2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, 2012-2016, 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"
47 #include <linux/pkt_sched.h>
68 u32 link_congs; /* # port sends blocked by congestion */
71 u32 max_queue_sz; /* send queue size high water mark */
72 u32 accu_queue_sz; /* used for send queue size profiling */
73 u32 queue_sz_counts; /* used for send queue size profiling */
74 u32 msg_length_counts; /* used for message length profiling */
75 u32 msg_lengths_total; /* used for message length profiling */
76 u32 msg_length_profile[7]; /* used for msg. length profiling */
80 * struct tipc_link - TIPC link data structure
81 * @addr: network address of link's peer node
82 * @name: link name character string
83 * @media_addr: media address to use when sending messages over link
85 * @net: pointer to namespace struct
86 * @refcnt: reference counter for permanent references (owner node & timer)
87 * @peer_session: link session # being used by peer end of link
88 * @peer_bearer_id: bearer id used by link's peer endpoint
89 * @bearer_id: local bearer id used by link
90 * @tolerance: minimum link continuity loss needed to reset link [in ms]
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 * @mon_state: cookie with information needed by link monitor
100 * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
101 * @exp_msg_count: # of tunnelled messages expected during link changeover
102 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
103 * @mtu: current maximum packet size for this link
104 * @advertised_mtu: advertised own mtu when link is being established
105 * @transmitq: queue for sent, non-acked messages
106 * @backlogq: queue for messages waiting to be sent
107 * @snt_nxt: next sequence number to use for outbound messages
108 * @last_retransmitted: sequence number of most recently retransmitted message
109 * @stale_count: # of identical retransmit requests made by peer
110 * @ackers: # of peers that needs to ack each packet before it can be released
111 * @acked: # last packet acked by a certain peer. Used for broadcast.
112 * @rcv_nxt: next sequence number to expect for inbound messages
113 * @deferred_queue: deferred queue saved OOS b'cast message received from node
114 * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
115 * @inputq: buffer queue for messages to be delivered upwards
116 * @namedq: buffer queue for name table messages to be delivered upwards
117 * @next_out: ptr to first unsent outbound message in queue
118 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
119 * @long_msg_seq_no: next identifier to use for outbound fragmented messages
120 * @reasm_buf: head of partially reassembled inbound message fragments
121 * @bc_rcvr: marks that this is a broadcast receiver link
122 * @stats: collects statistics regarding link activity
126 char name[TIPC_MAX_LINK_NAME];
129 /* Management and link supervision data */
140 char if_name[TIPC_MAX_IF_NAME];
143 struct tipc_mon_state mon_state;
148 struct sk_buff *failover_reasm_skb;
150 /* Max packet negotiation */
155 struct sk_buff_head transmq;
156 struct sk_buff_head backlogq;
160 struct sk_buff *target_bskb;
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;
185 unsigned long prev_retr;
192 struct tipc_stats stats;
196 * Error message prefixes
198 static const char *link_co_err = "Link tunneling error, ";
199 static const char *link_rst_msg = "Resetting link ";
201 /* Send states for broadcast NACKs
204 BC_NACK_SND_CONDITIONAL,
205 BC_NACK_SND_UNCONDITIONAL,
206 BC_NACK_SND_SUPPRESS,
209 #define TIPC_BC_RETR_LIMIT 10 /* [ms] */
212 * Interval between NACKs when packets arrive out of order
214 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
216 /* Wildcard value for link session numbers. When it is known that
217 * peer endpoint is down, any session number must be accepted.
219 #define ANY_SESSION 0x10000
224 LINK_ESTABLISHED = 0xe,
225 LINK_ESTABLISHING = 0xe << 4,
226 LINK_RESET = 0x1 << 8,
227 LINK_RESETTING = 0x2 << 12,
228 LINK_PEER_RESET = 0xd << 16,
229 LINK_FAILINGOVER = 0xf << 20,
230 LINK_SYNCHING = 0xc << 24
233 /* Link FSM state checking routines
235 static int link_is_up(struct tipc_link *l)
237 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
240 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
241 struct sk_buff_head *xmitq);
242 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
243 u16 rcvgap, int tolerance, int priority,
244 struct sk_buff_head *xmitq);
245 static void link_print(struct tipc_link *l, const char *str);
246 static int tipc_link_build_nack_msg(struct tipc_link *l,
247 struct sk_buff_head *xmitq);
248 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
249 struct sk_buff_head *xmitq);
250 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
253 * Simple non-static link routines (i.e. referenced outside this file)
255 bool tipc_link_is_up(struct tipc_link *l)
257 return link_is_up(l);
260 bool tipc_link_peer_is_down(struct tipc_link *l)
262 return l->state == LINK_PEER_RESET;
265 bool tipc_link_is_reset(struct tipc_link *l)
267 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
270 bool tipc_link_is_establishing(struct tipc_link *l)
272 return l->state == LINK_ESTABLISHING;
275 bool tipc_link_is_synching(struct tipc_link *l)
277 return l->state == LINK_SYNCHING;
280 bool tipc_link_is_failingover(struct tipc_link *l)
282 return l->state == LINK_FAILINGOVER;
285 bool tipc_link_is_blocked(struct tipc_link *l)
287 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
290 static bool link_is_bc_sndlink(struct tipc_link *l)
292 return !l->bc_sndlink;
295 static bool link_is_bc_rcvlink(struct tipc_link *l)
297 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
300 int tipc_link_is_active(struct tipc_link *l)
305 void tipc_link_set_active(struct tipc_link *l, bool active)
310 u32 tipc_link_id(struct tipc_link *l)
312 return l->peer_bearer_id << 16 | l->bearer_id;
315 int tipc_link_window(struct tipc_link *l)
320 int tipc_link_prio(struct tipc_link *l)
325 unsigned long tipc_link_tolerance(struct tipc_link *l)
330 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
335 char tipc_link_plane(struct tipc_link *l)
340 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
341 struct tipc_link *uc_l,
342 struct sk_buff_head *xmitq)
344 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
347 rcv_l->acked = snd_l->snd_nxt - 1;
348 snd_l->state = LINK_ESTABLISHED;
349 tipc_link_build_bc_init_msg(uc_l, xmitq);
352 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
353 struct tipc_link *rcv_l,
354 struct sk_buff_head *xmitq)
356 u16 ack = snd_l->snd_nxt - 1;
359 rcv_l->bc_peer_is_up = true;
360 rcv_l->state = LINK_ESTABLISHED;
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 u16 link_bc_rcv_gap(struct tipc_link *l)
378 struct sk_buff *skb = skb_peek(&l->deferdq);
381 if (more(l->snd_nxt, l->rcv_nxt))
382 gap = l->snd_nxt - l->rcv_nxt;
384 gap = buf_seqno(skb) - l->rcv_nxt;
388 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
393 int tipc_link_mtu(struct tipc_link *l)
398 u16 tipc_link_rcv_nxt(struct tipc_link *l)
403 u16 tipc_link_acked(struct tipc_link *l)
408 char *tipc_link_name(struct tipc_link *l)
414 * tipc_link_create - create a new link
415 * @n: pointer to associated node
416 * @if_name: associated interface name
417 * @bearer_id: id (index) of associated bearer
418 * @tolerance: link tolerance to be used by link
419 * @net_plane: network plane (A,B,c..) this link belongs to
420 * @mtu: mtu to be advertised by link
421 * @priority: priority to be used by link
422 * @window: send window to be used by link
423 * @session: session to be used by link
424 * @ownnode: identity of own node
425 * @peer: node id of peer node
426 * @peer_caps: bitmap describing peer node capabilities
427 * @bc_sndlink: the namespace global link used for broadcast sending
428 * @bc_rcvlink: the peer specific link used for broadcast reception
429 * @inputq: queue to put messages ready for delivery
430 * @namedq: queue to put binding table update messages ready for delivery
431 * @link: return value, pointer to put the created link
433 * Returns true if link was created, otherwise false
435 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
436 int tolerance, char net_plane, u32 mtu, int priority,
437 int window, u32 session, u32 ownnode, u32 peer,
439 struct tipc_link *bc_sndlink,
440 struct tipc_link *bc_rcvlink,
441 struct sk_buff_head *inputq,
442 struct sk_buff_head *namedq,
443 struct tipc_link **link)
447 l = kzalloc(sizeof(*l), GFP_ATOMIC);
451 l->session = session;
453 /* Note: peer i/f name is completed by reset/activate message */
454 sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
455 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
456 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
457 strcpy(l->if_name, if_name);
459 l->peer_caps = peer_caps;
461 l->peer_session = ANY_SESSION;
462 l->bearer_id = bearer_id;
463 l->tolerance = tolerance;
464 l->net_plane = net_plane;
465 l->advertised_mtu = mtu;
467 l->priority = priority;
468 tipc_link_set_queue_limits(l, window);
470 l->bc_sndlink = bc_sndlink;
471 l->bc_rcvlink = bc_rcvlink;
474 l->state = LINK_RESETTING;
475 __skb_queue_head_init(&l->transmq);
476 __skb_queue_head_init(&l->backlogq);
477 __skb_queue_head_init(&l->deferdq);
478 skb_queue_head_init(&l->wakeupq);
479 skb_queue_head_init(l->inputq);
484 * tipc_link_bc_create - create new link to be used for broadcast
485 * @n: pointer to associated node
486 * @mtu: mtu to be used
487 * @window: send window to be used
488 * @inputq: queue to put messages ready for delivery
489 * @namedq: queue to put binding table update messages ready for delivery
490 * @link: return value, pointer to put the created link
492 * Returns true if link was created, otherwise false
494 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
495 int mtu, int window, u16 peer_caps,
496 struct sk_buff_head *inputq,
497 struct sk_buff_head *namedq,
498 struct tipc_link *bc_sndlink,
499 struct tipc_link **link)
503 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
504 0, ownnode, peer, peer_caps, bc_sndlink,
505 NULL, inputq, namedq, link))
509 strcpy(l->name, tipc_bclink_name);
511 l->state = LINK_RESET;
515 /* Broadcast send link is always up */
516 if (link_is_bc_sndlink(l))
517 l->state = LINK_ESTABLISHED;
519 /* Disable replicast if even a single peer doesn't support it */
520 if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
521 tipc_bcast_disable_rcast(net);
527 * tipc_link_fsm_evt - link finite state machine
528 * @l: pointer to link
529 * @evt: state machine event to be processed
531 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
538 case LINK_PEER_RESET_EVT:
539 l->state = LINK_PEER_RESET;
542 l->state = LINK_RESET;
544 case LINK_FAILURE_EVT:
545 case LINK_FAILOVER_BEGIN_EVT:
546 case LINK_ESTABLISH_EVT:
547 case LINK_FAILOVER_END_EVT:
548 case LINK_SYNCH_BEGIN_EVT:
549 case LINK_SYNCH_END_EVT:
556 case LINK_PEER_RESET_EVT:
557 l->state = LINK_ESTABLISHING;
559 case LINK_FAILOVER_BEGIN_EVT:
560 l->state = LINK_FAILINGOVER;
561 case LINK_FAILURE_EVT:
563 case LINK_ESTABLISH_EVT:
564 case LINK_FAILOVER_END_EVT:
566 case LINK_SYNCH_BEGIN_EVT:
567 case LINK_SYNCH_END_EVT:
572 case LINK_PEER_RESET:
575 l->state = LINK_ESTABLISHING;
577 case LINK_PEER_RESET_EVT:
578 case LINK_ESTABLISH_EVT:
579 case LINK_FAILURE_EVT:
581 case LINK_SYNCH_BEGIN_EVT:
582 case LINK_SYNCH_END_EVT:
583 case LINK_FAILOVER_BEGIN_EVT:
584 case LINK_FAILOVER_END_EVT:
589 case LINK_FAILINGOVER:
591 case LINK_FAILOVER_END_EVT:
592 l->state = LINK_RESET;
594 case LINK_PEER_RESET_EVT:
596 case LINK_ESTABLISH_EVT:
597 case LINK_FAILURE_EVT:
599 case LINK_FAILOVER_BEGIN_EVT:
600 case LINK_SYNCH_BEGIN_EVT:
601 case LINK_SYNCH_END_EVT:
606 case LINK_ESTABLISHING:
608 case LINK_ESTABLISH_EVT:
609 l->state = LINK_ESTABLISHED;
611 case LINK_FAILOVER_BEGIN_EVT:
612 l->state = LINK_FAILINGOVER;
615 l->state = LINK_RESET;
617 case LINK_FAILURE_EVT:
618 case LINK_PEER_RESET_EVT:
619 case LINK_SYNCH_BEGIN_EVT:
620 case LINK_FAILOVER_END_EVT:
622 case LINK_SYNCH_END_EVT:
627 case LINK_ESTABLISHED:
629 case LINK_PEER_RESET_EVT:
630 l->state = LINK_PEER_RESET;
631 rc |= TIPC_LINK_DOWN_EVT;
633 case LINK_FAILURE_EVT:
634 l->state = LINK_RESETTING;
635 rc |= TIPC_LINK_DOWN_EVT;
638 l->state = LINK_RESET;
640 case LINK_ESTABLISH_EVT:
641 case LINK_SYNCH_END_EVT:
643 case LINK_SYNCH_BEGIN_EVT:
644 l->state = LINK_SYNCHING;
646 case LINK_FAILOVER_BEGIN_EVT:
647 case LINK_FAILOVER_END_EVT:
654 case LINK_PEER_RESET_EVT:
655 l->state = LINK_PEER_RESET;
656 rc |= TIPC_LINK_DOWN_EVT;
658 case LINK_FAILURE_EVT:
659 l->state = LINK_RESETTING;
660 rc |= TIPC_LINK_DOWN_EVT;
663 l->state = LINK_RESET;
665 case LINK_ESTABLISH_EVT:
666 case LINK_SYNCH_BEGIN_EVT:
668 case LINK_SYNCH_END_EVT:
669 l->state = LINK_ESTABLISHED;
671 case LINK_FAILOVER_BEGIN_EVT:
672 case LINK_FAILOVER_END_EVT:
678 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
682 pr_err("Illegal FSM event %x in state %x on link %s\n",
683 evt, l->state, l->name);
687 /* link_profile_stats - update statistical profiling of traffic
689 static void link_profile_stats(struct tipc_link *l)
692 struct tipc_msg *msg;
695 /* Update counters used in statistical profiling of send traffic */
696 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
697 l->stats.queue_sz_counts++;
699 skb = skb_peek(&l->transmq);
703 length = msg_size(msg);
705 if (msg_user(msg) == MSG_FRAGMENTER) {
706 if (msg_type(msg) != FIRST_FRAGMENT)
708 length = msg_size(msg_get_wrapped(msg));
710 l->stats.msg_lengths_total += length;
711 l->stats.msg_length_counts++;
713 l->stats.msg_length_profile[0]++;
714 else if (length <= 256)
715 l->stats.msg_length_profile[1]++;
716 else if (length <= 1024)
717 l->stats.msg_length_profile[2]++;
718 else if (length <= 4096)
719 l->stats.msg_length_profile[3]++;
720 else if (length <= 16384)
721 l->stats.msg_length_profile[4]++;
722 else if (length <= 32768)
723 l->stats.msg_length_profile[5]++;
725 l->stats.msg_length_profile[6]++;
728 /* tipc_link_timeout - perform periodic task as instructed from node timeout
730 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
737 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
738 u16 bc_acked = l->bc_rcvlink->acked;
739 struct tipc_mon_state *mstate = &l->mon_state;
742 case LINK_ESTABLISHED:
745 link_profile_stats(l);
746 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
747 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
748 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
749 state = bc_acked != bc_snt;
750 state |= l->bc_rcvlink->rcv_unacked;
751 state |= l->rcv_unacked;
752 state |= !skb_queue_empty(&l->transmq);
753 state |= !skb_queue_empty(&l->deferdq);
754 probe = mstate->probing;
755 probe |= l->silent_intv_cnt;
756 if (probe || mstate->monitoring)
757 l->silent_intv_cnt++;
760 setup = l->rst_cnt++ <= 4;
761 setup |= !(l->rst_cnt % 16);
764 case LINK_ESTABLISHING:
768 case LINK_PEER_RESET:
770 case LINK_FAILINGOVER:
776 if (state || probe || setup)
777 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, xmitq);
783 * link_schedule_user - schedule a message sender for wakeup after congestion
785 * @hdr: header of message that is being sent
786 * Create pseudo msg to send back to user when congestion abates
788 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
790 u32 dnode = tipc_own_addr(l->net);
791 u32 dport = msg_origport(hdr);
794 /* Create and schedule wakeup pseudo message */
795 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
796 dnode, l->addr, dport, 0, 0);
799 msg_set_dest_droppable(buf_msg(skb), true);
800 TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
801 skb_queue_tail(&l->wakeupq, skb);
802 l->stats.link_congs++;
807 * link_prepare_wakeup - prepare users for wakeup after congestion
809 * Wake up a number of waiting users, as permitted by available space
812 void link_prepare_wakeup(struct tipc_link *l)
814 struct sk_buff_head *wakeupq = &l->wakeupq;
815 struct sk_buff_head *inputq = l->inputq;
816 struct sk_buff *skb, *tmp;
817 struct sk_buff_head tmpq;
821 __skb_queue_head_init(&tmpq);
823 for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
824 avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
826 skb_queue_walk_safe(wakeupq, skb, tmp) {
827 imp = TIPC_SKB_CB(skb)->chain_imp;
831 __skb_unlink(skb, wakeupq);
832 __skb_queue_tail(&tmpq, skb);
835 spin_lock_bh(&inputq->lock);
836 skb_queue_splice_tail(&tmpq, inputq);
837 spin_unlock_bh(&inputq->lock);
841 void tipc_link_reset(struct tipc_link *l)
845 l->peer_session = ANY_SESSION;
847 l->mtu = l->advertised_mtu;
848 __skb_queue_purge(&l->transmq);
849 __skb_queue_purge(&l->deferdq);
850 skb_queue_splice_init(&l->wakeupq, l->inputq);
851 __skb_queue_purge(&l->backlogq);
852 for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
853 l->backlog[imp].len = 0;
854 l->backlog[imp].target_bskb = NULL;
856 kfree_skb(l->reasm_buf);
857 kfree_skb(l->failover_reasm_skb);
859 l->failover_reasm_skb = NULL;
864 l->silent_intv_cnt = 0;
867 l->bc_peer_is_up = false;
868 memset(&l->mon_state, 0, sizeof(l->mon_state));
869 tipc_link_reset_stats(l);
873 * tipc_link_xmit(): enqueue buffer list according to queue situation
875 * @list: chain of buffers containing message
876 * @xmitq: returned list of packets to be sent by caller
878 * Consumes the buffer chain.
879 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
880 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
882 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
883 struct sk_buff_head *xmitq)
885 unsigned int maxwin = l->window;
886 unsigned int mtu = l->mtu;
887 u16 ack = l->rcv_nxt - 1;
888 u16 seqno = l->snd_nxt;
889 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
890 struct sk_buff_head *transmq = &l->transmq;
891 struct sk_buff_head *backlogq = &l->backlogq;
892 struct sk_buff *skb, *_skb, **tskb;
893 int pkt_cnt = skb_queue_len(list);
894 struct tipc_msg *hdr;
901 hdr = buf_msg(skb_peek(list));
902 if (unlikely(msg_size(hdr) > mtu)) {
903 skb_queue_purge(list);
907 imp = msg_importance(hdr);
908 /* Allow oversubscription of one data msg per source at congestion */
909 if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
910 if (imp == TIPC_SYSTEM_IMPORTANCE) {
911 pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
914 rc = link_schedule_user(l, hdr);
918 l->stats.sent_fragmented++;
919 l->stats.sent_fragments += pkt_cnt;
922 /* Prepare each packet for sending, and add to relevant queue: */
923 while (skb_queue_len(list)) {
924 skb = skb_peek(list);
926 msg_set_seqno(hdr, seqno);
927 msg_set_ack(hdr, ack);
928 msg_set_bcast_ack(hdr, bc_ack);
930 if (likely(skb_queue_len(transmq) < maxwin)) {
931 _skb = skb_clone(skb, GFP_ATOMIC);
933 skb_queue_purge(list);
937 __skb_queue_tail(transmq, skb);
938 __skb_queue_tail(xmitq, _skb);
939 TIPC_SKB_CB(skb)->ackers = l->ackers;
941 l->stats.sent_pkts++;
945 tskb = &l->backlog[imp].target_bskb;
946 if (tipc_msg_bundle(*tskb, hdr, mtu)) {
947 kfree_skb(__skb_dequeue(list));
948 l->stats.sent_bundled++;
951 if (tipc_msg_make_bundle(tskb, hdr, mtu, l->addr)) {
952 kfree_skb(__skb_dequeue(list));
953 __skb_queue_tail(backlogq, *tskb);
954 l->backlog[imp].len++;
955 l->stats.sent_bundled++;
956 l->stats.sent_bundles++;
959 l->backlog[imp].target_bskb = NULL;
960 l->backlog[imp].len += skb_queue_len(list);
961 skb_queue_splice_tail_init(list, backlogq);
967 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
969 struct sk_buff *skb, *_skb;
970 struct tipc_msg *hdr;
971 u16 seqno = l->snd_nxt;
972 u16 ack = l->rcv_nxt - 1;
973 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
976 while (skb_queue_len(&l->transmq) < l->window) {
977 skb = skb_peek(&l->backlogq);
980 _skb = skb_clone(skb, GFP_ATOMIC);
983 __skb_dequeue(&l->backlogq);
985 imp = msg_importance(hdr);
986 l->backlog[imp].len--;
987 if (unlikely(skb == l->backlog[imp].target_bskb))
988 l->backlog[imp].target_bskb = NULL;
989 __skb_queue_tail(&l->transmq, skb);
990 __skb_queue_tail(xmitq, _skb);
991 TIPC_SKB_CB(skb)->ackers = l->ackers;
992 msg_set_seqno(hdr, seqno);
993 msg_set_ack(hdr, ack);
994 msg_set_bcast_ack(hdr, bc_ack);
996 l->stats.sent_pkts++;
1002 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
1004 struct tipc_msg *hdr = buf_msg(skb);
1006 pr_warn("Retransmission failure on link <%s>\n", l->name);
1007 link_print(l, "State of link ");
1008 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1009 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1010 pr_info("sqno %u, prev: %x, src: %x\n",
1011 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
1014 int tipc_link_retrans(struct tipc_link *l, struct tipc_link *nacker,
1015 u16 from, u16 to, struct sk_buff_head *xmitq)
1017 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1018 struct tipc_msg *hdr;
1019 u16 ack = l->rcv_nxt - 1;
1020 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1025 /* Detect repeated retransmit failures on same packet */
1026 if (nacker->last_retransm != buf_seqno(skb)) {
1027 nacker->last_retransm = buf_seqno(skb);
1028 nacker->stale_count = 1;
1029 } else if (++nacker->stale_count > 100) {
1030 link_retransmit_failure(l, skb);
1031 nacker->stale_count = 0;
1032 if (link_is_bc_sndlink(l))
1033 return TIPC_LINK_DOWN_EVT;
1034 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1037 /* Move forward to where retransmission should start */
1038 skb_queue_walk(&l->transmq, skb) {
1039 if (!less(buf_seqno(skb), from))
1043 skb_queue_walk_from(&l->transmq, skb) {
1044 if (more(buf_seqno(skb), to))
1047 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1050 hdr = buf_msg(_skb);
1051 msg_set_ack(hdr, ack);
1052 msg_set_bcast_ack(hdr, bc_ack);
1053 _skb->priority = TC_PRIO_CONTROL;
1054 __skb_queue_tail(xmitq, _skb);
1055 l->stats.retransmitted++;
1060 /* tipc_data_input - deliver data and name distr msgs to upper layer
1062 * Consumes buffer if message is of right type
1063 * Node lock must be held
1065 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1066 struct sk_buff_head *inputq)
1068 struct tipc_msg *hdr = buf_msg(skb);
1070 switch (msg_user(hdr)) {
1071 case TIPC_LOW_IMPORTANCE:
1072 case TIPC_MEDIUM_IMPORTANCE:
1073 case TIPC_HIGH_IMPORTANCE:
1074 case TIPC_CRITICAL_IMPORTANCE:
1075 if (unlikely(msg_type(hdr) == TIPC_MCAST_MSG)) {
1076 skb_queue_tail(l->bc_rcvlink->inputq, skb);
1080 skb_queue_tail(inputq, skb);
1082 case NAME_DISTRIBUTOR:
1083 l->bc_rcvlink->state = LINK_ESTABLISHED;
1084 skb_queue_tail(l->namedq, skb);
1087 case TUNNEL_PROTOCOL:
1088 case MSG_FRAGMENTER:
1089 case BCAST_PROTOCOL:
1092 pr_warn("Dropping received illegal msg type\n");
1098 /* tipc_link_input - process packet that has passed link protocol check
1102 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1103 struct sk_buff_head *inputq)
1105 struct tipc_msg *hdr = buf_msg(skb);
1106 struct sk_buff **reasm_skb = &l->reasm_buf;
1107 struct sk_buff *iskb;
1108 struct sk_buff_head tmpq;
1109 int usr = msg_user(hdr);
1114 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1115 if (msg_type(hdr) == SYNCH_MSG) {
1116 __skb_queue_purge(&l->deferdq);
1119 if (!tipc_msg_extract(skb, &iskb, &ipos))
1124 if (less(msg_seqno(hdr), l->drop_point))
1126 if (tipc_data_input(l, skb, inputq))
1128 usr = msg_user(hdr);
1129 reasm_skb = &l->failover_reasm_skb;
1132 if (usr == MSG_BUNDLER) {
1133 skb_queue_head_init(&tmpq);
1134 l->stats.recv_bundles++;
1135 l->stats.recv_bundled += msg_msgcnt(hdr);
1136 while (tipc_msg_extract(skb, &iskb, &pos))
1137 tipc_data_input(l, iskb, &tmpq);
1138 tipc_skb_queue_splice_tail(&tmpq, inputq);
1140 } else if (usr == MSG_FRAGMENTER) {
1141 l->stats.recv_fragments++;
1142 if (tipc_buf_append(reasm_skb, &skb)) {
1143 l->stats.recv_fragmented++;
1144 tipc_data_input(l, skb, inputq);
1145 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1146 pr_warn_ratelimited("Unable to build fragment list\n");
1147 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1150 } else if (usr == BCAST_PROTOCOL) {
1151 tipc_bcast_lock(l->net);
1152 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1153 tipc_bcast_unlock(l->net);
1160 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1162 bool released = false;
1163 struct sk_buff *skb, *tmp;
1165 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1166 if (more(buf_seqno(skb), acked))
1168 __skb_unlink(skb, &l->transmq);
1175 /* tipc_link_build_state_msg: prepare link state message for transmission
1177 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1178 * risk of ack storms towards the sender
1180 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1185 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1186 if (link_is_bc_rcvlink(l)) {
1187 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1191 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1192 l->snd_nxt = l->rcv_nxt;
1193 return TIPC_LINK_SND_STATE;
1198 l->stats.sent_acks++;
1199 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1203 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1205 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1207 int mtyp = RESET_MSG;
1208 struct sk_buff *skb;
1210 if (l->state == LINK_ESTABLISHING)
1211 mtyp = ACTIVATE_MSG;
1213 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1215 /* Inform peer that this endpoint is going down if applicable */
1216 skb = skb_peek_tail(xmitq);
1217 if (skb && (l->state == LINK_RESET))
1218 msg_set_peer_stopping(buf_msg(skb), 1);
1221 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1222 * Note that sending of broadcast NACK is coordinated among nodes, to
1223 * reduce the risk of NACK storms towards the sender
1225 static int tipc_link_build_nack_msg(struct tipc_link *l,
1226 struct sk_buff_head *xmitq)
1228 u32 def_cnt = ++l->stats.deferred_recv;
1231 if (link_is_bc_rcvlink(l)) {
1232 match1 = def_cnt & 0xf;
1233 match2 = tipc_own_addr(l->net) & 0xf;
1234 if (match1 == match2)
1235 return TIPC_LINK_SND_STATE;
1239 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1240 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1244 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1245 * @l: the link that should handle the message
1247 * @xmitq: queue to place packets to be sent after this call
1249 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1250 struct sk_buff_head *xmitq)
1252 struct sk_buff_head *defq = &l->deferdq;
1253 struct tipc_msg *hdr;
1254 u16 seqno, rcv_nxt, win_lim;
1259 seqno = msg_seqno(hdr);
1260 rcv_nxt = l->rcv_nxt;
1261 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1263 /* Verify and update link state */
1264 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1265 return tipc_link_proto_rcv(l, skb, xmitq);
1267 if (unlikely(!link_is_up(l))) {
1268 if (l->state == LINK_ESTABLISHING)
1269 rc = TIPC_LINK_UP_EVT;
1273 /* Don't send probe at next timeout expiration */
1274 l->silent_intv_cnt = 0;
1276 /* Drop if outside receive window */
1277 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1278 l->stats.duplicates++;
1282 /* Forward queues and wake up waiting users */
1283 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1284 tipc_link_advance_backlog(l, xmitq);
1285 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1286 link_prepare_wakeup(l);
1289 /* Defer delivery if sequence gap */
1290 if (unlikely(seqno != rcv_nxt)) {
1291 __tipc_skb_queue_sorted(defq, seqno, skb);
1292 rc |= tipc_link_build_nack_msg(l, xmitq);
1296 /* Deliver packet */
1298 l->stats.recv_pkts++;
1299 if (!tipc_data_input(l, skb, l->inputq))
1300 rc |= tipc_link_input(l, skb, l->inputq);
1301 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1302 rc |= tipc_link_build_state_msg(l, xmitq);
1303 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1305 } while ((skb = __skb_dequeue(defq)));
1313 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1314 u16 rcvgap, int tolerance, int priority,
1315 struct sk_buff_head *xmitq)
1317 struct tipc_link *bcl = l->bc_rcvlink;
1318 struct sk_buff *skb;
1319 struct tipc_msg *hdr;
1320 struct sk_buff_head *dfq = &l->deferdq;
1321 bool node_up = link_is_up(bcl);
1322 struct tipc_mon_state *mstate = &l->mon_state;
1326 /* Don't send protocol message during reset or link failover */
1327 if (tipc_link_is_blocked(l))
1330 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1333 if (!skb_queue_empty(dfq))
1334 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1336 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1337 tipc_max_domain_size, l->addr,
1338 tipc_own_addr(l->net), 0, 0, 0);
1343 data = msg_data(hdr);
1344 msg_set_session(hdr, l->session);
1345 msg_set_bearer_id(hdr, l->bearer_id);
1346 msg_set_net_plane(hdr, l->net_plane);
1347 msg_set_next_sent(hdr, l->snd_nxt);
1348 msg_set_ack(hdr, l->rcv_nxt - 1);
1349 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1350 msg_set_bc_ack_invalid(hdr, !node_up);
1351 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1352 msg_set_link_tolerance(hdr, tolerance);
1353 msg_set_linkprio(hdr, priority);
1354 msg_set_redundant_link(hdr, node_up);
1355 msg_set_seq_gap(hdr, 0);
1356 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1358 if (mtyp == STATE_MSG) {
1359 msg_set_seq_gap(hdr, rcvgap);
1360 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1361 msg_set_probe(hdr, probe);
1362 tipc_mon_prep(l->net, data, &dlen, mstate, l->bearer_id);
1363 msg_set_size(hdr, INT_H_SIZE + dlen);
1364 skb_trim(skb, INT_H_SIZE + dlen);
1365 l->stats.sent_states++;
1368 /* RESET_MSG or ACTIVATE_MSG */
1369 msg_set_max_pkt(hdr, l->advertised_mtu);
1370 strcpy(data, l->if_name);
1371 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1372 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1375 l->stats.sent_probes++;
1377 l->stats.sent_nacks++;
1378 skb->priority = TC_PRIO_CONTROL;
1379 __skb_queue_tail(xmitq, skb);
1382 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1383 * with contents of the link's transmit and backlog queues.
1385 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1386 int mtyp, struct sk_buff_head *xmitq)
1388 struct sk_buff *skb, *tnlskb;
1389 struct tipc_msg *hdr, tnlhdr;
1390 struct sk_buff_head *queue = &l->transmq;
1391 struct sk_buff_head tmpxq, tnlq;
1392 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1397 skb_queue_head_init(&tnlq);
1398 skb_queue_head_init(&tmpxq);
1400 /* At least one packet required for safe algorithm => add dummy */
1401 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1402 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1403 0, 0, TIPC_ERR_NO_PORT);
1405 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1408 skb_queue_tail(&tnlq, skb);
1409 tipc_link_xmit(l, &tnlq, &tmpxq);
1410 __skb_queue_purge(&tmpxq);
1412 /* Initialize reusable tunnel packet header */
1413 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1414 mtyp, INT_H_SIZE, l->addr);
1415 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1416 msg_set_msgcnt(&tnlhdr, pktcnt);
1417 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1419 /* Wrap each packet into a tunnel packet */
1420 skb_queue_walk(queue, skb) {
1422 if (queue == &l->backlogq)
1423 msg_set_seqno(hdr, seqno++);
1424 pktlen = msg_size(hdr);
1425 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1426 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1428 pr_warn("%sunable to send packet\n", link_co_err);
1431 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1432 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1433 __skb_queue_tail(&tnlq, tnlskb);
1435 if (queue != &l->backlogq) {
1436 queue = &l->backlogq;
1440 tipc_link_xmit(tnl, &tnlq, xmitq);
1442 if (mtyp == FAILOVER_MSG) {
1443 tnl->drop_point = l->rcv_nxt;
1444 tnl->failover_reasm_skb = l->reasm_buf;
1445 l->reasm_buf = NULL;
1449 /* tipc_link_proto_rcv(): receive link level protocol message :
1450 * Note that network plane id propagates through the network, and may
1451 * change at any time. The node with lowest numerical id determines
1454 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1455 struct sk_buff_head *xmitq)
1457 struct tipc_msg *hdr = buf_msg(skb);
1459 u16 ack = msg_ack(hdr);
1460 u16 gap = msg_seq_gap(hdr);
1461 u16 peers_snd_nxt = msg_next_sent(hdr);
1462 u16 peers_tol = msg_link_tolerance(hdr);
1463 u16 peers_prio = msg_linkprio(hdr);
1464 u16 rcv_nxt = l->rcv_nxt;
1465 u32 dlen = msg_data_sz(hdr);
1466 int mtyp = msg_type(hdr);
1474 if (tipc_link_is_blocked(l) || !xmitq)
1477 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1478 l->net_plane = msg_net_plane(hdr);
1482 data = msg_data(hdr);
1487 /* Ignore duplicate RESET with old session number */
1488 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1489 (l->peer_session != ANY_SESSION))
1495 /* Complete own link name with peer's interface name */
1496 if_name = strrchr(l->name, ':') + 1;
1497 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1499 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1501 strncpy(if_name, data, TIPC_MAX_IF_NAME);
1503 /* Update own tolerance if peer indicates a non-zero value */
1504 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1505 l->tolerance = peers_tol;
1507 /* Update own priority if peer's priority is higher */
1508 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1509 l->priority = peers_prio;
1511 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1512 if (msg_peer_stopping(hdr))
1513 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1514 else if ((mtyp == RESET_MSG) || !link_is_up(l))
1515 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1517 /* ACTIVATE_MSG takes up link if it was already locally reset */
1518 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1519 rc = TIPC_LINK_UP_EVT;
1521 l->peer_session = msg_session(hdr);
1522 l->peer_bearer_id = msg_bearer_id(hdr);
1523 if (l->mtu > msg_max_pkt(hdr))
1524 l->mtu = msg_max_pkt(hdr);
1529 /* Update own tolerance if peer indicates a non-zero value */
1530 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1531 l->tolerance = peers_tol;
1533 /* Update own prio if peer indicates a different value */
1534 if ((peers_prio != l->priority) &&
1535 in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
1536 l->priority = peers_prio;
1537 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1540 l->silent_intv_cnt = 0;
1541 l->stats.recv_states++;
1543 l->stats.recv_probes++;
1545 if (!link_is_up(l)) {
1546 if (l->state == LINK_ESTABLISHING)
1547 rc = TIPC_LINK_UP_EVT;
1550 tipc_mon_rcv(l->net, data, dlen, l->addr,
1551 &l->mon_state, l->bearer_id);
1553 /* Send NACK if peer has sent pkts we haven't received yet */
1554 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1555 rcvgap = peers_snd_nxt - l->rcv_nxt;
1556 if (rcvgap || (msg_probe(hdr)))
1557 tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1559 tipc_link_release_pkts(l, ack);
1561 /* If NACK, retransmit will now start at right position */
1563 rc = tipc_link_retrans(l, l, ack + 1, ack + gap, xmitq);
1564 l->stats.recv_nacks++;
1567 tipc_link_advance_backlog(l, xmitq);
1568 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1569 link_prepare_wakeup(l);
1576 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1578 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1580 struct sk_buff_head *xmitq)
1582 struct sk_buff *skb;
1583 struct tipc_msg *hdr;
1584 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1585 u16 ack = l->rcv_nxt - 1;
1586 u16 gap_to = peers_snd_nxt - 1;
1588 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1589 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1593 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1594 msg_set_bcast_ack(hdr, ack);
1595 msg_set_bcgap_after(hdr, ack);
1597 gap_to = buf_seqno(dfrd_skb) - 1;
1598 msg_set_bcgap_to(hdr, gap_to);
1599 msg_set_non_seq(hdr, bcast);
1600 __skb_queue_tail(xmitq, skb);
1604 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1606 * Give a newly added peer node the sequence number where it should
1607 * start receiving and acking broadcast packets.
1609 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1610 struct sk_buff_head *xmitq)
1612 struct sk_buff_head list;
1614 __skb_queue_head_init(&list);
1615 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1617 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
1618 tipc_link_xmit(l, &list, xmitq);
1621 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1623 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1625 int mtyp = msg_type(hdr);
1626 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1631 if (msg_user(hdr) == BCAST_PROTOCOL) {
1632 l->rcv_nxt = peers_snd_nxt;
1633 l->state = LINK_ESTABLISHED;
1637 if (l->peer_caps & TIPC_BCAST_SYNCH)
1640 if (msg_peer_node_is_up(hdr))
1643 /* Compatibility: accept older, less safe initial synch data */
1644 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1645 l->rcv_nxt = peers_snd_nxt;
1648 /* link_bc_retr eval()- check if the indicated range can be retransmitted now
1649 * - Adjust permitted range if there is overlap with previous retransmission
1651 static bool link_bc_retr_eval(struct tipc_link *l, u16 *from, u16 *to)
1653 unsigned long elapsed = jiffies_to_msecs(jiffies - l->prev_retr);
1655 if (less(*to, *from))
1658 /* New retransmission request */
1659 if ((elapsed > TIPC_BC_RETR_LIMIT) ||
1660 less(*to, l->prev_from) || more(*from, l->prev_to)) {
1661 l->prev_from = *from;
1663 l->prev_retr = jiffies;
1667 /* Inside range of previous retransmit */
1668 if (!less(*from, l->prev_from) && !more(*to, l->prev_to))
1671 /* Fully or partially outside previous range => exclude overlap */
1672 if (less(*from, l->prev_from)) {
1673 *to = l->prev_from - 1;
1674 l->prev_from = *from;
1676 if (more(*to, l->prev_to)) {
1677 *from = l->prev_to + 1;
1680 l->prev_retr = jiffies;
1684 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1686 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1687 struct sk_buff_head *xmitq)
1689 struct tipc_link *snd_l = l->bc_sndlink;
1690 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1691 u16 from = msg_bcast_ack(hdr) + 1;
1692 u16 to = from + msg_bc_gap(hdr) - 1;
1698 if (!msg_peer_node_is_up(hdr))
1701 /* Open when peer ackowledges our bcast init msg (pkt #1) */
1703 l->bc_peer_is_up = true;
1705 if (!l->bc_peer_is_up)
1708 l->stats.recv_nacks++;
1710 /* Ignore if peers_snd_nxt goes beyond receive window */
1711 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1714 if (link_bc_retr_eval(snd_l, &from, &to))
1715 rc = tipc_link_retrans(snd_l, l, from, to, xmitq);
1717 l->snd_nxt = peers_snd_nxt;
1718 if (link_bc_rcv_gap(l))
1719 rc |= TIPC_LINK_SND_STATE;
1721 /* Return now if sender supports nack via STATE messages */
1722 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
1725 /* Otherwise, be backwards compatible */
1727 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1728 l->nack_state = BC_NACK_SND_CONDITIONAL;
1732 /* Don't NACK if one was recently sent or peeked */
1733 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1734 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1738 /* Conditionally delay NACK sending until next synch rcv */
1739 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1740 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1741 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1745 /* Send NACK now but suppress next one */
1746 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1747 l->nack_state = BC_NACK_SND_SUPPRESS;
1751 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1752 struct sk_buff_head *xmitq)
1754 struct sk_buff *skb, *tmp;
1755 struct tipc_link *snd_l = l->bc_sndlink;
1757 if (!link_is_up(l) || !l->bc_peer_is_up)
1760 if (!more(acked, l->acked))
1763 /* Skip over packets peer has already acked */
1764 skb_queue_walk(&snd_l->transmq, skb) {
1765 if (more(buf_seqno(skb), l->acked))
1769 /* Update/release the packets peer is acking now */
1770 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1771 if (more(buf_seqno(skb), acked))
1773 if (!--TIPC_SKB_CB(skb)->ackers) {
1774 __skb_unlink(skb, &snd_l->transmq);
1779 tipc_link_advance_backlog(snd_l, xmitq);
1780 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1781 link_prepare_wakeup(snd_l);
1784 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1785 * This function is here for backwards compatibility, since
1786 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
1788 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1789 struct sk_buff_head *xmitq)
1791 struct tipc_msg *hdr = buf_msg(skb);
1792 u32 dnode = msg_destnode(hdr);
1793 int mtyp = msg_type(hdr);
1794 u16 acked = msg_bcast_ack(hdr);
1795 u16 from = acked + 1;
1796 u16 to = msg_bcgap_to(hdr);
1797 u16 peers_snd_nxt = to + 1;
1802 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1805 if (mtyp != STATE_MSG)
1808 if (dnode == tipc_own_addr(l->net)) {
1809 tipc_link_bc_ack_rcv(l, acked, xmitq);
1810 rc = tipc_link_retrans(l->bc_sndlink, l, from, to, xmitq);
1811 l->stats.recv_nacks++;
1815 /* Msg for other node => suppress own NACK at next sync if applicable */
1816 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1817 l->nack_state = BC_NACK_SND_SUPPRESS;
1822 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1824 int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1827 l->backlog[TIPC_LOW_IMPORTANCE].limit = max_t(u16, 50, win);
1828 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = max_t(u16, 100, win * 2);
1829 l->backlog[TIPC_HIGH_IMPORTANCE].limit = max_t(u16, 150, win * 3);
1830 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
1831 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1835 * link_reset_stats - reset link statistics
1836 * @l: pointer to link
1838 void tipc_link_reset_stats(struct tipc_link *l)
1840 memset(&l->stats, 0, sizeof(l->stats));
1843 static void link_print(struct tipc_link *l, const char *str)
1845 struct sk_buff *hskb = skb_peek(&l->transmq);
1846 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1847 u16 tail = l->snd_nxt - 1;
1849 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1850 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1851 skb_queue_len(&l->transmq), head, tail,
1852 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1855 /* Parse and validate nested (link) properties valid for media, bearer and link
1857 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1861 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1862 tipc_nl_prop_policy, NULL);
1866 if (props[TIPC_NLA_PROP_PRIO]) {
1869 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1870 if (prio > TIPC_MAX_LINK_PRI)
1874 if (props[TIPC_NLA_PROP_TOL]) {
1877 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1878 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1882 if (props[TIPC_NLA_PROP_WIN]) {
1885 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1886 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1893 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1896 struct nlattr *stats;
1903 struct nla_map map[] = {
1904 {TIPC_NLA_STATS_RX_INFO, 0},
1905 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1906 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1907 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1908 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1909 {TIPC_NLA_STATS_TX_INFO, 0},
1910 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1911 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1912 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1913 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1914 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1915 s->msg_length_counts : 1},
1916 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1917 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1918 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1919 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1920 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1921 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1922 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1923 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1924 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1925 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1926 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1927 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1928 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1929 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1930 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1931 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1932 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1933 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1934 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1935 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1936 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1937 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1938 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1941 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1945 for (i = 0; i < ARRAY_SIZE(map); i++)
1946 if (nla_put_u32(skb, map[i].key, map[i].val))
1949 nla_nest_end(skb, stats);
1953 nla_nest_cancel(skb, stats);
1958 /* Caller should hold appropriate locks to protect the link */
1959 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1960 struct tipc_link *link, int nlflags)
1964 struct nlattr *attrs;
1965 struct nlattr *prop;
1966 struct tipc_net *tn = net_generic(net, tipc_net_id);
1968 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1969 nlflags, TIPC_NL_LINK_GET);
1973 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1977 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1979 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1980 tipc_cluster_mask(tn->own_addr)))
1982 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1984 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
1986 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
1989 if (tipc_link_is_up(link))
1990 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1993 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
1996 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1999 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2001 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2003 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2006 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2008 nla_nest_end(msg->skb, prop);
2010 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2014 nla_nest_end(msg->skb, attrs);
2015 genlmsg_end(msg->skb, hdr);
2020 nla_nest_cancel(msg->skb, prop);
2022 nla_nest_cancel(msg->skb, attrs);
2024 genlmsg_cancel(msg->skb, hdr);
2029 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2030 struct tipc_stats *stats)
2033 struct nlattr *nest;
2040 struct nla_map map[] = {
2041 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2042 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2043 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2044 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2045 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2046 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2047 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2048 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2049 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2050 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2051 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2052 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2053 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2054 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2055 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2056 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2057 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2058 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2059 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2060 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2063 nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
2067 for (i = 0; i < ARRAY_SIZE(map); i++)
2068 if (nla_put_u32(skb, map[i].key, map[i].val))
2071 nla_nest_end(skb, nest);
2075 nla_nest_cancel(skb, nest);
2080 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2084 struct nlattr *attrs;
2085 struct nlattr *prop;
2086 struct tipc_net *tn = net_generic(net, tipc_net_id);
2087 struct tipc_link *bcl = tn->bcl;
2092 tipc_bcast_lock(net);
2094 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2095 NLM_F_MULTI, TIPC_NL_LINK_GET);
2097 tipc_bcast_unlock(net);
2101 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
2105 /* The broadcast link is always up */
2106 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2109 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2111 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2113 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2115 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2118 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2121 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2123 nla_nest_end(msg->skb, prop);
2125 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2129 tipc_bcast_unlock(net);
2130 nla_nest_end(msg->skb, attrs);
2131 genlmsg_end(msg->skb, hdr);
2136 nla_nest_cancel(msg->skb, prop);
2138 nla_nest_cancel(msg->skb, attrs);
2140 tipc_bcast_unlock(net);
2141 genlmsg_cancel(msg->skb, hdr);
2146 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2147 struct sk_buff_head *xmitq)
2150 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, tol, 0, xmitq);
2153 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2154 struct sk_buff_head *xmitq)
2157 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, prio, xmitq);
2160 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2162 l->abort_limit = limit;