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];
126 struct tipc_media_addr *media_addr;
129 /* Management and link supervision data */
134 unsigned long keepalive_intv;
141 unchar hdr[INT_H_SIZE];
142 unchar body[TIPC_MAX_IF_NAME];
144 struct tipc_msg *pmsg;
150 struct sk_buff *failover_reasm_skb;
152 /* Max packet negotiation */
157 struct sk_buff_head transmq;
158 struct sk_buff_head backlogq;
171 struct sk_buff_head deferdq;
172 struct sk_buff_head *inputq;
173 struct sk_buff_head *namedq;
175 /* Congestion handling */
176 struct sk_buff_head wakeupq;
178 /* Fragmentation/reassembly */
179 struct sk_buff *reasm_buf;
184 struct tipc_link *bc_rcvlink;
185 struct tipc_link *bc_sndlink;
190 struct tipc_stats stats;
194 * Error message prefixes
196 static const char *link_co_err = "Link tunneling error, ";
197 static const char *link_rst_msg = "Resetting link ";
199 /* Properties valid for media, bearar and link */
200 static const struct nla_policy tipc_nl_prop_policy[TIPC_NLA_PROP_MAX + 1] = {
201 [TIPC_NLA_PROP_UNSPEC] = { .type = NLA_UNSPEC },
202 [TIPC_NLA_PROP_PRIO] = { .type = NLA_U32 },
203 [TIPC_NLA_PROP_TOL] = { .type = NLA_U32 },
204 [TIPC_NLA_PROP_WIN] = { .type = NLA_U32 }
207 /* Send states for broadcast NACKs
210 BC_NACK_SND_CONDITIONAL,
211 BC_NACK_SND_UNCONDITIONAL,
212 BC_NACK_SND_SUPPRESS,
216 * Interval between NACKs when packets arrive out of order
218 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
220 * Out-of-range value for link session numbers
222 #define WILDCARD_SESSION 0x10000
227 LINK_ESTABLISHED = 0xe,
228 LINK_ESTABLISHING = 0xe << 4,
229 LINK_RESET = 0x1 << 8,
230 LINK_RESETTING = 0x2 << 12,
231 LINK_PEER_RESET = 0xd << 16,
232 LINK_FAILINGOVER = 0xf << 20,
233 LINK_SYNCHING = 0xc << 24
236 /* Link FSM state checking routines
238 static int link_is_up(struct tipc_link *l)
240 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
243 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
244 struct sk_buff_head *xmitq);
245 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
246 u16 rcvgap, int tolerance, int priority,
247 struct sk_buff_head *xmitq);
248 static void link_print(struct tipc_link *l, const char *str);
249 static void tipc_link_build_nack_msg(struct tipc_link *l,
250 struct sk_buff_head *xmitq);
251 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
252 struct sk_buff_head *xmitq);
253 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
256 * Simple non-static link routines (i.e. referenced outside this file)
258 bool tipc_link_is_up(struct tipc_link *l)
260 return link_is_up(l);
263 bool tipc_link_peer_is_down(struct tipc_link *l)
265 return l->state == LINK_PEER_RESET;
268 bool tipc_link_is_reset(struct tipc_link *l)
270 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
273 bool tipc_link_is_establishing(struct tipc_link *l)
275 return l->state == LINK_ESTABLISHING;
278 bool tipc_link_is_synching(struct tipc_link *l)
280 return l->state == LINK_SYNCHING;
283 bool tipc_link_is_failingover(struct tipc_link *l)
285 return l->state == LINK_FAILINGOVER;
288 bool tipc_link_is_blocked(struct tipc_link *l)
290 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
293 static bool link_is_bc_sndlink(struct tipc_link *l)
295 return !l->bc_sndlink;
298 static bool link_is_bc_rcvlink(struct tipc_link *l)
300 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
303 int tipc_link_is_active(struct tipc_link *l)
308 void tipc_link_set_active(struct tipc_link *l, bool active)
313 u32 tipc_link_id(struct tipc_link *l)
315 return l->peer_bearer_id << 16 | l->bearer_id;
318 int tipc_link_window(struct tipc_link *l)
323 int tipc_link_prio(struct tipc_link *l)
328 unsigned long tipc_link_tolerance(struct tipc_link *l)
333 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
338 char tipc_link_plane(struct tipc_link *l)
343 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
344 struct tipc_link *uc_l,
345 struct sk_buff_head *xmitq)
347 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
350 rcv_l->acked = snd_l->snd_nxt - 1;
351 snd_l->state = LINK_ESTABLISHED;
352 tipc_link_build_bc_init_msg(uc_l, xmitq);
355 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
356 struct tipc_link *rcv_l,
357 struct sk_buff_head *xmitq)
359 u16 ack = snd_l->snd_nxt - 1;
362 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
363 tipc_link_reset(rcv_l);
364 rcv_l->state = LINK_RESET;
365 if (!snd_l->ackers) {
366 tipc_link_reset(snd_l);
367 snd_l->state = LINK_RESET;
368 __skb_queue_purge(xmitq);
372 int tipc_link_bc_peers(struct tipc_link *l)
377 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
382 int tipc_link_mtu(struct tipc_link *l)
387 u16 tipc_link_rcv_nxt(struct tipc_link *l)
392 u16 tipc_link_acked(struct tipc_link *l)
397 char *tipc_link_name(struct tipc_link *l)
402 static u32 link_own_addr(struct tipc_link *l)
404 return msg_prevnode(l->pmsg);
407 void tipc_link_reinit(struct tipc_link *l, u32 addr)
409 msg_set_prevnode(l->pmsg, addr);
413 * tipc_link_create - create a new link
414 * @n: pointer to associated node
415 * @if_name: associated interface name
416 * @bearer_id: id (index) of associated bearer
417 * @tolerance: link tolerance to be used by link
418 * @net_plane: network plane (A,B,c..) this link belongs to
419 * @mtu: mtu to be advertised by link
420 * @priority: priority to be used by link
421 * @window: send window to be used by link
422 * @session: session to be used by link
423 * @ownnode: identity of own node
424 * @peer: node id of peer node
425 * @peer_caps: bitmap describing peer node capabilities
426 * @bc_sndlink: the namespace global link used for broadcast sending
427 * @bc_rcvlink: the peer specific link used for broadcast reception
428 * @inputq: queue to put messages ready for delivery
429 * @namedq: queue to put binding table update messages ready for delivery
430 * @link: return value, pointer to put the created link
432 * Returns true if link was created, otherwise false
434 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
435 int tolerance, char net_plane, u32 mtu, int priority,
436 int window, u32 session, u32 ownnode, u32 peer,
438 struct tipc_link *bc_sndlink,
439 struct tipc_link *bc_rcvlink,
440 struct sk_buff_head *inputq,
441 struct sk_buff_head *namedq,
442 struct tipc_link **link)
445 struct tipc_msg *hdr;
447 l = kzalloc(sizeof(*l), GFP_ATOMIC);
451 l->pmsg = (struct tipc_msg *)&l->proto_msg;
453 tipc_msg_init(ownnode, hdr, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, peer);
454 msg_set_size(hdr, sizeof(l->proto_msg));
455 msg_set_session(hdr, session);
456 msg_set_bearer_id(hdr, l->bearer_id);
458 /* Note: peer i/f name is completed by reset/activate message */
459 sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
460 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
461 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
462 strcpy((char *)msg_data(hdr), if_name);
465 l->peer_caps = peer_caps;
467 l->peer_session = WILDCARD_SESSION;
468 l->bearer_id = bearer_id;
469 l->tolerance = tolerance;
470 l->net_plane = net_plane;
471 l->advertised_mtu = mtu;
473 l->priority = priority;
474 tipc_link_set_queue_limits(l, window);
476 l->bc_sndlink = bc_sndlink;
477 l->bc_rcvlink = bc_rcvlink;
480 l->state = LINK_RESETTING;
481 __skb_queue_head_init(&l->transmq);
482 __skb_queue_head_init(&l->backlogq);
483 __skb_queue_head_init(&l->deferdq);
484 skb_queue_head_init(&l->wakeupq);
485 skb_queue_head_init(l->inputq);
490 * tipc_link_bc_create - create new link to be used for broadcast
491 * @n: pointer to associated node
492 * @mtu: mtu to be used
493 * @window: send window to be used
494 * @inputq: queue to put messages ready for delivery
495 * @namedq: queue to put binding table update messages ready for delivery
496 * @link: return value, pointer to put the created link
498 * Returns true if link was created, otherwise false
500 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
501 int mtu, int window, u16 peer_caps,
502 struct sk_buff_head *inputq,
503 struct sk_buff_head *namedq,
504 struct tipc_link *bc_sndlink,
505 struct tipc_link **link)
509 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
510 0, ownnode, peer, peer_caps, bc_sndlink,
511 NULL, inputq, namedq, link))
515 strcpy(l->name, tipc_bclink_name);
517 l->state = LINK_RESET;
521 /* Broadcast send link is always up */
522 if (link_is_bc_sndlink(l))
523 l->state = LINK_ESTABLISHED;
529 * tipc_link_fsm_evt - link finite state machine
530 * @l: pointer to link
531 * @evt: state machine event to be processed
533 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
540 case LINK_PEER_RESET_EVT:
541 l->state = LINK_PEER_RESET;
544 l->state = LINK_RESET;
546 case LINK_FAILURE_EVT:
547 case LINK_FAILOVER_BEGIN_EVT:
548 case LINK_ESTABLISH_EVT:
549 case LINK_FAILOVER_END_EVT:
550 case LINK_SYNCH_BEGIN_EVT:
551 case LINK_SYNCH_END_EVT:
558 case LINK_PEER_RESET_EVT:
559 l->state = LINK_ESTABLISHING;
561 case LINK_FAILOVER_BEGIN_EVT:
562 l->state = LINK_FAILINGOVER;
563 case LINK_FAILURE_EVT:
565 case LINK_ESTABLISH_EVT:
566 case LINK_FAILOVER_END_EVT:
568 case LINK_SYNCH_BEGIN_EVT:
569 case LINK_SYNCH_END_EVT:
574 case LINK_PEER_RESET:
577 l->state = LINK_ESTABLISHING;
579 case LINK_PEER_RESET_EVT:
580 case LINK_ESTABLISH_EVT:
581 case LINK_FAILURE_EVT:
583 case LINK_SYNCH_BEGIN_EVT:
584 case LINK_SYNCH_END_EVT:
585 case LINK_FAILOVER_BEGIN_EVT:
586 case LINK_FAILOVER_END_EVT:
591 case LINK_FAILINGOVER:
593 case LINK_FAILOVER_END_EVT:
594 l->state = LINK_RESET;
596 case LINK_PEER_RESET_EVT:
598 case LINK_ESTABLISH_EVT:
599 case LINK_FAILURE_EVT:
601 case LINK_FAILOVER_BEGIN_EVT:
602 case LINK_SYNCH_BEGIN_EVT:
603 case LINK_SYNCH_END_EVT:
608 case LINK_ESTABLISHING:
610 case LINK_ESTABLISH_EVT:
611 l->state = LINK_ESTABLISHED;
613 case LINK_FAILOVER_BEGIN_EVT:
614 l->state = LINK_FAILINGOVER;
617 l->state = LINK_RESET;
619 case LINK_FAILURE_EVT:
620 case LINK_PEER_RESET_EVT:
621 case LINK_SYNCH_BEGIN_EVT:
622 case LINK_FAILOVER_END_EVT:
624 case LINK_SYNCH_END_EVT:
629 case LINK_ESTABLISHED:
631 case LINK_PEER_RESET_EVT:
632 l->state = LINK_PEER_RESET;
633 rc |= TIPC_LINK_DOWN_EVT;
635 case LINK_FAILURE_EVT:
636 l->state = LINK_RESETTING;
637 rc |= TIPC_LINK_DOWN_EVT;
640 l->state = LINK_RESET;
642 case LINK_ESTABLISH_EVT:
643 case LINK_SYNCH_END_EVT:
645 case LINK_SYNCH_BEGIN_EVT:
646 l->state = LINK_SYNCHING;
648 case LINK_FAILOVER_BEGIN_EVT:
649 case LINK_FAILOVER_END_EVT:
656 case LINK_PEER_RESET_EVT:
657 l->state = LINK_PEER_RESET;
658 rc |= TIPC_LINK_DOWN_EVT;
660 case LINK_FAILURE_EVT:
661 l->state = LINK_RESETTING;
662 rc |= TIPC_LINK_DOWN_EVT;
665 l->state = LINK_RESET;
667 case LINK_ESTABLISH_EVT:
668 case LINK_SYNCH_BEGIN_EVT:
670 case LINK_SYNCH_END_EVT:
671 l->state = LINK_ESTABLISHED;
673 case LINK_FAILOVER_BEGIN_EVT:
674 case LINK_FAILOVER_END_EVT:
680 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
684 pr_err("Illegal FSM event %x in state %x on link %s\n",
685 evt, l->state, l->name);
689 /* link_profile_stats - update statistical profiling of traffic
691 static void link_profile_stats(struct tipc_link *l)
694 struct tipc_msg *msg;
697 /* Update counters used in statistical profiling of send traffic */
698 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
699 l->stats.queue_sz_counts++;
701 skb = skb_peek(&l->transmq);
705 length = msg_size(msg);
707 if (msg_user(msg) == MSG_FRAGMENTER) {
708 if (msg_type(msg) != FIRST_FRAGMENT)
710 length = msg_size(msg_get_wrapped(msg));
712 l->stats.msg_lengths_total += length;
713 l->stats.msg_length_counts++;
715 l->stats.msg_length_profile[0]++;
716 else if (length <= 256)
717 l->stats.msg_length_profile[1]++;
718 else if (length <= 1024)
719 l->stats.msg_length_profile[2]++;
720 else if (length <= 4096)
721 l->stats.msg_length_profile[3]++;
722 else if (length <= 16384)
723 l->stats.msg_length_profile[4]++;
724 else if (length <= 32768)
725 l->stats.msg_length_profile[5]++;
727 l->stats.msg_length_profile[6]++;
730 /* tipc_link_timeout - perform periodic task as instructed from node timeout
732 /* tipc_link_timeout - perform periodic task as instructed from node timeout
734 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
737 int mtyp = STATE_MSG;
740 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
741 u16 bc_acked = l->bc_rcvlink->acked;
742 bool bc_up = link_is_up(l->bc_rcvlink);
744 link_profile_stats(l);
747 case LINK_ESTABLISHED:
749 if (!l->silent_intv_cnt) {
750 if (bc_up && (bc_acked != bc_snt))
752 } else if (l->silent_intv_cnt <= l->abort_limit) {
756 rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
758 l->silent_intv_cnt++;
764 case LINK_ESTABLISHING:
768 case LINK_PEER_RESET:
770 case LINK_FAILINGOVER:
777 tipc_link_build_proto_msg(l, mtyp, prb, 0, 0, 0, xmitq);
783 * link_schedule_user - schedule a message sender for wakeup after congestion
784 * @link: congested link
785 * @list: message that was attempted sent
786 * Create pseudo msg to send back to user when congestion abates
787 * Does not consume buffer list
789 static int link_schedule_user(struct tipc_link *link, struct sk_buff_head *list)
791 struct tipc_msg *msg = buf_msg(skb_peek(list));
792 int imp = msg_importance(msg);
793 u32 oport = msg_origport(msg);
794 u32 addr = link_own_addr(link);
797 /* This really cannot happen... */
798 if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
799 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
802 /* Non-blocking sender: */
803 if (TIPC_SKB_CB(skb_peek(list))->wakeup_pending)
806 /* Create and schedule wakeup pseudo message */
807 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
808 addr, addr, oport, 0, 0);
811 TIPC_SKB_CB(skb)->chain_sz = skb_queue_len(list);
812 TIPC_SKB_CB(skb)->chain_imp = imp;
813 skb_queue_tail(&link->wakeupq, skb);
814 link->stats.link_congs++;
819 * link_prepare_wakeup - prepare users for wakeup after congestion
820 * @link: congested link
821 * Move a number of waiting users, as permitted by available space in
822 * the send queue, from link wait queue to node wait queue for wakeup
824 void link_prepare_wakeup(struct tipc_link *l)
826 int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
828 struct sk_buff *skb, *tmp;
830 skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
831 imp = TIPC_SKB_CB(skb)->chain_imp;
832 lim = l->window + l->backlog[imp].limit;
833 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
834 if ((pnd[imp] + l->backlog[imp].len) >= lim)
836 skb_unlink(skb, &l->wakeupq);
837 skb_queue_tail(l->inputq, skb);
841 void tipc_link_reset(struct tipc_link *l)
843 /* Link is down, accept any session */
844 l->peer_session = WILDCARD_SESSION;
846 /* If peer is up, it only accepts an incremented session number */
847 msg_set_session(l->pmsg, msg_session(l->pmsg) + 1);
849 /* Prepare for renewed mtu size negotiation */
850 l->mtu = l->advertised_mtu;
852 /* Clean up all queues and counters: */
853 __skb_queue_purge(&l->transmq);
854 __skb_queue_purge(&l->deferdq);
855 skb_queue_splice_init(&l->wakeupq, l->inputq);
856 __skb_queue_purge(&l->backlogq);
857 l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
858 l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
859 l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
860 l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
861 l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
862 kfree_skb(l->reasm_buf);
863 kfree_skb(l->failover_reasm_skb);
865 l->failover_reasm_skb = NULL;
870 l->silent_intv_cnt = 0;
871 l->stats.recv_info = 0;
873 l->bc_peer_is_up = false;
874 tipc_link_reset_stats(l);
878 * tipc_link_xmit(): enqueue buffer list according to queue situation
880 * @list: chain of buffers containing message
881 * @xmitq: returned list of packets to be sent by caller
883 * Consumes the buffer chain, except when returning -ELINKCONG,
884 * since the caller then may want to make more send attempts.
885 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
886 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
888 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
889 struct sk_buff_head *xmitq)
891 struct tipc_msg *hdr = buf_msg(skb_peek(list));
892 unsigned int maxwin = l->window;
893 unsigned int i, imp = msg_importance(hdr);
894 unsigned int mtu = l->mtu;
895 u16 ack = l->rcv_nxt - 1;
896 u16 seqno = l->snd_nxt;
897 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
898 struct sk_buff_head *transmq = &l->transmq;
899 struct sk_buff_head *backlogq = &l->backlogq;
900 struct sk_buff *skb, *_skb, *bskb;
902 /* Match msg importance against this and all higher backlog limits: */
903 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
904 if (unlikely(l->backlog[i].len >= l->backlog[i].limit))
905 return link_schedule_user(l, list);
907 if (unlikely(msg_size(hdr) > mtu))
910 /* Prepare each packet for sending, and add to relevant queue: */
911 while (skb_queue_len(list)) {
912 skb = skb_peek(list);
914 msg_set_seqno(hdr, seqno);
915 msg_set_ack(hdr, ack);
916 msg_set_bcast_ack(hdr, bc_ack);
918 if (likely(skb_queue_len(transmq) < maxwin)) {
919 _skb = skb_clone(skb, GFP_ATOMIC);
923 __skb_queue_tail(transmq, skb);
924 __skb_queue_tail(xmitq, _skb);
925 TIPC_SKB_CB(skb)->ackers = l->ackers;
930 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
931 kfree_skb(__skb_dequeue(list));
932 l->stats.sent_bundled++;
935 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
936 kfree_skb(__skb_dequeue(list));
937 __skb_queue_tail(backlogq, bskb);
938 l->backlog[msg_importance(buf_msg(bskb))].len++;
939 l->stats.sent_bundled++;
940 l->stats.sent_bundles++;
943 l->backlog[imp].len += skb_queue_len(list);
944 skb_queue_splice_tail_init(list, backlogq);
950 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
952 struct sk_buff *skb, *_skb;
953 struct tipc_msg *hdr;
954 u16 seqno = l->snd_nxt;
955 u16 ack = l->rcv_nxt - 1;
956 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
958 while (skb_queue_len(&l->transmq) < l->window) {
959 skb = skb_peek(&l->backlogq);
962 _skb = skb_clone(skb, GFP_ATOMIC);
965 __skb_dequeue(&l->backlogq);
967 l->backlog[msg_importance(hdr)].len--;
968 __skb_queue_tail(&l->transmq, skb);
969 __skb_queue_tail(xmitq, _skb);
970 TIPC_SKB_CB(skb)->ackers = l->ackers;
971 msg_set_seqno(hdr, seqno);
972 msg_set_ack(hdr, ack);
973 msg_set_bcast_ack(hdr, bc_ack);
980 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
982 struct tipc_msg *hdr = buf_msg(skb);
984 pr_warn("Retransmission failure on link <%s>\n", l->name);
985 link_print(l, "Resetting link ");
986 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
987 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
988 pr_info("sqno %u, prev: %x, src: %x\n",
989 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
992 int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
993 struct sk_buff_head *xmitq)
995 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
996 struct tipc_msg *hdr;
997 u16 ack = l->rcv_nxt - 1;
998 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1003 /* Detect repeated retransmit failures on same packet */
1004 if (likely(l->last_retransm != buf_seqno(skb))) {
1005 l->last_retransm = buf_seqno(skb);
1007 } else if (++l->stale_count > 100) {
1008 link_retransmit_failure(l, skb);
1009 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1012 /* Move forward to where retransmission should start */
1013 skb_queue_walk(&l->transmq, skb) {
1014 if (!less(buf_seqno(skb), from))
1018 skb_queue_walk_from(&l->transmq, skb) {
1019 if (more(buf_seqno(skb), to))
1022 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1025 hdr = buf_msg(_skb);
1026 msg_set_ack(hdr, ack);
1027 msg_set_bcast_ack(hdr, bc_ack);
1028 _skb->priority = TC_PRIO_CONTROL;
1029 __skb_queue_tail(xmitq, _skb);
1030 l->stats.retransmitted++;
1035 /* tipc_data_input - deliver data and name distr msgs to upper layer
1037 * Consumes buffer if message is of right type
1038 * Node lock must be held
1040 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1041 struct sk_buff_head *inputq)
1043 switch (msg_user(buf_msg(skb))) {
1044 case TIPC_LOW_IMPORTANCE:
1045 case TIPC_MEDIUM_IMPORTANCE:
1046 case TIPC_HIGH_IMPORTANCE:
1047 case TIPC_CRITICAL_IMPORTANCE:
1049 skb_queue_tail(inputq, skb);
1051 case NAME_DISTRIBUTOR:
1052 l->bc_rcvlink->state = LINK_ESTABLISHED;
1053 skb_queue_tail(l->namedq, skb);
1056 case TUNNEL_PROTOCOL:
1057 case MSG_FRAGMENTER:
1058 case BCAST_PROTOCOL:
1061 pr_warn("Dropping received illegal msg type\n");
1067 /* tipc_link_input - process packet that has passed link protocol check
1071 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1072 struct sk_buff_head *inputq)
1074 struct tipc_msg *hdr = buf_msg(skb);
1075 struct sk_buff **reasm_skb = &l->reasm_buf;
1076 struct sk_buff *iskb;
1077 struct sk_buff_head tmpq;
1078 int usr = msg_user(hdr);
1083 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1084 if (msg_type(hdr) == SYNCH_MSG) {
1085 __skb_queue_purge(&l->deferdq);
1088 if (!tipc_msg_extract(skb, &iskb, &ipos))
1093 if (less(msg_seqno(hdr), l->drop_point))
1095 if (tipc_data_input(l, skb, inputq))
1097 usr = msg_user(hdr);
1098 reasm_skb = &l->failover_reasm_skb;
1101 if (usr == MSG_BUNDLER) {
1102 skb_queue_head_init(&tmpq);
1103 l->stats.recv_bundles++;
1104 l->stats.recv_bundled += msg_msgcnt(hdr);
1105 while (tipc_msg_extract(skb, &iskb, &pos))
1106 tipc_data_input(l, iskb, &tmpq);
1107 tipc_skb_queue_splice_tail(&tmpq, inputq);
1109 } else if (usr == MSG_FRAGMENTER) {
1110 l->stats.recv_fragments++;
1111 if (tipc_buf_append(reasm_skb, &skb)) {
1112 l->stats.recv_fragmented++;
1113 tipc_data_input(l, skb, inputq);
1114 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1115 pr_warn_ratelimited("Unable to build fragment list\n");
1116 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1119 } else if (usr == BCAST_PROTOCOL) {
1120 tipc_bcast_lock(l->net);
1121 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1122 tipc_bcast_unlock(l->net);
1129 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1131 bool released = false;
1132 struct sk_buff *skb, *tmp;
1134 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1135 if (more(buf_seqno(skb), acked))
1137 __skb_unlink(skb, &l->transmq);
1144 /* tipc_link_build_ack_msg: prepare link acknowledge message for transmission
1146 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1147 * risk of ack storms towards the sender
1149 int tipc_link_build_ack_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1154 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1155 if (link_is_bc_rcvlink(l)) {
1156 if (((l->rcv_nxt ^ link_own_addr(l)) & 0xf) != 0xf)
1159 return TIPC_LINK_SND_BC_ACK;
1164 l->stats.sent_acks++;
1165 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1169 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1171 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1173 int mtyp = RESET_MSG;
1175 if (l->state == LINK_ESTABLISHING)
1176 mtyp = ACTIVATE_MSG;
1178 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1181 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1183 static void tipc_link_build_nack_msg(struct tipc_link *l,
1184 struct sk_buff_head *xmitq)
1186 u32 def_cnt = ++l->stats.deferred_recv;
1188 if (link_is_bc_rcvlink(l))
1191 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1192 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1195 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1196 * @l: the link that should handle the message
1198 * @xmitq: queue to place packets to be sent after this call
1200 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1201 struct sk_buff_head *xmitq)
1203 struct sk_buff_head *defq = &l->deferdq;
1204 struct tipc_msg *hdr;
1205 u16 seqno, rcv_nxt, win_lim;
1210 seqno = msg_seqno(hdr);
1211 rcv_nxt = l->rcv_nxt;
1212 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1214 /* Verify and update link state */
1215 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1216 return tipc_link_proto_rcv(l, skb, xmitq);
1218 if (unlikely(!link_is_up(l))) {
1219 if (l->state == LINK_ESTABLISHING)
1220 rc = TIPC_LINK_UP_EVT;
1224 /* Don't send probe at next timeout expiration */
1225 l->silent_intv_cnt = 0;
1227 /* Drop if outside receive window */
1228 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1229 l->stats.duplicates++;
1233 /* Forward queues and wake up waiting users */
1234 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1235 tipc_link_advance_backlog(l, xmitq);
1236 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1237 link_prepare_wakeup(l);
1240 /* Defer delivery if sequence gap */
1241 if (unlikely(seqno != rcv_nxt)) {
1242 __tipc_skb_queue_sorted(defq, seqno, skb);
1243 tipc_link_build_nack_msg(l, xmitq);
1247 /* Deliver packet */
1249 l->stats.recv_info++;
1250 if (!tipc_data_input(l, skb, l->inputq))
1251 rc |= tipc_link_input(l, skb, l->inputq);
1252 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1253 rc |= tipc_link_build_ack_msg(l, xmitq);
1254 if (unlikely(rc & ~TIPC_LINK_SND_BC_ACK))
1256 } while ((skb = __skb_dequeue(defq)));
1265 * Send protocol message to the other endpoint.
1267 static void tipc_link_proto_xmit(struct tipc_link *l, u32 msg_typ,
1268 int probe_msg, u32 gap, u32 tolerance,
1271 struct sk_buff *skb = NULL;
1272 struct sk_buff_head xmitq;
1274 __skb_queue_head_init(&xmitq);
1275 tipc_link_build_proto_msg(l, msg_typ, probe_msg, gap,
1276 tolerance, priority, &xmitq);
1277 skb = __skb_dequeue(&xmitq);
1280 tipc_bearer_xmit_skb(l->net, l->bearer_id, skb, l->media_addr);
1284 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1285 u16 rcvgap, int tolerance, int priority,
1286 struct sk_buff_head *xmitq)
1288 struct sk_buff *skb = NULL;
1289 struct tipc_msg *hdr = l->pmsg;
1290 bool node_up = link_is_up(l->bc_rcvlink);
1292 /* Don't send protocol message during reset or link failover */
1293 if (tipc_link_is_blocked(l))
1296 msg_set_type(hdr, mtyp);
1297 msg_set_net_plane(hdr, l->net_plane);
1298 msg_set_next_sent(hdr, l->snd_nxt);
1299 msg_set_ack(hdr, l->rcv_nxt - 1);
1300 msg_set_bcast_ack(hdr, l->bc_rcvlink->rcv_nxt - 1);
1301 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1302 msg_set_link_tolerance(hdr, tolerance);
1303 msg_set_linkprio(hdr, priority);
1304 msg_set_redundant_link(hdr, node_up);
1305 msg_set_seq_gap(hdr, 0);
1307 /* Compatibility: created msg must not be in sequence with pkt flow */
1308 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1310 if (mtyp == STATE_MSG) {
1311 if (!tipc_link_is_up(l))
1314 /* Override rcvgap if there are packets in deferred queue */
1315 if (!skb_queue_empty(&l->deferdq))
1316 rcvgap = buf_seqno(skb_peek(&l->deferdq)) - l->rcv_nxt;
1318 msg_set_seq_gap(hdr, rcvgap);
1319 l->stats.sent_nacks++;
1321 msg_set_probe(hdr, probe);
1323 l->stats.sent_probes++;
1324 l->stats.sent_states++;
1327 /* RESET_MSG or ACTIVATE_MSG */
1328 msg_set_max_pkt(hdr, l->advertised_mtu);
1329 msg_set_ack(hdr, l->rcv_nxt - 1);
1330 msg_set_next_sent(hdr, 1);
1332 skb = tipc_buf_acquire(msg_size(hdr));
1335 skb_copy_to_linear_data(skb, hdr, msg_size(hdr));
1336 skb->priority = TC_PRIO_CONTROL;
1337 __skb_queue_tail(xmitq, skb);
1340 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1341 * with contents of the link's transmit and backlog queues.
1343 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1344 int mtyp, struct sk_buff_head *xmitq)
1346 struct sk_buff *skb, *tnlskb;
1347 struct tipc_msg *hdr, tnlhdr;
1348 struct sk_buff_head *queue = &l->transmq;
1349 struct sk_buff_head tmpxq, tnlq;
1350 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1355 skb_queue_head_init(&tnlq);
1356 skb_queue_head_init(&tmpxq);
1358 /* At least one packet required for safe algorithm => add dummy */
1359 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1360 BASIC_H_SIZE, 0, l->addr, link_own_addr(l),
1361 0, 0, TIPC_ERR_NO_PORT);
1363 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1366 skb_queue_tail(&tnlq, skb);
1367 tipc_link_xmit(l, &tnlq, &tmpxq);
1368 __skb_queue_purge(&tmpxq);
1370 /* Initialize reusable tunnel packet header */
1371 tipc_msg_init(link_own_addr(l), &tnlhdr, TUNNEL_PROTOCOL,
1372 mtyp, INT_H_SIZE, l->addr);
1373 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1374 msg_set_msgcnt(&tnlhdr, pktcnt);
1375 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1377 /* Wrap each packet into a tunnel packet */
1378 skb_queue_walk(queue, skb) {
1380 if (queue == &l->backlogq)
1381 msg_set_seqno(hdr, seqno++);
1382 pktlen = msg_size(hdr);
1383 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1384 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE);
1386 pr_warn("%sunable to send packet\n", link_co_err);
1389 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1390 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1391 __skb_queue_tail(&tnlq, tnlskb);
1393 if (queue != &l->backlogq) {
1394 queue = &l->backlogq;
1398 tipc_link_xmit(tnl, &tnlq, xmitq);
1400 if (mtyp == FAILOVER_MSG) {
1401 tnl->drop_point = l->rcv_nxt;
1402 tnl->failover_reasm_skb = l->reasm_buf;
1403 l->reasm_buf = NULL;
1407 /* tipc_link_proto_rcv(): receive link level protocol message :
1408 * Note that network plane id propagates through the network, and may
1409 * change at any time. The node with lowest numerical id determines
1412 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1413 struct sk_buff_head *xmitq)
1415 struct tipc_msg *hdr = buf_msg(skb);
1417 u16 ack = msg_ack(hdr);
1418 u16 gap = msg_seq_gap(hdr);
1419 u16 peers_snd_nxt = msg_next_sent(hdr);
1420 u16 peers_tol = msg_link_tolerance(hdr);
1421 u16 peers_prio = msg_linkprio(hdr);
1422 u16 rcv_nxt = l->rcv_nxt;
1423 int mtyp = msg_type(hdr);
1427 if (tipc_link_is_blocked(l) || !xmitq)
1430 if (link_own_addr(l) > msg_prevnode(hdr))
1431 l->net_plane = msg_net_plane(hdr);
1436 /* Ignore duplicate RESET with old session number */
1437 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1438 (l->peer_session != WILDCARD_SESSION))
1446 /* Complete own link name with peer's interface name */
1447 if_name = strrchr(l->name, ':') + 1;
1448 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1450 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1452 strncpy(if_name, msg_data(hdr), TIPC_MAX_IF_NAME);
1454 /* Update own tolerance if peer indicates a non-zero value */
1455 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1456 l->tolerance = peers_tol;
1458 /* Update own priority if peer's priority is higher */
1459 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1460 l->priority = peers_prio;
1462 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1463 if ((mtyp == RESET_MSG) || !link_is_up(l))
1464 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1466 /* ACTIVATE_MSG takes up link if it was already locally reset */
1467 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1468 rc = TIPC_LINK_UP_EVT;
1470 l->peer_session = msg_session(hdr);
1471 l->peer_bearer_id = msg_bearer_id(hdr);
1472 if (l->mtu > msg_max_pkt(hdr))
1473 l->mtu = msg_max_pkt(hdr);
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 l->silent_intv_cnt = 0;
1483 l->stats.recv_states++;
1485 l->stats.recv_probes++;
1487 if (!link_is_up(l)) {
1488 if (l->state == LINK_ESTABLISHING)
1489 rc = TIPC_LINK_UP_EVT;
1493 /* Send NACK if peer has sent pkts we haven't received yet */
1494 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1495 rcvgap = peers_snd_nxt - l->rcv_nxt;
1496 if (rcvgap || (msg_probe(hdr)))
1497 tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1499 tipc_link_release_pkts(l, ack);
1501 /* If NACK, retransmit will now start at right position */
1503 rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
1504 l->stats.recv_nacks++;
1507 tipc_link_advance_backlog(l, xmitq);
1508 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1509 link_prepare_wakeup(l);
1516 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1518 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1520 struct sk_buff_head *xmitq)
1522 struct sk_buff *skb;
1523 struct tipc_msg *hdr;
1524 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1525 u16 ack = l->rcv_nxt - 1;
1526 u16 gap_to = peers_snd_nxt - 1;
1528 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1529 0, l->addr, link_own_addr(l), 0, 0, 0);
1533 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1534 msg_set_bcast_ack(hdr, ack);
1535 msg_set_bcgap_after(hdr, ack);
1537 gap_to = buf_seqno(dfrd_skb) - 1;
1538 msg_set_bcgap_to(hdr, gap_to);
1539 msg_set_non_seq(hdr, bcast);
1540 __skb_queue_tail(xmitq, skb);
1544 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1546 * Give a newly added peer node the sequence number where it should
1547 * start receiving and acking broadcast packets.
1549 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1550 struct sk_buff_head *xmitq)
1552 struct sk_buff_head list;
1554 __skb_queue_head_init(&list);
1555 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1557 tipc_link_xmit(l, &list, xmitq);
1560 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1562 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1564 int mtyp = msg_type(hdr);
1565 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1570 if (msg_user(hdr) == BCAST_PROTOCOL) {
1571 l->rcv_nxt = peers_snd_nxt;
1572 l->state = LINK_ESTABLISHED;
1576 if (l->peer_caps & TIPC_BCAST_SYNCH)
1579 if (msg_peer_node_is_up(hdr))
1582 /* Compatibility: accept older, less safe initial synch data */
1583 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1584 l->rcv_nxt = peers_snd_nxt;
1587 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1589 void tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1590 struct sk_buff_head *xmitq)
1592 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1597 if (!msg_peer_node_is_up(hdr))
1600 l->bc_peer_is_up = true;
1602 /* Ignore if peers_snd_nxt goes beyond receive window */
1603 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1606 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1607 l->nack_state = BC_NACK_SND_CONDITIONAL;
1611 /* Don't NACK if one was recently sent or peeked */
1612 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1613 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1617 /* Conditionally delay NACK sending until next synch rcv */
1618 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1619 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1620 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1624 /* Send NACK now but suppress next one */
1625 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1626 l->nack_state = BC_NACK_SND_SUPPRESS;
1629 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1630 struct sk_buff_head *xmitq)
1632 struct sk_buff *skb, *tmp;
1633 struct tipc_link *snd_l = l->bc_sndlink;
1635 if (!link_is_up(l) || !l->bc_peer_is_up)
1638 if (!more(acked, l->acked))
1641 /* Skip over packets peer has already acked */
1642 skb_queue_walk(&snd_l->transmq, skb) {
1643 if (more(buf_seqno(skb), l->acked))
1647 /* Update/release the packets peer is acking now */
1648 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1649 if (more(buf_seqno(skb), acked))
1651 if (!--TIPC_SKB_CB(skb)->ackers) {
1652 __skb_unlink(skb, &snd_l->transmq);
1657 tipc_link_advance_backlog(snd_l, xmitq);
1658 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1659 link_prepare_wakeup(snd_l);
1662 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1664 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1665 struct sk_buff_head *xmitq)
1667 struct tipc_msg *hdr = buf_msg(skb);
1668 u32 dnode = msg_destnode(hdr);
1669 int mtyp = msg_type(hdr);
1670 u16 acked = msg_bcast_ack(hdr);
1671 u16 from = acked + 1;
1672 u16 to = msg_bcgap_to(hdr);
1673 u16 peers_snd_nxt = to + 1;
1678 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1681 if (mtyp != STATE_MSG)
1684 if (dnode == link_own_addr(l)) {
1685 tipc_link_bc_ack_rcv(l, acked, xmitq);
1686 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1687 l->stats.recv_nacks++;
1691 /* Msg for other node => suppress own NACK at next sync if applicable */
1692 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1693 l->nack_state = BC_NACK_SND_SUPPRESS;
1698 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1700 int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1703 l->backlog[TIPC_LOW_IMPORTANCE].limit = win / 2;
1704 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = win;
1705 l->backlog[TIPC_HIGH_IMPORTANCE].limit = win / 2 * 3;
1706 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = win * 2;
1707 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1711 * link_reset_stats - reset link statistics
1712 * @l: pointer to link
1714 void tipc_link_reset_stats(struct tipc_link *l)
1716 memset(&l->stats, 0, sizeof(l->stats));
1717 if (!link_is_bc_sndlink(l)) {
1718 l->stats.sent_info = l->snd_nxt;
1719 l->stats.recv_info = l->rcv_nxt;
1723 static void link_print(struct tipc_link *l, const char *str)
1725 struct sk_buff *hskb = skb_peek(&l->transmq);
1726 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1727 u16 tail = l->snd_nxt - 1;
1729 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1730 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1731 skb_queue_len(&l->transmq), head, tail,
1732 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1735 /* Parse and validate nested (link) properties valid for media, bearer and link
1737 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1741 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1742 tipc_nl_prop_policy);
1746 if (props[TIPC_NLA_PROP_PRIO]) {
1749 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1750 if (prio > TIPC_MAX_LINK_PRI)
1754 if (props[TIPC_NLA_PROP_TOL]) {
1757 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1758 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1762 if (props[TIPC_NLA_PROP_WIN]) {
1765 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1766 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1773 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1776 struct nlattr *stats;
1783 struct nla_map map[] = {
1784 {TIPC_NLA_STATS_RX_INFO, s->recv_info},
1785 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1786 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1787 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1788 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1789 {TIPC_NLA_STATS_TX_INFO, s->sent_info},
1790 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1791 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1792 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1793 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1794 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1795 s->msg_length_counts : 1},
1796 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1797 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1798 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1799 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1800 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1801 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1802 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1803 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1804 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1805 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1806 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1807 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1808 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1809 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1810 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1811 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1812 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1813 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1814 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1815 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1816 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1817 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1818 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1821 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1825 for (i = 0; i < ARRAY_SIZE(map); i++)
1826 if (nla_put_u32(skb, map[i].key, map[i].val))
1829 nla_nest_end(skb, stats);
1833 nla_nest_cancel(skb, stats);
1838 /* Caller should hold appropriate locks to protect the link */
1839 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1840 struct tipc_link *link, int nlflags)
1844 struct nlattr *attrs;
1845 struct nlattr *prop;
1846 struct tipc_net *tn = net_generic(net, tipc_net_id);
1848 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1849 nlflags, TIPC_NL_LINK_GET);
1853 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1857 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1859 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1860 tipc_cluster_mask(tn->own_addr)))
1862 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1864 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->rcv_nxt))
1866 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->snd_nxt))
1869 if (tipc_link_is_up(link))
1870 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1873 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
1876 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1879 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1881 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
1883 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
1886 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1888 nla_nest_end(msg->skb, prop);
1890 err = __tipc_nl_add_stats(msg->skb, &link->stats);
1894 nla_nest_end(msg->skb, attrs);
1895 genlmsg_end(msg->skb, hdr);
1900 nla_nest_cancel(msg->skb, prop);
1902 nla_nest_cancel(msg->skb, attrs);
1904 genlmsg_cancel(msg->skb, hdr);
1909 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
1910 struct tipc_stats *stats)
1913 struct nlattr *nest;
1920 struct nla_map map[] = {
1921 {TIPC_NLA_STATS_RX_INFO, stats->recv_info},
1922 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
1923 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
1924 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
1925 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
1926 {TIPC_NLA_STATS_TX_INFO, stats->sent_info},
1927 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
1928 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
1929 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
1930 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
1931 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
1932 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
1933 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
1934 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
1935 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
1936 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
1937 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
1938 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
1939 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
1940 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
1943 nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1947 for (i = 0; i < ARRAY_SIZE(map); i++)
1948 if (nla_put_u32(skb, map[i].key, map[i].val))
1951 nla_nest_end(skb, nest);
1955 nla_nest_cancel(skb, nest);
1960 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
1964 struct nlattr *attrs;
1965 struct nlattr *prop;
1966 struct tipc_net *tn = net_generic(net, tipc_net_id);
1967 struct tipc_link *bcl = tn->bcl;
1972 tipc_bcast_lock(net);
1974 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1975 NLM_F_MULTI, TIPC_NL_LINK_GET);
1979 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1983 /* The broadcast link is always up */
1984 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1987 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
1989 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
1991 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, bcl->rcv_nxt))
1993 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, bcl->snd_nxt))
1996 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1999 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2001 nla_nest_end(msg->skb, prop);
2003 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2007 tipc_bcast_unlock(net);
2008 nla_nest_end(msg->skb, attrs);
2009 genlmsg_end(msg->skb, hdr);
2014 nla_nest_cancel(msg->skb, prop);
2016 nla_nest_cancel(msg->skb, attrs);
2018 tipc_bcast_unlock(net);
2019 genlmsg_cancel(msg->skb, hdr);
2024 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol)
2027 tipc_link_proto_xmit(l, STATE_MSG, 0, 0, tol, 0);
2030 void tipc_link_set_prio(struct tipc_link *l, u32 prio)
2033 tipc_link_proto_xmit(l, STATE_MSG, 0, 0, 0, prio);
2036 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2038 l->abort_limit = limit;