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 u32 sent_info; /* used in counting # sent packets */
50 u32 recv_info; /* used in counting # recv'd packets */
67 u32 link_congs; /* # port sends blocked by congestion */
70 u32 max_queue_sz; /* send queue size high water mark */
71 u32 accu_queue_sz; /* used for send queue size profiling */
72 u32 queue_sz_counts; /* used for send queue size profiling */
73 u32 msg_length_counts; /* used for message length profiling */
74 u32 msg_lengths_total; /* used for message length profiling */
75 u32 msg_length_profile[7]; /* used for msg. length profiling */
79 * struct tipc_link - TIPC link data structure
80 * @addr: network address of link's peer node
81 * @name: link name character string
82 * @media_addr: media address to use when sending messages over link
84 * @net: pointer to namespace struct
85 * @refcnt: reference counter for permanent references (owner node & timer)
86 * @peer_session: link session # being used by peer end of link
87 * @peer_bearer_id: bearer id used by link's peer endpoint
88 * @bearer_id: local bearer id used by link
89 * @tolerance: minimum link continuity loss needed to reset link [in ms]
90 * @keepalive_intv: link keepalive timer interval
91 * @abort_limit: # of unacknowledged continuity probes needed to reset link
92 * @state: current state of link FSM
93 * @peer_caps: bitmap describing capabilities of peer node
94 * @silent_intv_cnt: # of timer intervals without any reception from peer
95 * @proto_msg: template for control messages generated by link
96 * @pmsg: convenience pointer to "proto_msg" field
97 * @priority: current link priority
98 * @net_plane: current link network plane ('A' through 'H')
99 * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
100 * @exp_msg_count: # of tunnelled messages expected during link changeover
101 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
102 * @mtu: current maximum packet size for this link
103 * @advertised_mtu: advertised own mtu when link is being established
104 * @transmitq: queue for sent, non-acked messages
105 * @backlogq: queue for messages waiting to be sent
106 * @snt_nxt: next sequence number to use for outbound messages
107 * @last_retransmitted: sequence number of most recently retransmitted message
108 * @stale_count: # of identical retransmit requests made by peer
109 * @ackers: # of peers that needs to ack each packet before it can be released
110 * @acked: # last packet acked by a certain peer. Used for broadcast.
111 * @rcv_nxt: next sequence number to expect for inbound messages
112 * @deferred_queue: deferred queue saved OOS b'cast message received from node
113 * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
114 * @inputq: buffer queue for messages to be delivered upwards
115 * @namedq: buffer queue for name table messages to be delivered upwards
116 * @next_out: ptr to first unsent outbound message in queue
117 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
118 * @long_msg_seq_no: next identifier to use for outbound fragmented messages
119 * @reasm_buf: head of partially reassembled inbound message fragments
120 * @bc_rcvr: marks that this is a broadcast receiver link
121 * @stats: collects statistics regarding link activity
125 char name[TIPC_MAX_LINK_NAME];
128 /* Management and link supervision data */
133 unsigned long keepalive_intv;
140 unchar hdr[INT_H_SIZE];
141 unchar body[TIPC_MAX_IF_NAME];
143 struct tipc_msg *pmsg;
149 struct sk_buff *failover_reasm_skb;
151 /* Max packet negotiation */
156 struct sk_buff_head transmq;
157 struct sk_buff_head backlogq;
170 struct sk_buff_head deferdq;
171 struct sk_buff_head *inputq;
172 struct sk_buff_head *namedq;
174 /* Congestion handling */
175 struct sk_buff_head wakeupq;
177 /* Fragmentation/reassembly */
178 struct sk_buff *reasm_buf;
183 struct tipc_link *bc_rcvlink;
184 struct tipc_link *bc_sndlink;
189 struct tipc_stats stats;
193 * Error message prefixes
195 static const char *link_co_err = "Link tunneling error, ";
196 static const char *link_rst_msg = "Resetting link ";
198 /* Properties valid for media, bearar and link */
199 static const struct nla_policy tipc_nl_prop_policy[TIPC_NLA_PROP_MAX + 1] = {
200 [TIPC_NLA_PROP_UNSPEC] = { .type = NLA_UNSPEC },
201 [TIPC_NLA_PROP_PRIO] = { .type = NLA_U32 },
202 [TIPC_NLA_PROP_TOL] = { .type = NLA_U32 },
203 [TIPC_NLA_PROP_WIN] = { .type = NLA_U32 }
206 /* Send states for broadcast NACKs
209 BC_NACK_SND_CONDITIONAL,
210 BC_NACK_SND_UNCONDITIONAL,
211 BC_NACK_SND_SUPPRESS,
215 * Interval between NACKs when packets arrive out of order
217 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
219 * Out-of-range value for link session numbers
221 #define WILDCARD_SESSION 0x10000
226 LINK_ESTABLISHED = 0xe,
227 LINK_ESTABLISHING = 0xe << 4,
228 LINK_RESET = 0x1 << 8,
229 LINK_RESETTING = 0x2 << 12,
230 LINK_PEER_RESET = 0xd << 16,
231 LINK_FAILINGOVER = 0xf << 20,
232 LINK_SYNCHING = 0xc << 24
235 /* Link FSM state checking routines
237 static int link_is_up(struct tipc_link *l)
239 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
242 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
243 struct sk_buff_head *xmitq);
244 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
245 u16 rcvgap, int tolerance, int priority,
246 struct sk_buff_head *xmitq);
247 static void link_print(struct tipc_link *l, const char *str);
248 static void tipc_link_build_nack_msg(struct tipc_link *l,
249 struct sk_buff_head *xmitq);
250 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
251 struct sk_buff_head *xmitq);
252 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
255 * Simple non-static link routines (i.e. referenced outside this file)
257 bool tipc_link_is_up(struct tipc_link *l)
259 return link_is_up(l);
262 bool tipc_link_peer_is_down(struct tipc_link *l)
264 return l->state == LINK_PEER_RESET;
267 bool tipc_link_is_reset(struct tipc_link *l)
269 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
272 bool tipc_link_is_establishing(struct tipc_link *l)
274 return l->state == LINK_ESTABLISHING;
277 bool tipc_link_is_synching(struct tipc_link *l)
279 return l->state == LINK_SYNCHING;
282 bool tipc_link_is_failingover(struct tipc_link *l)
284 return l->state == LINK_FAILINGOVER;
287 bool tipc_link_is_blocked(struct tipc_link *l)
289 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
292 static bool link_is_bc_sndlink(struct tipc_link *l)
294 return !l->bc_sndlink;
297 static bool link_is_bc_rcvlink(struct tipc_link *l)
299 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
302 int tipc_link_is_active(struct tipc_link *l)
307 void tipc_link_set_active(struct tipc_link *l, bool active)
312 u32 tipc_link_id(struct tipc_link *l)
314 return l->peer_bearer_id << 16 | l->bearer_id;
317 int tipc_link_window(struct tipc_link *l)
322 int tipc_link_prio(struct tipc_link *l)
327 unsigned long tipc_link_tolerance(struct tipc_link *l)
332 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
337 char tipc_link_plane(struct tipc_link *l)
342 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
343 struct tipc_link *uc_l,
344 struct sk_buff_head *xmitq)
346 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
349 rcv_l->acked = snd_l->snd_nxt - 1;
350 snd_l->state = LINK_ESTABLISHED;
351 tipc_link_build_bc_init_msg(uc_l, xmitq);
354 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
355 struct tipc_link *rcv_l,
356 struct sk_buff_head *xmitq)
358 u16 ack = snd_l->snd_nxt - 1;
361 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
362 tipc_link_reset(rcv_l);
363 rcv_l->state = LINK_RESET;
364 if (!snd_l->ackers) {
365 tipc_link_reset(snd_l);
366 snd_l->state = LINK_RESET;
367 __skb_queue_purge(xmitq);
371 int tipc_link_bc_peers(struct tipc_link *l)
376 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
381 int tipc_link_mtu(struct tipc_link *l)
386 u16 tipc_link_rcv_nxt(struct tipc_link *l)
391 u16 tipc_link_acked(struct tipc_link *l)
396 char *tipc_link_name(struct tipc_link *l)
401 static u32 link_own_addr(struct tipc_link *l)
403 return msg_prevnode(l->pmsg);
406 void tipc_link_reinit(struct tipc_link *l, u32 addr)
408 msg_set_prevnode(l->pmsg, addr);
412 * tipc_link_create - create a new link
413 * @n: pointer to associated node
414 * @if_name: associated interface name
415 * @bearer_id: id (index) of associated bearer
416 * @tolerance: link tolerance to be used by link
417 * @net_plane: network plane (A,B,c..) this link belongs to
418 * @mtu: mtu to be advertised by link
419 * @priority: priority to be used by link
420 * @window: send window to be used by link
421 * @session: session to be used by link
422 * @ownnode: identity of own node
423 * @peer: node id of peer node
424 * @peer_caps: bitmap describing peer node capabilities
425 * @bc_sndlink: the namespace global link used for broadcast sending
426 * @bc_rcvlink: the peer specific link used for broadcast reception
427 * @inputq: queue to put messages ready for delivery
428 * @namedq: queue to put binding table update messages ready for delivery
429 * @link: return value, pointer to put the created link
431 * Returns true if link was created, otherwise false
433 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
434 int tolerance, char net_plane, u32 mtu, int priority,
435 int window, u32 session, u32 ownnode, u32 peer,
437 struct tipc_link *bc_sndlink,
438 struct tipc_link *bc_rcvlink,
439 struct sk_buff_head *inputq,
440 struct sk_buff_head *namedq,
441 struct tipc_link **link)
444 struct tipc_msg *hdr;
446 l = kzalloc(sizeof(*l), GFP_ATOMIC);
450 l->pmsg = (struct tipc_msg *)&l->proto_msg;
452 tipc_msg_init(ownnode, hdr, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, peer);
453 msg_set_size(hdr, sizeof(l->proto_msg));
454 msg_set_session(hdr, session);
455 msg_set_bearer_id(hdr, l->bearer_id);
457 /* Note: peer i/f name is completed by reset/activate message */
458 sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
459 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
460 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
461 strcpy((char *)msg_data(hdr), if_name);
464 l->peer_caps = peer_caps;
466 l->peer_session = WILDCARD_SESSION;
467 l->bearer_id = bearer_id;
468 l->tolerance = tolerance;
469 l->net_plane = net_plane;
470 l->advertised_mtu = mtu;
472 l->priority = priority;
473 tipc_link_set_queue_limits(l, window);
475 l->bc_sndlink = bc_sndlink;
476 l->bc_rcvlink = bc_rcvlink;
479 l->state = LINK_RESETTING;
480 __skb_queue_head_init(&l->transmq);
481 __skb_queue_head_init(&l->backlogq);
482 __skb_queue_head_init(&l->deferdq);
483 skb_queue_head_init(&l->wakeupq);
484 skb_queue_head_init(l->inputq);
489 * tipc_link_bc_create - create new link to be used for broadcast
490 * @n: pointer to associated node
491 * @mtu: mtu to be used
492 * @window: send window to be used
493 * @inputq: queue to put messages ready for delivery
494 * @namedq: queue to put binding table update messages ready for delivery
495 * @link: return value, pointer to put the created link
497 * Returns true if link was created, otherwise false
499 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
500 int mtu, int window, u16 peer_caps,
501 struct sk_buff_head *inputq,
502 struct sk_buff_head *namedq,
503 struct tipc_link *bc_sndlink,
504 struct tipc_link **link)
508 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
509 0, ownnode, peer, peer_caps, bc_sndlink,
510 NULL, inputq, namedq, link))
514 strcpy(l->name, tipc_bclink_name);
516 l->state = LINK_RESET;
520 /* Broadcast send link is always up */
521 if (link_is_bc_sndlink(l))
522 l->state = LINK_ESTABLISHED;
528 * tipc_link_fsm_evt - link finite state machine
529 * @l: pointer to link
530 * @evt: state machine event to be processed
532 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
539 case LINK_PEER_RESET_EVT:
540 l->state = LINK_PEER_RESET;
543 l->state = LINK_RESET;
545 case LINK_FAILURE_EVT:
546 case LINK_FAILOVER_BEGIN_EVT:
547 case LINK_ESTABLISH_EVT:
548 case LINK_FAILOVER_END_EVT:
549 case LINK_SYNCH_BEGIN_EVT:
550 case LINK_SYNCH_END_EVT:
557 case LINK_PEER_RESET_EVT:
558 l->state = LINK_ESTABLISHING;
560 case LINK_FAILOVER_BEGIN_EVT:
561 l->state = LINK_FAILINGOVER;
562 case LINK_FAILURE_EVT:
564 case LINK_ESTABLISH_EVT:
565 case LINK_FAILOVER_END_EVT:
567 case LINK_SYNCH_BEGIN_EVT:
568 case LINK_SYNCH_END_EVT:
573 case LINK_PEER_RESET:
576 l->state = LINK_ESTABLISHING;
578 case LINK_PEER_RESET_EVT:
579 case LINK_ESTABLISH_EVT:
580 case LINK_FAILURE_EVT:
582 case LINK_SYNCH_BEGIN_EVT:
583 case LINK_SYNCH_END_EVT:
584 case LINK_FAILOVER_BEGIN_EVT:
585 case LINK_FAILOVER_END_EVT:
590 case LINK_FAILINGOVER:
592 case LINK_FAILOVER_END_EVT:
593 l->state = LINK_RESET;
595 case LINK_PEER_RESET_EVT:
597 case LINK_ESTABLISH_EVT:
598 case LINK_FAILURE_EVT:
600 case LINK_FAILOVER_BEGIN_EVT:
601 case LINK_SYNCH_BEGIN_EVT:
602 case LINK_SYNCH_END_EVT:
607 case LINK_ESTABLISHING:
609 case LINK_ESTABLISH_EVT:
610 l->state = LINK_ESTABLISHED;
612 case LINK_FAILOVER_BEGIN_EVT:
613 l->state = LINK_FAILINGOVER;
616 l->state = LINK_RESET;
618 case LINK_FAILURE_EVT:
619 case LINK_PEER_RESET_EVT:
620 case LINK_SYNCH_BEGIN_EVT:
621 case LINK_FAILOVER_END_EVT:
623 case LINK_SYNCH_END_EVT:
628 case LINK_ESTABLISHED:
630 case LINK_PEER_RESET_EVT:
631 l->state = LINK_PEER_RESET;
632 rc |= TIPC_LINK_DOWN_EVT;
634 case LINK_FAILURE_EVT:
635 l->state = LINK_RESETTING;
636 rc |= TIPC_LINK_DOWN_EVT;
639 l->state = LINK_RESET;
641 case LINK_ESTABLISH_EVT:
642 case LINK_SYNCH_END_EVT:
644 case LINK_SYNCH_BEGIN_EVT:
645 l->state = LINK_SYNCHING;
647 case LINK_FAILOVER_BEGIN_EVT:
648 case LINK_FAILOVER_END_EVT:
655 case LINK_PEER_RESET_EVT:
656 l->state = LINK_PEER_RESET;
657 rc |= TIPC_LINK_DOWN_EVT;
659 case LINK_FAILURE_EVT:
660 l->state = LINK_RESETTING;
661 rc |= TIPC_LINK_DOWN_EVT;
664 l->state = LINK_RESET;
666 case LINK_ESTABLISH_EVT:
667 case LINK_SYNCH_BEGIN_EVT:
669 case LINK_SYNCH_END_EVT:
670 l->state = LINK_ESTABLISHED;
672 case LINK_FAILOVER_BEGIN_EVT:
673 case LINK_FAILOVER_END_EVT:
679 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
683 pr_err("Illegal FSM event %x in state %x on link %s\n",
684 evt, l->state, l->name);
688 /* link_profile_stats - update statistical profiling of traffic
690 static void link_profile_stats(struct tipc_link *l)
693 struct tipc_msg *msg;
696 /* Update counters used in statistical profiling of send traffic */
697 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
698 l->stats.queue_sz_counts++;
700 skb = skb_peek(&l->transmq);
704 length = msg_size(msg);
706 if (msg_user(msg) == MSG_FRAGMENTER) {
707 if (msg_type(msg) != FIRST_FRAGMENT)
709 length = msg_size(msg_get_wrapped(msg));
711 l->stats.msg_lengths_total += length;
712 l->stats.msg_length_counts++;
714 l->stats.msg_length_profile[0]++;
715 else if (length <= 256)
716 l->stats.msg_length_profile[1]++;
717 else if (length <= 1024)
718 l->stats.msg_length_profile[2]++;
719 else if (length <= 4096)
720 l->stats.msg_length_profile[3]++;
721 else if (length <= 16384)
722 l->stats.msg_length_profile[4]++;
723 else if (length <= 32768)
724 l->stats.msg_length_profile[5]++;
726 l->stats.msg_length_profile[6]++;
729 /* tipc_link_timeout - perform periodic task as instructed from node timeout
731 /* tipc_link_timeout - perform periodic task as instructed from node timeout
733 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
736 int mtyp = STATE_MSG;
739 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
740 u16 bc_acked = l->bc_rcvlink->acked;
741 bool bc_up = link_is_up(l->bc_rcvlink);
743 link_profile_stats(l);
746 case LINK_ESTABLISHED:
748 if (!l->silent_intv_cnt) {
749 if (bc_up && (bc_acked != bc_snt))
751 } else if (l->silent_intv_cnt <= l->abort_limit) {
755 rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
757 l->silent_intv_cnt++;
763 case LINK_ESTABLISHING:
767 case LINK_PEER_RESET:
769 case LINK_FAILINGOVER:
776 tipc_link_build_proto_msg(l, mtyp, prb, 0, 0, 0, xmitq);
782 * link_schedule_user - schedule a message sender for wakeup after congestion
783 * @link: congested link
784 * @list: message that was attempted sent
785 * Create pseudo msg to send back to user when congestion abates
786 * Does not consume buffer list
788 static int link_schedule_user(struct tipc_link *link, struct sk_buff_head *list)
790 struct tipc_msg *msg = buf_msg(skb_peek(list));
791 int imp = msg_importance(msg);
792 u32 oport = msg_origport(msg);
793 u32 addr = link_own_addr(link);
796 /* This really cannot happen... */
797 if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
798 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
801 /* Non-blocking sender: */
802 if (TIPC_SKB_CB(skb_peek(list))->wakeup_pending)
805 /* Create and schedule wakeup pseudo message */
806 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
807 addr, addr, oport, 0, 0);
810 TIPC_SKB_CB(skb)->chain_sz = skb_queue_len(list);
811 TIPC_SKB_CB(skb)->chain_imp = imp;
812 skb_queue_tail(&link->wakeupq, skb);
813 link->stats.link_congs++;
818 * link_prepare_wakeup - prepare users for wakeup after congestion
819 * @link: congested link
820 * Move a number of waiting users, as permitted by available space in
821 * the send queue, from link wait queue to node wait queue for wakeup
823 void link_prepare_wakeup(struct tipc_link *l)
825 int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
827 struct sk_buff *skb, *tmp;
829 skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
830 imp = TIPC_SKB_CB(skb)->chain_imp;
831 lim = l->window + l->backlog[imp].limit;
832 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
833 if ((pnd[imp] + l->backlog[imp].len) >= lim)
835 skb_unlink(skb, &l->wakeupq);
836 skb_queue_tail(l->inputq, skb);
840 void tipc_link_reset(struct tipc_link *l)
842 /* Link is down, accept any session */
843 l->peer_session = WILDCARD_SESSION;
845 /* If peer is up, it only accepts an incremented session number */
846 msg_set_session(l->pmsg, msg_session(l->pmsg) + 1);
848 /* Prepare for renewed mtu size negotiation */
849 l->mtu = l->advertised_mtu;
851 /* Clean up all queues and counters: */
852 __skb_queue_purge(&l->transmq);
853 __skb_queue_purge(&l->deferdq);
854 skb_queue_splice_init(&l->wakeupq, l->inputq);
855 __skb_queue_purge(&l->backlogq);
856 l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
857 l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
858 l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
859 l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
860 l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
861 kfree_skb(l->reasm_buf);
862 kfree_skb(l->failover_reasm_skb);
864 l->failover_reasm_skb = NULL;
869 l->silent_intv_cnt = 0;
870 l->stats.recv_info = 0;
872 l->bc_peer_is_up = false;
873 tipc_link_reset_stats(l);
877 * tipc_link_xmit(): enqueue buffer list according to queue situation
879 * @list: chain of buffers containing message
880 * @xmitq: returned list of packets to be sent by caller
882 * Consumes the buffer chain, except when returning -ELINKCONG,
883 * since the caller then may want to make more send attempts.
884 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
885 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
887 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
888 struct sk_buff_head *xmitq)
890 struct tipc_msg *hdr = buf_msg(skb_peek(list));
891 unsigned int maxwin = l->window;
892 unsigned int i, imp = msg_importance(hdr);
893 unsigned int mtu = l->mtu;
894 u16 ack = l->rcv_nxt - 1;
895 u16 seqno = l->snd_nxt;
896 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
897 struct sk_buff_head *transmq = &l->transmq;
898 struct sk_buff_head *backlogq = &l->backlogq;
899 struct sk_buff *skb, *_skb, *bskb;
901 /* Match msg importance against this and all higher backlog limits: */
902 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
903 if (unlikely(l->backlog[i].len >= l->backlog[i].limit))
904 return link_schedule_user(l, list);
906 if (unlikely(msg_size(hdr) > mtu)) {
907 skb_queue_purge(list);
911 /* Prepare each packet for sending, and add to relevant queue: */
912 while (skb_queue_len(list)) {
913 skb = skb_peek(list);
915 msg_set_seqno(hdr, seqno);
916 msg_set_ack(hdr, ack);
917 msg_set_bcast_ack(hdr, bc_ack);
919 if (likely(skb_queue_len(transmq) < maxwin)) {
920 _skb = skb_clone(skb, GFP_ATOMIC);
922 skb_queue_purge(list);
926 __skb_queue_tail(transmq, skb);
927 __skb_queue_tail(xmitq, _skb);
928 TIPC_SKB_CB(skb)->ackers = l->ackers;
933 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
934 kfree_skb(__skb_dequeue(list));
935 l->stats.sent_bundled++;
938 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
939 kfree_skb(__skb_dequeue(list));
940 __skb_queue_tail(backlogq, bskb);
941 l->backlog[msg_importance(buf_msg(bskb))].len++;
942 l->stats.sent_bundled++;
943 l->stats.sent_bundles++;
946 l->backlog[imp].len += skb_queue_len(list);
947 skb_queue_splice_tail_init(list, backlogq);
953 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
955 struct sk_buff *skb, *_skb;
956 struct tipc_msg *hdr;
957 u16 seqno = l->snd_nxt;
958 u16 ack = l->rcv_nxt - 1;
959 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
961 while (skb_queue_len(&l->transmq) < l->window) {
962 skb = skb_peek(&l->backlogq);
965 _skb = skb_clone(skb, GFP_ATOMIC);
968 __skb_dequeue(&l->backlogq);
970 l->backlog[msg_importance(hdr)].len--;
971 __skb_queue_tail(&l->transmq, skb);
972 __skb_queue_tail(xmitq, _skb);
973 TIPC_SKB_CB(skb)->ackers = l->ackers;
974 msg_set_seqno(hdr, seqno);
975 msg_set_ack(hdr, ack);
976 msg_set_bcast_ack(hdr, bc_ack);
983 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
985 struct tipc_msg *hdr = buf_msg(skb);
987 pr_warn("Retransmission failure on link <%s>\n", l->name);
988 link_print(l, "Resetting link ");
989 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
990 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
991 pr_info("sqno %u, prev: %x, src: %x\n",
992 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
995 int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
996 struct sk_buff_head *xmitq)
998 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
999 struct tipc_msg *hdr;
1000 u16 ack = l->rcv_nxt - 1;
1001 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1006 /* Detect repeated retransmit failures on same packet */
1007 if (likely(l->last_retransm != buf_seqno(skb))) {
1008 l->last_retransm = buf_seqno(skb);
1010 } else if (++l->stale_count > 100) {
1011 link_retransmit_failure(l, skb);
1012 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1015 /* Move forward to where retransmission should start */
1016 skb_queue_walk(&l->transmq, skb) {
1017 if (!less(buf_seqno(skb), from))
1021 skb_queue_walk_from(&l->transmq, skb) {
1022 if (more(buf_seqno(skb), to))
1025 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1028 hdr = buf_msg(_skb);
1029 msg_set_ack(hdr, ack);
1030 msg_set_bcast_ack(hdr, bc_ack);
1031 _skb->priority = TC_PRIO_CONTROL;
1032 __skb_queue_tail(xmitq, _skb);
1033 l->stats.retransmitted++;
1038 /* tipc_data_input - deliver data and name distr msgs to upper layer
1040 * Consumes buffer if message is of right type
1041 * Node lock must be held
1043 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1044 struct sk_buff_head *inputq)
1046 switch (msg_user(buf_msg(skb))) {
1047 case TIPC_LOW_IMPORTANCE:
1048 case TIPC_MEDIUM_IMPORTANCE:
1049 case TIPC_HIGH_IMPORTANCE:
1050 case TIPC_CRITICAL_IMPORTANCE:
1052 skb_queue_tail(inputq, skb);
1054 case NAME_DISTRIBUTOR:
1055 l->bc_rcvlink->state = LINK_ESTABLISHED;
1056 skb_queue_tail(l->namedq, skb);
1059 case TUNNEL_PROTOCOL:
1060 case MSG_FRAGMENTER:
1061 case BCAST_PROTOCOL:
1064 pr_warn("Dropping received illegal msg type\n");
1070 /* tipc_link_input - process packet that has passed link protocol check
1074 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1075 struct sk_buff_head *inputq)
1077 struct tipc_msg *hdr = buf_msg(skb);
1078 struct sk_buff **reasm_skb = &l->reasm_buf;
1079 struct sk_buff *iskb;
1080 struct sk_buff_head tmpq;
1081 int usr = msg_user(hdr);
1086 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1087 if (msg_type(hdr) == SYNCH_MSG) {
1088 __skb_queue_purge(&l->deferdq);
1091 if (!tipc_msg_extract(skb, &iskb, &ipos))
1096 if (less(msg_seqno(hdr), l->drop_point))
1098 if (tipc_data_input(l, skb, inputq))
1100 usr = msg_user(hdr);
1101 reasm_skb = &l->failover_reasm_skb;
1104 if (usr == MSG_BUNDLER) {
1105 skb_queue_head_init(&tmpq);
1106 l->stats.recv_bundles++;
1107 l->stats.recv_bundled += msg_msgcnt(hdr);
1108 while (tipc_msg_extract(skb, &iskb, &pos))
1109 tipc_data_input(l, iskb, &tmpq);
1110 tipc_skb_queue_splice_tail(&tmpq, inputq);
1112 } else if (usr == MSG_FRAGMENTER) {
1113 l->stats.recv_fragments++;
1114 if (tipc_buf_append(reasm_skb, &skb)) {
1115 l->stats.recv_fragmented++;
1116 tipc_data_input(l, skb, inputq);
1117 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1118 pr_warn_ratelimited("Unable to build fragment list\n");
1119 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1122 } else if (usr == BCAST_PROTOCOL) {
1123 tipc_bcast_lock(l->net);
1124 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1125 tipc_bcast_unlock(l->net);
1132 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1134 bool released = false;
1135 struct sk_buff *skb, *tmp;
1137 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1138 if (more(buf_seqno(skb), acked))
1140 __skb_unlink(skb, &l->transmq);
1147 /* tipc_link_build_ack_msg: prepare link acknowledge message for transmission
1149 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1150 * risk of ack storms towards the sender
1152 int tipc_link_build_ack_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1157 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1158 if (link_is_bc_rcvlink(l)) {
1159 if (((l->rcv_nxt ^ link_own_addr(l)) & 0xf) != 0xf)
1162 return TIPC_LINK_SND_BC_ACK;
1167 l->stats.sent_acks++;
1168 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1172 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1174 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1176 int mtyp = RESET_MSG;
1178 if (l->state == LINK_ESTABLISHING)
1179 mtyp = ACTIVATE_MSG;
1181 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1184 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1186 static void tipc_link_build_nack_msg(struct tipc_link *l,
1187 struct sk_buff_head *xmitq)
1189 u32 def_cnt = ++l->stats.deferred_recv;
1191 if (link_is_bc_rcvlink(l))
1194 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1195 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1198 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1199 * @l: the link that should handle the message
1201 * @xmitq: queue to place packets to be sent after this call
1203 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1204 struct sk_buff_head *xmitq)
1206 struct sk_buff_head *defq = &l->deferdq;
1207 struct tipc_msg *hdr;
1208 u16 seqno, rcv_nxt, win_lim;
1213 seqno = msg_seqno(hdr);
1214 rcv_nxt = l->rcv_nxt;
1215 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1217 /* Verify and update link state */
1218 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1219 return tipc_link_proto_rcv(l, skb, xmitq);
1221 if (unlikely(!link_is_up(l))) {
1222 if (l->state == LINK_ESTABLISHING)
1223 rc = TIPC_LINK_UP_EVT;
1227 /* Don't send probe at next timeout expiration */
1228 l->silent_intv_cnt = 0;
1230 /* Drop if outside receive window */
1231 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1232 l->stats.duplicates++;
1236 /* Forward queues and wake up waiting users */
1237 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1238 tipc_link_advance_backlog(l, xmitq);
1239 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1240 link_prepare_wakeup(l);
1243 /* Defer delivery if sequence gap */
1244 if (unlikely(seqno != rcv_nxt)) {
1245 __tipc_skb_queue_sorted(defq, seqno, skb);
1246 tipc_link_build_nack_msg(l, xmitq);
1250 /* Deliver packet */
1252 l->stats.recv_info++;
1253 if (!tipc_data_input(l, skb, l->inputq))
1254 rc |= tipc_link_input(l, skb, l->inputq);
1255 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1256 rc |= tipc_link_build_ack_msg(l, xmitq);
1257 if (unlikely(rc & ~TIPC_LINK_SND_BC_ACK))
1259 } while ((skb = __skb_dequeue(defq)));
1267 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1268 u16 rcvgap, int tolerance, int priority,
1269 struct sk_buff_head *xmitq)
1271 struct sk_buff *skb = NULL;
1272 struct tipc_msg *hdr = l->pmsg;
1273 bool node_up = link_is_up(l->bc_rcvlink);
1275 /* Don't send protocol message during reset or link failover */
1276 if (tipc_link_is_blocked(l))
1279 msg_set_type(hdr, mtyp);
1280 msg_set_net_plane(hdr, l->net_plane);
1281 msg_set_next_sent(hdr, l->snd_nxt);
1282 msg_set_ack(hdr, l->rcv_nxt - 1);
1283 msg_set_bcast_ack(hdr, l->bc_rcvlink->rcv_nxt - 1);
1284 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1285 msg_set_link_tolerance(hdr, tolerance);
1286 msg_set_linkprio(hdr, priority);
1287 msg_set_redundant_link(hdr, node_up);
1288 msg_set_seq_gap(hdr, 0);
1290 /* Compatibility: created msg must not be in sequence with pkt flow */
1291 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1293 if (mtyp == STATE_MSG) {
1294 if (!tipc_link_is_up(l))
1297 /* Override rcvgap if there are packets in deferred queue */
1298 if (!skb_queue_empty(&l->deferdq))
1299 rcvgap = buf_seqno(skb_peek(&l->deferdq)) - l->rcv_nxt;
1301 msg_set_seq_gap(hdr, rcvgap);
1302 l->stats.sent_nacks++;
1304 msg_set_probe(hdr, probe);
1306 l->stats.sent_probes++;
1307 l->stats.sent_states++;
1310 /* RESET_MSG or ACTIVATE_MSG */
1311 msg_set_max_pkt(hdr, l->advertised_mtu);
1312 msg_set_ack(hdr, l->rcv_nxt - 1);
1313 msg_set_next_sent(hdr, 1);
1315 skb = tipc_buf_acquire(msg_size(hdr));
1318 skb_copy_to_linear_data(skb, hdr, msg_size(hdr));
1319 skb->priority = TC_PRIO_CONTROL;
1320 __skb_queue_tail(xmitq, skb);
1323 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1324 * with contents of the link's transmit and backlog queues.
1326 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1327 int mtyp, struct sk_buff_head *xmitq)
1329 struct sk_buff *skb, *tnlskb;
1330 struct tipc_msg *hdr, tnlhdr;
1331 struct sk_buff_head *queue = &l->transmq;
1332 struct sk_buff_head tmpxq, tnlq;
1333 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1338 skb_queue_head_init(&tnlq);
1339 skb_queue_head_init(&tmpxq);
1341 /* At least one packet required for safe algorithm => add dummy */
1342 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1343 BASIC_H_SIZE, 0, l->addr, link_own_addr(l),
1344 0, 0, TIPC_ERR_NO_PORT);
1346 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1349 skb_queue_tail(&tnlq, skb);
1350 tipc_link_xmit(l, &tnlq, &tmpxq);
1351 __skb_queue_purge(&tmpxq);
1353 /* Initialize reusable tunnel packet header */
1354 tipc_msg_init(link_own_addr(l), &tnlhdr, TUNNEL_PROTOCOL,
1355 mtyp, INT_H_SIZE, l->addr);
1356 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1357 msg_set_msgcnt(&tnlhdr, pktcnt);
1358 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1360 /* Wrap each packet into a tunnel packet */
1361 skb_queue_walk(queue, skb) {
1363 if (queue == &l->backlogq)
1364 msg_set_seqno(hdr, seqno++);
1365 pktlen = msg_size(hdr);
1366 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1367 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE);
1369 pr_warn("%sunable to send packet\n", link_co_err);
1372 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1373 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1374 __skb_queue_tail(&tnlq, tnlskb);
1376 if (queue != &l->backlogq) {
1377 queue = &l->backlogq;
1381 tipc_link_xmit(tnl, &tnlq, xmitq);
1383 if (mtyp == FAILOVER_MSG) {
1384 tnl->drop_point = l->rcv_nxt;
1385 tnl->failover_reasm_skb = l->reasm_buf;
1386 l->reasm_buf = NULL;
1390 /* tipc_link_proto_rcv(): receive link level protocol message :
1391 * Note that network plane id propagates through the network, and may
1392 * change at any time. The node with lowest numerical id determines
1395 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1396 struct sk_buff_head *xmitq)
1398 struct tipc_msg *hdr = buf_msg(skb);
1400 u16 ack = msg_ack(hdr);
1401 u16 gap = msg_seq_gap(hdr);
1402 u16 peers_snd_nxt = msg_next_sent(hdr);
1403 u16 peers_tol = msg_link_tolerance(hdr);
1404 u16 peers_prio = msg_linkprio(hdr);
1405 u16 rcv_nxt = l->rcv_nxt;
1406 int mtyp = msg_type(hdr);
1410 if (tipc_link_is_blocked(l) || !xmitq)
1413 if (link_own_addr(l) > msg_prevnode(hdr))
1414 l->net_plane = msg_net_plane(hdr);
1419 /* Ignore duplicate RESET with old session number */
1420 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1421 (l->peer_session != WILDCARD_SESSION))
1429 /* Complete own link name with peer's interface name */
1430 if_name = strrchr(l->name, ':') + 1;
1431 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1433 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1435 strncpy(if_name, msg_data(hdr), TIPC_MAX_IF_NAME);
1437 /* Update own tolerance if peer indicates a non-zero value */
1438 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1439 l->tolerance = peers_tol;
1441 /* Update own priority if peer's priority is higher */
1442 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1443 l->priority = peers_prio;
1445 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1446 if ((mtyp == RESET_MSG) || !link_is_up(l))
1447 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1449 /* ACTIVATE_MSG takes up link if it was already locally reset */
1450 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1451 rc = TIPC_LINK_UP_EVT;
1453 l->peer_session = msg_session(hdr);
1454 l->peer_bearer_id = msg_bearer_id(hdr);
1455 if (l->mtu > msg_max_pkt(hdr))
1456 l->mtu = msg_max_pkt(hdr);
1461 /* Update own tolerance if peer indicates a non-zero value */
1462 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1463 l->tolerance = peers_tol;
1465 if (peers_prio && in_range(peers_prio, TIPC_MIN_LINK_PRI,
1466 TIPC_MAX_LINK_PRI)) {
1467 l->priority = peers_prio;
1468 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1471 l->silent_intv_cnt = 0;
1472 l->stats.recv_states++;
1474 l->stats.recv_probes++;
1476 if (!link_is_up(l)) {
1477 if (l->state == LINK_ESTABLISHING)
1478 rc = TIPC_LINK_UP_EVT;
1482 /* Send NACK if peer has sent pkts we haven't received yet */
1483 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1484 rcvgap = peers_snd_nxt - l->rcv_nxt;
1485 if (rcvgap || (msg_probe(hdr)))
1486 tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1488 tipc_link_release_pkts(l, ack);
1490 /* If NACK, retransmit will now start at right position */
1492 rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
1493 l->stats.recv_nacks++;
1496 tipc_link_advance_backlog(l, xmitq);
1497 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1498 link_prepare_wakeup(l);
1505 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1507 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1509 struct sk_buff_head *xmitq)
1511 struct sk_buff *skb;
1512 struct tipc_msg *hdr;
1513 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1514 u16 ack = l->rcv_nxt - 1;
1515 u16 gap_to = peers_snd_nxt - 1;
1517 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1518 0, l->addr, link_own_addr(l), 0, 0, 0);
1522 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1523 msg_set_bcast_ack(hdr, ack);
1524 msg_set_bcgap_after(hdr, ack);
1526 gap_to = buf_seqno(dfrd_skb) - 1;
1527 msg_set_bcgap_to(hdr, gap_to);
1528 msg_set_non_seq(hdr, bcast);
1529 __skb_queue_tail(xmitq, skb);
1533 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1535 * Give a newly added peer node the sequence number where it should
1536 * start receiving and acking broadcast packets.
1538 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1539 struct sk_buff_head *xmitq)
1541 struct sk_buff_head list;
1543 __skb_queue_head_init(&list);
1544 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1546 tipc_link_xmit(l, &list, xmitq);
1549 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1551 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1553 int mtyp = msg_type(hdr);
1554 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1559 if (msg_user(hdr) == BCAST_PROTOCOL) {
1560 l->rcv_nxt = peers_snd_nxt;
1561 l->state = LINK_ESTABLISHED;
1565 if (l->peer_caps & TIPC_BCAST_SYNCH)
1568 if (msg_peer_node_is_up(hdr))
1571 /* Compatibility: accept older, less safe initial synch data */
1572 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1573 l->rcv_nxt = peers_snd_nxt;
1576 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1578 void tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1579 struct sk_buff_head *xmitq)
1581 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1586 if (!msg_peer_node_is_up(hdr))
1589 l->bc_peer_is_up = true;
1591 /* Ignore if peers_snd_nxt goes beyond receive window */
1592 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1595 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1596 l->nack_state = BC_NACK_SND_CONDITIONAL;
1600 /* Don't NACK if one was recently sent or peeked */
1601 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1602 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1606 /* Conditionally delay NACK sending until next synch rcv */
1607 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1608 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1609 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1613 /* Send NACK now but suppress next one */
1614 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1615 l->nack_state = BC_NACK_SND_SUPPRESS;
1618 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1619 struct sk_buff_head *xmitq)
1621 struct sk_buff *skb, *tmp;
1622 struct tipc_link *snd_l = l->bc_sndlink;
1624 if (!link_is_up(l) || !l->bc_peer_is_up)
1627 if (!more(acked, l->acked))
1630 /* Skip over packets peer has already acked */
1631 skb_queue_walk(&snd_l->transmq, skb) {
1632 if (more(buf_seqno(skb), l->acked))
1636 /* Update/release the packets peer is acking now */
1637 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1638 if (more(buf_seqno(skb), acked))
1640 if (!--TIPC_SKB_CB(skb)->ackers) {
1641 __skb_unlink(skb, &snd_l->transmq);
1646 tipc_link_advance_backlog(snd_l, xmitq);
1647 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1648 link_prepare_wakeup(snd_l);
1651 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1653 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1654 struct sk_buff_head *xmitq)
1656 struct tipc_msg *hdr = buf_msg(skb);
1657 u32 dnode = msg_destnode(hdr);
1658 int mtyp = msg_type(hdr);
1659 u16 acked = msg_bcast_ack(hdr);
1660 u16 from = acked + 1;
1661 u16 to = msg_bcgap_to(hdr);
1662 u16 peers_snd_nxt = to + 1;
1667 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1670 if (mtyp != STATE_MSG)
1673 if (dnode == link_own_addr(l)) {
1674 tipc_link_bc_ack_rcv(l, acked, xmitq);
1675 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1676 l->stats.recv_nacks++;
1680 /* Msg for other node => suppress own NACK at next sync if applicable */
1681 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1682 l->nack_state = BC_NACK_SND_SUPPRESS;
1687 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1689 int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1692 l->backlog[TIPC_LOW_IMPORTANCE].limit = win / 2;
1693 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = win;
1694 l->backlog[TIPC_HIGH_IMPORTANCE].limit = win / 2 * 3;
1695 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = win * 2;
1696 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1700 * link_reset_stats - reset link statistics
1701 * @l: pointer to link
1703 void tipc_link_reset_stats(struct tipc_link *l)
1705 memset(&l->stats, 0, sizeof(l->stats));
1706 if (!link_is_bc_sndlink(l)) {
1707 l->stats.sent_info = l->snd_nxt;
1708 l->stats.recv_info = l->rcv_nxt;
1712 static void link_print(struct tipc_link *l, const char *str)
1714 struct sk_buff *hskb = skb_peek(&l->transmq);
1715 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1716 u16 tail = l->snd_nxt - 1;
1718 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1719 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1720 skb_queue_len(&l->transmq), head, tail,
1721 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1724 /* Parse and validate nested (link) properties valid for media, bearer and link
1726 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1730 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1731 tipc_nl_prop_policy);
1735 if (props[TIPC_NLA_PROP_PRIO]) {
1738 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1739 if (prio > TIPC_MAX_LINK_PRI)
1743 if (props[TIPC_NLA_PROP_TOL]) {
1746 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1747 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1751 if (props[TIPC_NLA_PROP_WIN]) {
1754 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1755 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1762 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1765 struct nlattr *stats;
1772 struct nla_map map[] = {
1773 {TIPC_NLA_STATS_RX_INFO, s->recv_info},
1774 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1775 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1776 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1777 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1778 {TIPC_NLA_STATS_TX_INFO, s->sent_info},
1779 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1780 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1781 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1782 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1783 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1784 s->msg_length_counts : 1},
1785 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1786 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1787 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1788 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1789 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1790 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1791 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1792 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1793 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1794 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1795 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1796 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1797 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1798 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1799 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1800 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1801 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1802 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1803 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1804 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1805 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1806 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1807 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1810 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1814 for (i = 0; i < ARRAY_SIZE(map); i++)
1815 if (nla_put_u32(skb, map[i].key, map[i].val))
1818 nla_nest_end(skb, stats);
1822 nla_nest_cancel(skb, stats);
1827 /* Caller should hold appropriate locks to protect the link */
1828 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1829 struct tipc_link *link, int nlflags)
1833 struct nlattr *attrs;
1834 struct nlattr *prop;
1835 struct tipc_net *tn = net_generic(net, tipc_net_id);
1837 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1838 nlflags, TIPC_NL_LINK_GET);
1842 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1846 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1848 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1849 tipc_cluster_mask(tn->own_addr)))
1851 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1853 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->rcv_nxt))
1855 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->snd_nxt))
1858 if (tipc_link_is_up(link))
1859 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1862 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
1865 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1868 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1870 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
1872 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
1875 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1877 nla_nest_end(msg->skb, prop);
1879 err = __tipc_nl_add_stats(msg->skb, &link->stats);
1883 nla_nest_end(msg->skb, attrs);
1884 genlmsg_end(msg->skb, hdr);
1889 nla_nest_cancel(msg->skb, prop);
1891 nla_nest_cancel(msg->skb, attrs);
1893 genlmsg_cancel(msg->skb, hdr);
1898 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
1899 struct tipc_stats *stats)
1902 struct nlattr *nest;
1909 struct nla_map map[] = {
1910 {TIPC_NLA_STATS_RX_INFO, stats->recv_info},
1911 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
1912 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
1913 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
1914 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
1915 {TIPC_NLA_STATS_TX_INFO, stats->sent_info},
1916 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
1917 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
1918 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
1919 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
1920 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
1921 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
1922 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
1923 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
1924 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
1925 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
1926 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
1927 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
1928 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
1929 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
1932 nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1936 for (i = 0; i < ARRAY_SIZE(map); i++)
1937 if (nla_put_u32(skb, map[i].key, map[i].val))
1940 nla_nest_end(skb, nest);
1944 nla_nest_cancel(skb, nest);
1949 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
1953 struct nlattr *attrs;
1954 struct nlattr *prop;
1955 struct tipc_net *tn = net_generic(net, tipc_net_id);
1956 struct tipc_link *bcl = tn->bcl;
1961 tipc_bcast_lock(net);
1963 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1964 NLM_F_MULTI, TIPC_NL_LINK_GET);
1968 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1972 /* The broadcast link is always up */
1973 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1976 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
1978 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
1980 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, bcl->rcv_nxt))
1982 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, bcl->snd_nxt))
1985 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1988 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
1990 nla_nest_end(msg->skb, prop);
1992 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
1996 tipc_bcast_unlock(net);
1997 nla_nest_end(msg->skb, attrs);
1998 genlmsg_end(msg->skb, hdr);
2003 nla_nest_cancel(msg->skb, prop);
2005 nla_nest_cancel(msg->skb, attrs);
2007 tipc_bcast_unlock(net);
2008 genlmsg_cancel(msg->skb, hdr);
2013 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2014 struct sk_buff_head *xmitq)
2017 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, tol, 0, xmitq);
2020 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2021 struct sk_buff_head *xmitq)
2024 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, prio, xmitq);
2027 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2029 l->abort_limit = limit;