);
1557 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1558 msg_set_bcast_ack(hdr, ack);
1559 msg_set_bcgap_after(hdr, ack);
1561 gap_to = buf_seqno(dfrd_skb) - 1;
1562 msg_set_bcgap_to(hdr, gap_to);
1563 msg_set_non_seq(hdr, bcast);
1564 __skb_queue_tail(xmitq, skb);
1568 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1570 * Give a newly added peer node the sequence number where it should
1571 * start receiving and acking broadcast packets.
1573 void tipc_link_build_bc_init_msg(struct tipc_link *l,
1574 struct sk_buff_head *xmitq)
1576 struct sk_buff_head list;
1578 __skb_queue_head_init(&list);
1579 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1581 tipc_link_xmit(l, &list, xmitq);
1584 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1586 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1588 int mtyp = msg_type(hdr);
1589 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1594 if (msg_user(hdr) == BCAST_PROTOCOL) {
1595 l->rcv_nxt = peers_snd_nxt;
1596 l->state = LINK_ESTABLISHED;
1600 if (l->peer_caps & TIPC_BCAST_SYNCH)
1603 if (msg_peer_node_is_up(hdr))
1606 /* Compatibility: accept older, less safe initial synch data */
1607 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1608 l->rcv_nxt = peers_snd_nxt;
1611 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1613 void tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1614 struct sk_buff_head *xmitq)
1616 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1621 if (!msg_peer_node_is_up(hdr))
1624 l->bc_peer_is_up = true;
1626 /* Ignore if peers_snd_nxt goes beyond receive window */
1627 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1630 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1631 l->nack_state = BC_NACK_SND_CONDITIONAL;
1635 /* Don't NACK if one was recently sent or peeked */
1636 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1637 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1641 /* Conditionally delay NACK sending until next synch rcv */
1642 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1643 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1644 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1648 /* Send NACK now but suppress next one */
1649 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1650 l->nack_state = BC_NACK_SND_SUPPRESS;
1653 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1654 struct sk_buff_head *xmitq)
1656 struct sk_buff *skb, *tmp;
1657 struct tipc_link *snd_l = l->bc_sndlink;
1659 if (!link_is_up(l) || !l->bc_peer_is_up)
1662 if (!more(acked, l->acked))
1665 /* Skip over packets peer has already acked */
1666 skb_queue_walk(&snd_l->transmq, skb) {
1667 if (more(buf_seqno(skb), l->acked))
1671 /* Update/release the packets peer is acking now */
1672 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1673 if (more(buf_seqno(skb), acked))
1675 if (!--TIPC_SKB_CB(skb)->ackers) {
1676 __skb_unlink(skb, &snd_l->transmq);
1681 tipc_link_advance_backlog(snd_l, xmitq);
1682 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1683 link_prepare_wakeup(snd_l);
1686 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1688 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1689 struct sk_buff_head *xmitq)
1691 struct tipc_msg *hdr = buf_msg(skb);
1692 u32 dnode = msg_destnode(hdr);
1693 int mtyp = msg_type(hdr);
1694 u16 acked = msg_bcast_ack(hdr);
1695 u16 from = acked + 1;
1696 u16 to = msg_bcgap_to(hdr);
1697 u16 peers_snd_nxt = to + 1;
1702 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1705 if (mtyp != STATE_MSG)
1708 if (dnode == link_own_addr(l)) {
1709 tipc_link_bc_ack_rcv(l, acked, xmitq);
1710 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1711 l->stats.recv_nacks++;
1715 /* Msg for other node => suppress own NACK at next sync if applicable */
1716 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1717 l->nack_state = BC_NACK_SND_SUPPRESS;
1722 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1724 int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1727 l->backlog[TIPC_LOW_IMPORTANCE].limit = win / 2;
1728 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = win;
1729 l->backlog[TIPC_HIGH_IMPORTANCE].limit = win / 2 * 3;
1730 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = win * 2;
1731 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1734 /* tipc_link_find_owner - locate owner node of link by link's name
1735 * @net: the applicable net namespace
1736 * @name: pointer to link name string
1737 * @bearer_id: pointer to index in 'node->links' array where the link was found.
1739 * Returns pointer to node owning the link, or 0 if no matching link is found.
1741 static struct tipc_node *tipc_link_find_owner(struct net *net,
1742 const char *link_name,
1743 unsigned int *bearer_id)
1745 struct tipc_net *tn = net_generic(net, tipc_net_id);
1746 struct tipc_link *l_ptr;
1747 struct tipc_node *n_ptr;
1748 struct tipc_node *found_node = NULL;
1753 list_for_each_entry_rcu(n_ptr, &tn->node_list, list) {
1754 tipc_node_lock(n_ptr);
1755 for (i = 0; i < MAX_BEARERS; i++) {
1756 l_ptr = n_ptr->links[i].link;
1757 if (l_ptr && !strcmp(l_ptr->name, link_name)) {
1763 tipc_node_unlock(n_ptr);
1773 * link_reset_statistics - reset link statistics
1774 * @l_ptr: pointer to link
1776 static void link_reset_statistics(struct tipc_link *l_ptr)
1778 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
1779 l_ptr->stats.sent_info = l_ptr->snd_nxt;
1780 l_ptr->stats.recv_info = l_ptr->rcv_nxt;
1783 static void link_print(struct tipc_link *l, const char *str)
1785 struct sk_buff *hskb = skb_peek(&l->transmq);
1786 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1787 u16 tail = l->snd_nxt - 1;
1789 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1790 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1791 skb_queue_len(&l->transmq), head, tail,
1792 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1795 /* Parse and validate nested (link) properties valid for media, bearer and link
1797 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1801 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1802 tipc_nl_prop_policy);
1806 if (props[TIPC_NLA_PROP_PRIO]) {
1809 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1810 if (prio > TIPC_MAX_LINK_PRI)
1814 if (props[TIPC_NLA_PROP_TOL]) {
1817 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1818 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1822 if (props[TIPC_NLA_PROP_WIN]) {
1825 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1826 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1833 int tipc_nl_link_set(struct sk_buff *skb, struct genl_info *info)
1839 struct tipc_link *link;
1840 struct tipc_node *node;
1841 struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
1842 struct net *net = sock_net(skb->sk);
1844 if (!info->attrs[TIPC_NLA_LINK])
1847 err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
1848 info->attrs[TIPC_NLA_LINK],
1849 tipc_nl_link_policy);
1853 if (!attrs[TIPC_NLA_LINK_NAME])
1856 name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
1858 if (strcmp(name, tipc_bclink_name) == 0)
1859 return tipc_nl_bc_link_set(net, attrs);
1861 node = tipc_link_find_owner(net, name, &bearer_id);
1865 tipc_node_lock(node);
1867 link = node->links[bearer_id].link;
1873 if (attrs[TIPC_NLA_LINK_PROP]) {
1874 struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
1876 err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP],
1883 if (props[TIPC_NLA_PROP_TOL]) {
1886 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1887 link->tolerance = tol;
1888 tipc_link_proto_xmit(link, STATE_MSG, 0, 0, tol, 0);
1890 if (props[TIPC_NLA_PROP_PRIO]) {
1893 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1894 link->priority = prio;
1895 tipc_link_proto_xmit(link, STATE_MSG, 0, 0, 0, prio);
1897 if (props[TIPC_NLA_PROP_WIN]) {
1900 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1901 tipc_link_set_queue_limits(link, win);
1906 tipc_node_unlock(node);
1911 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1914 struct nlattr *stats;
1921 struct nla_map map[] = {
1922 {TIPC_NLA_STATS_RX_INFO, s->recv_info},
1923 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1924 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1925 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1926 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1927 {TIPC_NLA_STATS_TX_INFO, s->sent_info},
1928 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1929 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1930 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1931 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1932 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1933 s->msg_length_counts : 1},
1934 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1935 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1936 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1937 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1938 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1939 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1940 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1941 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1942 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1943 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1944 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1945 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1946 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1947 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1948 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1949 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1950 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1951 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1952 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1953 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1954 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1955 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1956 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1959 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1963 for (i = 0; i < ARRAY_SIZE(map); i++)
1964 if (nla_put_u32(skb, map[i].key, map[i].val))
1967 nla_nest_end(skb, stats);
1971 nla_nest_cancel(skb, stats);
1976 /* Caller should hold appropriate locks to protect the link */
1977 static int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1978 struct tipc_link *link, int nlflags)
1982 struct nlattr *attrs;
1983 struct nlattr *prop;
1984 struct tipc_net *tn = net_generic(net, tipc_net_id);
1986 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1987 nlflags, TIPC_NL_LINK_GET);
1991 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1995 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1997 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1998 tipc_cluster_mask(tn->own_addr)))
2000 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2002 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->rcv_nxt))
2004 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->snd_nxt))
2007 if (tipc_link_is_up(link))
2008 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2011 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2014 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2017 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2019 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2021 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2024 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2026 nla_nest_end(msg->skb, prop);
2028 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2032 nla_nest_end(msg->skb, attrs);
2033 genlmsg_end(msg->skb, hdr);
2038 nla_nest_cancel(msg->skb, prop);
2040 nla_nest_cancel(msg->skb, attrs);
2042 genlmsg_cancel(msg->skb, hdr);
2047 /* Caller should hold node lock */
2048 static int __tipc_nl_add_node_links(struct net *net, struct tipc_nl_msg *msg,
2049 struct tipc_node *node, u32 *prev_link)
2054 for (i = *prev_link; i < MAX_BEARERS; i++) {
2057 if (!node->links[i].link)
2060 err = __tipc_nl_add_link(net, msg,
2061 node->links[i].link, NLM_F_MULTI);
2070 int tipc_nl_link_dump(struct sk_buff *skb, struct netlink_callback *cb)
2072 struct net *net = sock_net(skb->sk);
2073 struct tipc_net *tn = net_generic(net, tipc_net_id);
2074 struct tipc_node *node;
2075 struct tipc_nl_msg msg;
2076 u32 prev_node = cb->args[0];
2077 u32 prev_link = cb->args[1];
2078 int done = cb->args[2];
2085 msg.portid = NETLINK_CB(cb->skb).portid;
2086 msg.seq = cb->nlh->nlmsg_seq;
2090 node = tipc_node_find(net, prev_node);
2092 /* We never set seq or call nl_dump_check_consistent()
2093 * this means that setting prev_seq here will cause the
2094 * consistence check to fail in the netlink callback
2095 * handler. Resulting in the last NLMSG_DONE message
2096 * having the NLM_F_DUMP_INTR flag set.
2101 tipc_node_put(node);
2103 list_for_each_entry_continue_rcu(node, &tn->node_list,
2105 tipc_node_lock(node);
2106 err = __tipc_nl_add_node_links(net, &msg, node,
2108 tipc_node_unlock(node);
2112 prev_node = node->addr;
2115 err = tipc_nl_add_bc_link(net, &msg);
2119 list_for_each_entry_rcu(node, &tn->node_list, list) {
2120 tipc_node_lock(node);
2121 err = __tipc_nl_add_node_links(net, &msg, node,
2123 tipc_node_unlock(node);
2127 prev_node = node->addr;
2134 cb->args[0] = prev_node;
2135 cb->args[1] = prev_link;
2141 int tipc_nl_link_get(struct sk_buff *skb, struct genl_info *info)
2143 struct net *net = genl_info_net(info);
2144 struct tipc_nl_msg msg;
2148 msg.portid = info->snd_portid;
2149 msg.seq = info->snd_seq;
2151 if (!info->attrs[TIPC_NLA_LINK_NAME])
2153 name = nla_data(info->attrs[TIPC_NLA_LINK_NAME]);
2155 msg.skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
2159 if (strcmp(name, tipc_bclink_name) == 0) {
2160 err = tipc_nl_add_bc_link(net, &msg);
2162 nlmsg_free(msg.skb);
2167 struct tipc_node *node;
2168 struct tipc_link *link;
2170 node = tipc_link_find_owner(net, name, &bearer_id);
2174 tipc_node_lock(node);
2175 link = node->links[bearer_id].link;
2177 tipc_node_unlock(node);
2178 nlmsg_free(msg.skb);
2182 err = __tipc_nl_add_link(net, &msg, link, 0);
2183 tipc_node_unlock(node);
2185 nlmsg_free(msg.skb);
2190 return genlmsg_reply(msg.skb, info);
2193 int tipc_nl_link_reset_stats(struct sk_buff *skb, struct genl_info *info)
2197 unsigned int bearer_id;
2198 struct tipc_link *link;
2199 struct tipc_node *node;
2200 struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
2201 struct net *net = sock_net(skb->sk);
2203 if (!info->attrs[TIPC_NLA_LINK])
2206 err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
2207 info->attrs[TIPC_NLA_LINK],
2208 tipc_nl_link_policy);
2212 if (!attrs[TIPC_NLA_LINK_NAME])
2215 link_name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
2217 if (strcmp(link_name, tipc_bclink_name) == 0) {
2218 err = tipc_bclink_reset_stats(net);
2224 node = tipc_link_find_owner(net, link_name, &bearer_id);
2228 tipc_node_lock(node);
2230 link = node->links[bearer_id].link;
2232 tipc_node_unlock(node);
2236 link_reset_statistics(link);
2238 tipc_node_unlock(node);