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"
48 #include <linux/pkt_sched.h>
69 u32 link_congs; /* # port sends blocked by congestion */
72 u32 max_queue_sz; /* send queue size high water mark */
73 u32 accu_queue_sz; /* used for send queue size profiling */
74 u32 queue_sz_counts; /* used for send queue size profiling */
75 u32 msg_length_counts; /* used for message length profiling */
76 u32 msg_lengths_total; /* used for message length profiling */
77 u32 msg_length_profile[7]; /* used for msg. length profiling */
81 * struct tipc_link - TIPC link data structure
82 * @addr: network address of link's peer node
83 * @name: link name character string
84 * @media_addr: media address to use when sending messages over link
86 * @net: pointer to namespace struct
87 * @refcnt: reference counter for permanent references (owner node & timer)
88 * @peer_session: link session # being used by peer end of link
89 * @peer_bearer_id: bearer id used by link's peer endpoint
90 * @bearer_id: local bearer id used by link
91 * @tolerance: minimum link continuity loss needed to reset link [in ms]
92 * @abort_limit: # of unacknowledged continuity probes needed to reset link
93 * @state: current state of link FSM
94 * @peer_caps: bitmap describing capabilities of peer node
95 * @silent_intv_cnt: # of timer intervals without any reception from peer
96 * @proto_msg: template for control messages generated by link
97 * @pmsg: convenience pointer to "proto_msg" field
98 * @priority: current link priority
99 * @net_plane: current link network plane ('A' through 'H')
100 * @mon_state: cookie with information needed by link monitor
101 * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
102 * @exp_msg_count: # of tunnelled messages expected during link changeover
103 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
104 * @mtu: current maximum packet size for this link
105 * @advertised_mtu: advertised own mtu when link is being established
106 * @transmitq: queue for sent, non-acked messages
107 * @backlogq: queue for messages waiting to be sent
108 * @snt_nxt: next sequence number to use for outbound messages
109 * @ackers: # of peers that needs to ack each packet before it can be released
110 * @acked: # last packet acked by a certain peer. Used for broadcast.
111 * @rcv_nxt: next sequence number to expect for inbound messages
112 * @deferred_queue: deferred queue saved OOS b'cast message received from node
113 * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
114 * @inputq: buffer queue for messages to be delivered upwards
115 * @namedq: buffer queue for name table messages to be delivered upwards
116 * @next_out: ptr to first unsent outbound message in queue
117 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
118 * @long_msg_seq_no: next identifier to use for outbound fragmented messages
119 * @reasm_buf: head of partially reassembled inbound message fragments
120 * @bc_rcvr: marks that this is a broadcast receiver link
121 * @stats: collects statistics regarding link activity
125 char name[TIPC_MAX_LINK_NAME];
128 /* Management and link supervision data */
142 char if_name[TIPC_MAX_IF_NAME];
145 struct tipc_mon_state mon_state;
150 struct sk_buff *failover_reasm_skb;
151 struct sk_buff_head failover_deferdq;
153 /* Max packet negotiation */
158 struct sk_buff_head transmq;
159 struct sk_buff_head backlogq;
163 struct sk_buff *target_bskb;
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;
180 struct sk_buff *reasm_tnlmsg;
185 struct tipc_link *bc_rcvlink;
186 struct tipc_link *bc_sndlink;
191 struct tipc_stats stats;
195 * Error message prefixes
197 static const char *link_co_err = "Link tunneling error, ";
198 static const char *link_rst_msg = "Resetting link ";
200 /* Send states for broadcast NACKs
203 BC_NACK_SND_CONDITIONAL,
204 BC_NACK_SND_UNCONDITIONAL,
205 BC_NACK_SND_SUPPRESS,
208 #define TIPC_BC_RETR_LIM (jiffies + msecs_to_jiffies(10))
209 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
212 * Interval between NACKs when packets arrive out of order
214 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
219 LINK_ESTABLISHED = 0xe,
220 LINK_ESTABLISHING = 0xe << 4,
221 LINK_RESET = 0x1 << 8,
222 LINK_RESETTING = 0x2 << 12,
223 LINK_PEER_RESET = 0xd << 16,
224 LINK_FAILINGOVER = 0xf << 20,
225 LINK_SYNCHING = 0xc << 24
228 /* Link FSM state checking routines
230 static int link_is_up(struct tipc_link *l)
232 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
235 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
236 struct sk_buff_head *xmitq);
237 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
238 bool probe_reply, u16 rcvgap,
239 int tolerance, int priority,
240 struct sk_buff_head *xmitq);
241 static void link_print(struct tipc_link *l, const char *str);
242 static int tipc_link_build_nack_msg(struct tipc_link *l,
243 struct sk_buff_head *xmitq);
244 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
245 struct sk_buff_head *xmitq);
246 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
247 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data);
248 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
249 struct tipc_gap_ack_blks *ga,
250 struct sk_buff_head *xmitq);
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 void tipc_link_set_active(struct tipc_link *l, bool active)
305 u32 tipc_link_id(struct tipc_link *l)
307 return l->peer_bearer_id << 16 | l->bearer_id;
310 int tipc_link_window(struct tipc_link *l)
315 int tipc_link_prio(struct tipc_link *l)
320 unsigned long tipc_link_tolerance(struct tipc_link *l)
325 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
330 char tipc_link_plane(struct tipc_link *l)
335 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
337 l->peer_caps = capabilities;
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 trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
363 tipc_link_reset(rcv_l);
364 rcv_l->state = LINK_RESET;
365 if (!snd_l->ackers) {
366 trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
367 tipc_link_reset(snd_l);
368 snd_l->state = LINK_RESET;
369 __skb_queue_purge(xmitq);
373 int tipc_link_bc_peers(struct tipc_link *l)
378 static u16 link_bc_rcv_gap(struct tipc_link *l)
380 struct sk_buff *skb = skb_peek(&l->deferdq);
383 if (more(l->snd_nxt, l->rcv_nxt))
384 gap = l->snd_nxt - l->rcv_nxt;
386 gap = buf_seqno(skb) - l->rcv_nxt;
390 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
395 int tipc_link_mtu(struct tipc_link *l)
400 u16 tipc_link_rcv_nxt(struct tipc_link *l)
405 u16 tipc_link_acked(struct tipc_link *l)
410 char *tipc_link_name(struct tipc_link *l)
415 u32 tipc_link_state(struct tipc_link *l)
421 * tipc_link_create - create a new link
422 * @n: pointer to associated node
423 * @if_name: associated interface name
424 * @bearer_id: id (index) of associated bearer
425 * @tolerance: link tolerance to be used by link
426 * @net_plane: network plane (A,B,c..) this link belongs to
427 * @mtu: mtu to be advertised by link
428 * @priority: priority to be used by link
429 * @window: send window to be used by link
430 * @session: session to be used by link
431 * @ownnode: identity of own node
432 * @peer: node id of peer node
433 * @peer_caps: bitmap describing peer node capabilities
434 * @bc_sndlink: the namespace global link used for broadcast sending
435 * @bc_rcvlink: the peer specific link used for broadcast reception
436 * @inputq: queue to put messages ready for delivery
437 * @namedq: queue to put binding table update messages ready for delivery
438 * @link: return value, pointer to put the created link
440 * Returns true if link was created, otherwise false
442 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
443 int tolerance, char net_plane, u32 mtu, int priority,
444 int window, u32 session, u32 self,
445 u32 peer, u8 *peer_id, u16 peer_caps,
446 struct tipc_link *bc_sndlink,
447 struct tipc_link *bc_rcvlink,
448 struct sk_buff_head *inputq,
449 struct sk_buff_head *namedq,
450 struct tipc_link **link)
452 char peer_str[NODE_ID_STR_LEN] = {0,};
453 char self_str[NODE_ID_STR_LEN] = {0,};
456 l = kzalloc(sizeof(*l), GFP_ATOMIC);
460 l->session = session;
462 /* Set link name for unicast links only */
464 tipc_nodeid2string(self_str, tipc_own_id(net));
465 if (strlen(self_str) > 16)
466 sprintf(self_str, "%x", self);
467 tipc_nodeid2string(peer_str, peer_id);
468 if (strlen(peer_str) > 16)
469 sprintf(peer_str, "%x", peer);
471 /* Peer i/f name will be completed by reset/activate message */
472 snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
473 self_str, if_name, peer_str);
475 strcpy(l->if_name, if_name);
477 l->peer_caps = peer_caps;
479 l->in_session = false;
480 l->bearer_id = bearer_id;
481 l->tolerance = tolerance;
483 bc_rcvlink->tolerance = tolerance;
484 l->net_plane = net_plane;
485 l->advertised_mtu = mtu;
487 l->priority = priority;
488 tipc_link_set_queue_limits(l, window);
490 l->bc_sndlink = bc_sndlink;
491 l->bc_rcvlink = bc_rcvlink;
494 l->state = LINK_RESETTING;
495 __skb_queue_head_init(&l->transmq);
496 __skb_queue_head_init(&l->backlogq);
497 __skb_queue_head_init(&l->deferdq);
498 __skb_queue_head_init(&l->failover_deferdq);
499 skb_queue_head_init(&l->wakeupq);
500 skb_queue_head_init(l->inputq);
505 * tipc_link_bc_create - create new link to be used for broadcast
506 * @n: pointer to associated node
507 * @mtu: mtu to be used initially if no peers
508 * @window: send window to be used
509 * @inputq: queue to put messages ready for delivery
510 * @namedq: queue to put binding table update messages ready for delivery
511 * @link: return value, pointer to put the created link
513 * Returns true if link was created, otherwise false
515 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
516 int mtu, int window, u16 peer_caps,
517 struct sk_buff_head *inputq,
518 struct sk_buff_head *namedq,
519 struct tipc_link *bc_sndlink,
520 struct tipc_link **link)
524 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
525 0, ownnode, peer, NULL, peer_caps, bc_sndlink,
526 NULL, inputq, namedq, link))
530 strcpy(l->name, tipc_bclink_name);
531 trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
533 l->state = LINK_RESET;
537 /* Broadcast send link is always up */
538 if (link_is_bc_sndlink(l))
539 l->state = LINK_ESTABLISHED;
541 /* Disable replicast if even a single peer doesn't support it */
542 if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
543 tipc_bcast_disable_rcast(net);
549 * tipc_link_fsm_evt - link finite state machine
550 * @l: pointer to link
551 * @evt: state machine event to be processed
553 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
556 int old_state = l->state;
561 case LINK_PEER_RESET_EVT:
562 l->state = LINK_PEER_RESET;
565 l->state = LINK_RESET;
567 case LINK_FAILURE_EVT:
568 case LINK_FAILOVER_BEGIN_EVT:
569 case LINK_ESTABLISH_EVT:
570 case LINK_FAILOVER_END_EVT:
571 case LINK_SYNCH_BEGIN_EVT:
572 case LINK_SYNCH_END_EVT:
579 case LINK_PEER_RESET_EVT:
580 l->state = LINK_ESTABLISHING;
582 case LINK_FAILOVER_BEGIN_EVT:
583 l->state = LINK_FAILINGOVER;
584 case LINK_FAILURE_EVT:
586 case LINK_ESTABLISH_EVT:
587 case LINK_FAILOVER_END_EVT:
589 case LINK_SYNCH_BEGIN_EVT:
590 case LINK_SYNCH_END_EVT:
595 case LINK_PEER_RESET:
598 l->state = LINK_ESTABLISHING;
600 case LINK_PEER_RESET_EVT:
601 case LINK_ESTABLISH_EVT:
602 case LINK_FAILURE_EVT:
604 case LINK_SYNCH_BEGIN_EVT:
605 case LINK_SYNCH_END_EVT:
606 case LINK_FAILOVER_BEGIN_EVT:
607 case LINK_FAILOVER_END_EVT:
612 case LINK_FAILINGOVER:
614 case LINK_FAILOVER_END_EVT:
615 l->state = LINK_RESET;
617 case LINK_PEER_RESET_EVT:
619 case LINK_ESTABLISH_EVT:
620 case LINK_FAILURE_EVT:
622 case LINK_FAILOVER_BEGIN_EVT:
623 case LINK_SYNCH_BEGIN_EVT:
624 case LINK_SYNCH_END_EVT:
629 case LINK_ESTABLISHING:
631 case LINK_ESTABLISH_EVT:
632 l->state = LINK_ESTABLISHED;
634 case LINK_FAILOVER_BEGIN_EVT:
635 l->state = LINK_FAILINGOVER;
638 l->state = LINK_RESET;
640 case LINK_FAILURE_EVT:
641 case LINK_PEER_RESET_EVT:
642 case LINK_SYNCH_BEGIN_EVT:
643 case LINK_FAILOVER_END_EVT:
645 case LINK_SYNCH_END_EVT:
650 case LINK_ESTABLISHED:
652 case LINK_PEER_RESET_EVT:
653 l->state = LINK_PEER_RESET;
654 rc |= TIPC_LINK_DOWN_EVT;
656 case LINK_FAILURE_EVT:
657 l->state = LINK_RESETTING;
658 rc |= TIPC_LINK_DOWN_EVT;
661 l->state = LINK_RESET;
663 case LINK_ESTABLISH_EVT:
664 case LINK_SYNCH_END_EVT:
666 case LINK_SYNCH_BEGIN_EVT:
667 l->state = LINK_SYNCHING;
669 case LINK_FAILOVER_BEGIN_EVT:
670 case LINK_FAILOVER_END_EVT:
677 case LINK_PEER_RESET_EVT:
678 l->state = LINK_PEER_RESET;
679 rc |= TIPC_LINK_DOWN_EVT;
681 case LINK_FAILURE_EVT:
682 l->state = LINK_RESETTING;
683 rc |= TIPC_LINK_DOWN_EVT;
686 l->state = LINK_RESET;
688 case LINK_ESTABLISH_EVT:
689 case LINK_SYNCH_BEGIN_EVT:
691 case LINK_SYNCH_END_EVT:
692 l->state = LINK_ESTABLISHED;
694 case LINK_FAILOVER_BEGIN_EVT:
695 case LINK_FAILOVER_END_EVT:
701 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
703 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
706 pr_err("Illegal FSM event %x in state %x on link %s\n",
707 evt, l->state, l->name);
708 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
712 /* link_profile_stats - update statistical profiling of traffic
714 static void link_profile_stats(struct tipc_link *l)
717 struct tipc_msg *msg;
720 /* Update counters used in statistical profiling of send traffic */
721 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
722 l->stats.queue_sz_counts++;
724 skb = skb_peek(&l->transmq);
728 length = msg_size(msg);
730 if (msg_user(msg) == MSG_FRAGMENTER) {
731 if (msg_type(msg) != FIRST_FRAGMENT)
733 length = msg_size(msg_inner_hdr(msg));
735 l->stats.msg_lengths_total += length;
736 l->stats.msg_length_counts++;
738 l->stats.msg_length_profile[0]++;
739 else if (length <= 256)
740 l->stats.msg_length_profile[1]++;
741 else if (length <= 1024)
742 l->stats.msg_length_profile[2]++;
743 else if (length <= 4096)
744 l->stats.msg_length_profile[3]++;
745 else if (length <= 16384)
746 l->stats.msg_length_profile[4]++;
747 else if (length <= 32768)
748 l->stats.msg_length_profile[5]++;
750 l->stats.msg_length_profile[6]++;
754 * tipc_link_too_silent - check if link is "too silent"
755 * @l: tipc link to be checked
757 * Returns true if the link 'silent_intv_cnt' is about to reach the
758 * 'abort_limit' value, otherwise false
760 bool tipc_link_too_silent(struct tipc_link *l)
762 return (l->silent_intv_cnt + 2 > l->abort_limit);
765 /* tipc_link_timeout - perform periodic task as instructed from node timeout
767 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
774 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
775 u16 bc_acked = l->bc_rcvlink->acked;
776 struct tipc_mon_state *mstate = &l->mon_state;
778 trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
779 trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
781 case LINK_ESTABLISHED:
784 link_profile_stats(l);
785 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
786 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
787 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
788 state = bc_acked != bc_snt;
789 state |= l->bc_rcvlink->rcv_unacked;
790 state |= l->rcv_unacked;
791 state |= !skb_queue_empty(&l->transmq);
792 state |= !skb_queue_empty(&l->deferdq);
793 probe = mstate->probing;
794 probe |= l->silent_intv_cnt;
795 if (probe || mstate->monitoring)
796 l->silent_intv_cnt++;
799 setup = l->rst_cnt++ <= 4;
800 setup |= !(l->rst_cnt % 16);
803 case LINK_ESTABLISHING:
807 case LINK_PEER_RESET:
809 case LINK_FAILINGOVER:
815 if (state || probe || setup)
816 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
822 * link_schedule_user - schedule a message sender for wakeup after congestion
824 * @hdr: header of message that is being sent
825 * Create pseudo msg to send back to user when congestion abates
827 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
829 u32 dnode = tipc_own_addr(l->net);
830 u32 dport = msg_origport(hdr);
833 /* Create and schedule wakeup pseudo message */
834 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
835 dnode, l->addr, dport, 0, 0);
838 msg_set_dest_droppable(buf_msg(skb), true);
839 TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
840 skb_queue_tail(&l->wakeupq, skb);
841 l->stats.link_congs++;
842 trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
847 * link_prepare_wakeup - prepare users for wakeup after congestion
849 * Wake up a number of waiting users, as permitted by available space
852 static void link_prepare_wakeup(struct tipc_link *l)
854 struct sk_buff_head *wakeupq = &l->wakeupq;
855 struct sk_buff_head *inputq = l->inputq;
856 struct sk_buff *skb, *tmp;
857 struct sk_buff_head tmpq;
861 __skb_queue_head_init(&tmpq);
863 for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
864 avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
866 skb_queue_walk_safe(wakeupq, skb, tmp) {
867 imp = TIPC_SKB_CB(skb)->chain_imp;
871 __skb_unlink(skb, wakeupq);
872 __skb_queue_tail(&tmpq, skb);
875 spin_lock_bh(&inputq->lock);
876 skb_queue_splice_tail(&tmpq, inputq);
877 spin_unlock_bh(&inputq->lock);
881 void tipc_link_reset(struct tipc_link *l)
883 struct sk_buff_head list;
886 __skb_queue_head_init(&list);
888 l->in_session = false;
889 /* Force re-synch of peer session number before establishing */
892 l->mtu = l->advertised_mtu;
894 spin_lock_bh(&l->wakeupq.lock);
895 skb_queue_splice_init(&l->wakeupq, &list);
896 spin_unlock_bh(&l->wakeupq.lock);
898 spin_lock_bh(&l->inputq->lock);
899 skb_queue_splice_init(&list, l->inputq);
900 spin_unlock_bh(&l->inputq->lock);
902 __skb_queue_purge(&l->transmq);
903 __skb_queue_purge(&l->deferdq);
904 __skb_queue_purge(&l->backlogq);
905 __skb_queue_purge(&l->failover_deferdq);
906 for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
907 l->backlog[imp].len = 0;
908 l->backlog[imp].target_bskb = NULL;
910 kfree_skb(l->reasm_buf);
911 kfree_skb(l->reasm_tnlmsg);
912 kfree_skb(l->failover_reasm_skb);
914 l->reasm_tnlmsg = NULL;
915 l->failover_reasm_skb = NULL;
919 l->snd_nxt_state = 1;
920 l->rcv_nxt_state = 1;
922 l->silent_intv_cnt = 0;
924 l->bc_peer_is_up = false;
925 memset(&l->mon_state, 0, sizeof(l->mon_state));
926 tipc_link_reset_stats(l);
930 * tipc_link_xmit(): enqueue buffer list according to queue situation
932 * @list: chain of buffers containing message
933 * @xmitq: returned list of packets to be sent by caller
935 * Consumes the buffer chain.
936 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
937 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
939 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
940 struct sk_buff_head *xmitq)
942 struct tipc_msg *hdr = buf_msg(skb_peek(list));
943 unsigned int maxwin = l->window;
944 int imp = msg_importance(hdr);
945 unsigned int mtu = l->mtu;
946 u16 ack = l->rcv_nxt - 1;
947 u16 seqno = l->snd_nxt;
948 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
949 struct sk_buff_head *transmq = &l->transmq;
950 struct sk_buff_head *backlogq = &l->backlogq;
951 struct sk_buff *skb, *_skb, **tskb;
952 int pkt_cnt = skb_queue_len(list);
955 if (unlikely(msg_size(hdr) > mtu)) {
956 pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
957 skb_queue_len(list), msg_user(hdr),
958 msg_type(hdr), msg_size(hdr), mtu);
959 __skb_queue_purge(list);
963 /* Allow oversubscription of one data msg per source at congestion */
964 if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
965 if (imp == TIPC_SYSTEM_IMPORTANCE) {
966 pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
969 rc = link_schedule_user(l, hdr);
973 l->stats.sent_fragmented++;
974 l->stats.sent_fragments += pkt_cnt;
977 /* Prepare each packet for sending, and add to relevant queue: */
978 while (skb_queue_len(list)) {
979 skb = skb_peek(list);
981 msg_set_seqno(hdr, seqno);
982 msg_set_ack(hdr, ack);
983 msg_set_bcast_ack(hdr, bc_ack);
985 if (likely(skb_queue_len(transmq) < maxwin)) {
986 _skb = skb_clone(skb, GFP_ATOMIC);
988 __skb_queue_purge(list);
992 __skb_queue_tail(transmq, skb);
993 /* next retransmit attempt */
994 if (link_is_bc_sndlink(l))
995 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
996 __skb_queue_tail(xmitq, _skb);
997 TIPC_SKB_CB(skb)->ackers = l->ackers;
999 l->stats.sent_pkts++;
1003 tskb = &l->backlog[imp].target_bskb;
1004 if (tipc_msg_bundle(*tskb, hdr, mtu)) {
1005 kfree_skb(__skb_dequeue(list));
1006 l->stats.sent_bundled++;
1009 if (tipc_msg_make_bundle(tskb, hdr, mtu, l->addr)) {
1010 kfree_skb(__skb_dequeue(list));
1011 __skb_queue_tail(backlogq, *tskb);
1012 l->backlog[imp].len++;
1013 l->stats.sent_bundled++;
1014 l->stats.sent_bundles++;
1017 l->backlog[imp].target_bskb = NULL;
1018 l->backlog[imp].len += skb_queue_len(list);
1019 skb_queue_splice_tail_init(list, backlogq);
1025 static void tipc_link_advance_backlog(struct tipc_link *l,
1026 struct sk_buff_head *xmitq)
1028 struct sk_buff *skb, *_skb;
1029 struct tipc_msg *hdr;
1030 u16 seqno = l->snd_nxt;
1031 u16 ack = l->rcv_nxt - 1;
1032 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1035 while (skb_queue_len(&l->transmq) < l->window) {
1036 skb = skb_peek(&l->backlogq);
1039 _skb = skb_clone(skb, GFP_ATOMIC);
1042 __skb_dequeue(&l->backlogq);
1044 imp = msg_importance(hdr);
1045 l->backlog[imp].len--;
1046 if (unlikely(skb == l->backlog[imp].target_bskb))
1047 l->backlog[imp].target_bskb = NULL;
1048 __skb_queue_tail(&l->transmq, skb);
1049 /* next retransmit attempt */
1050 if (link_is_bc_sndlink(l))
1051 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1053 __skb_queue_tail(xmitq, _skb);
1054 TIPC_SKB_CB(skb)->ackers = l->ackers;
1055 msg_set_seqno(hdr, seqno);
1056 msg_set_ack(hdr, ack);
1057 msg_set_bcast_ack(hdr, bc_ack);
1059 l->stats.sent_pkts++;
1066 * link_retransmit_failure() - Detect repeated retransmit failures
1067 * @l: tipc link sender
1068 * @r: tipc link receiver (= l in case of unicast)
1069 * @rc: returned code
1071 * Return: true if the repeated retransmit failures happens, otherwise
1074 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1077 struct sk_buff *skb = skb_peek(&l->transmq);
1078 struct tipc_msg *hdr;
1083 if (!TIPC_SKB_CB(skb)->retr_cnt)
1086 if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1087 msecs_to_jiffies(r->tolerance)))
1091 if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1094 pr_warn("Retransmission failure on link <%s>\n", l->name);
1095 link_print(l, "State of link ");
1096 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1097 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1098 pr_info("sqno %u, prev: %x, dest: %x\n",
1099 msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1100 pr_info("retr_stamp %d, retr_cnt %d\n",
1101 jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1102 TIPC_SKB_CB(skb)->retr_cnt);
1104 trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1105 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1106 trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1108 if (link_is_bc_sndlink(l)) {
1109 r->state = LINK_RESET;
1110 *rc = TIPC_LINK_DOWN_EVT;
1112 *rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1118 /* tipc_link_bc_retrans() - retransmit zero or more packets
1119 * @l: the link to transmit on
1120 * @r: the receiving link ordering the retransmit. Same as l if unicast
1121 * @from: retransmit from (inclusive) this sequence number
1122 * @to: retransmit to (inclusive) this sequence number
1123 * xmitq: queue for accumulating the retransmitted packets
1125 static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r,
1126 u16 from, u16 to, struct sk_buff_head *xmitq)
1128 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1129 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1130 u16 ack = l->rcv_nxt - 1;
1131 struct tipc_msg *hdr;
1139 trace_tipc_link_retrans(r, from, to, &l->transmq);
1141 if (link_retransmit_failure(l, r, &rc))
1144 skb_queue_walk(&l->transmq, skb) {
1146 if (less(msg_seqno(hdr), from))
1148 if (more(msg_seqno(hdr), to))
1151 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1153 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1154 _skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE, GFP_ATOMIC);
1157 hdr = buf_msg(_skb);
1158 msg_set_ack(hdr, ack);
1159 msg_set_bcast_ack(hdr, bc_ack);
1160 _skb->priority = TC_PRIO_CONTROL;
1161 __skb_queue_tail(xmitq, _skb);
1162 l->stats.retransmitted++;
1164 /* Increase actual retrans counter & mark first time */
1165 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1166 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1171 /* tipc_data_input - deliver data and name distr msgs to upper layer
1173 * Consumes buffer if message is of right type
1174 * Node lock must be held
1176 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1177 struct sk_buff_head *inputq)
1179 struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1180 struct tipc_msg *hdr = buf_msg(skb);
1182 switch (msg_user(hdr)) {
1183 case TIPC_LOW_IMPORTANCE:
1184 case TIPC_MEDIUM_IMPORTANCE:
1185 case TIPC_HIGH_IMPORTANCE:
1186 case TIPC_CRITICAL_IMPORTANCE:
1187 if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1188 skb_queue_tail(mc_inputq, skb);
1193 skb_queue_tail(inputq, skb);
1195 case GROUP_PROTOCOL:
1196 skb_queue_tail(mc_inputq, skb);
1198 case NAME_DISTRIBUTOR:
1199 l->bc_rcvlink->state = LINK_ESTABLISHED;
1200 skb_queue_tail(l->namedq, skb);
1203 case TUNNEL_PROTOCOL:
1204 case MSG_FRAGMENTER:
1205 case BCAST_PROTOCOL:
1208 pr_warn("Dropping received illegal msg type\n");
1214 /* tipc_link_input - process packet that has passed link protocol check
1218 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1219 struct sk_buff_head *inputq,
1220 struct sk_buff **reasm_skb)
1222 struct tipc_msg *hdr = buf_msg(skb);
1223 struct sk_buff *iskb;
1224 struct sk_buff_head tmpq;
1225 int usr = msg_user(hdr);
1228 if (usr == MSG_BUNDLER) {
1229 skb_queue_head_init(&tmpq);
1230 l->stats.recv_bundles++;
1231 l->stats.recv_bundled += msg_msgcnt(hdr);
1232 while (tipc_msg_extract(skb, &iskb, &pos))
1233 tipc_data_input(l, iskb, &tmpq);
1234 tipc_skb_queue_splice_tail(&tmpq, inputq);
1236 } else if (usr == MSG_FRAGMENTER) {
1237 l->stats.recv_fragments++;
1238 if (tipc_buf_append(reasm_skb, &skb)) {
1239 l->stats.recv_fragmented++;
1240 tipc_data_input(l, skb, inputq);
1241 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1242 pr_warn_ratelimited("Unable to build fragment list\n");
1243 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1246 } else if (usr == BCAST_PROTOCOL) {
1247 tipc_bcast_lock(l->net);
1248 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1249 tipc_bcast_unlock(l->net);
1256 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1257 * inner message along with the ones in the old link's
1260 * @skb: TUNNEL_PROTOCOL message
1261 * @inputq: queue to put messages ready for delivery
1263 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1264 struct sk_buff_head *inputq)
1266 struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1267 struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1268 struct sk_buff_head *fdefq = &l->failover_deferdq;
1269 struct tipc_msg *hdr = buf_msg(skb);
1270 struct sk_buff *iskb;
1275 if (msg_type(hdr) == SYNCH_MSG) {
1280 /* Not a fragment? */
1281 if (likely(!msg_nof_fragms(hdr))) {
1282 if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1283 pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1284 skb_queue_len(fdefq));
1289 /* Set fragment type for buf_append */
1290 if (msg_fragm_no(hdr) == 1)
1291 msg_set_type(hdr, FIRST_FRAGMENT);
1292 else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1293 msg_set_type(hdr, FRAGMENT);
1295 msg_set_type(hdr, LAST_FRAGMENT);
1297 if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1298 /* Successful but non-complete reassembly? */
1299 if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1301 pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1302 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1308 seqno = buf_seqno(iskb);
1309 if (unlikely(less(seqno, l->drop_point))) {
1313 if (unlikely(seqno != l->drop_point)) {
1314 __tipc_skb_queue_sorted(fdefq, seqno, iskb);
1319 if (!tipc_data_input(l, iskb, inputq))
1320 rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1323 } while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1328 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1330 bool released = false;
1331 struct sk_buff *skb, *tmp;
1333 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1334 if (more(buf_seqno(skb), acked))
1336 __skb_unlink(skb, &l->transmq);
1343 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1344 * @l: tipc link that data have come with gaps in sequence if any
1345 * @data: data buffer to store the Gap ACK blocks after built
1347 * returns the actual allocated memory size
1349 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data)
1351 struct sk_buff *skb = skb_peek(&l->deferdq);
1352 struct tipc_gap_ack_blks *ga = data;
1353 u16 len, expect, seqno = 0;
1359 expect = buf_seqno(skb);
1360 skb_queue_walk(&l->deferdq, skb) {
1361 seqno = buf_seqno(skb);
1362 if (unlikely(more(seqno, expect))) {
1363 ga->gacks[n].ack = htons(expect - 1);
1364 ga->gacks[n].gap = htons(seqno - expect);
1365 if (++n >= MAX_GAP_ACK_BLKS) {
1366 pr_info_ratelimited("Too few Gap ACK blocks!\n");
1369 } else if (unlikely(less(seqno, expect))) {
1370 pr_warn("Unexpected skb in deferdq!\n");
1377 ga->gacks[n].ack = htons(seqno);
1378 ga->gacks[n].gap = 0;
1382 len = tipc_gap_ack_blks_sz(n);
1383 ga->len = htons(len);
1388 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1389 * acked packets, also doing retransmissions if
1391 * @l: tipc link with transmq queue to be advanced
1392 * @acked: seqno of last packet acked by peer without any gaps before
1393 * @gap: # of gap packets
1394 * @ga: buffer pointer to Gap ACK blocks from peer
1395 * @xmitq: queue for accumulating the retransmitted packets if any
1397 * In case of a repeated retransmit failures, the call will return shortly
1398 * with a returned code (e.g. TIPC_LINK_DOWN_EVT)
1400 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
1401 struct tipc_gap_ack_blks *ga,
1402 struct sk_buff_head *xmitq)
1404 struct sk_buff *skb, *_skb, *tmp;
1405 struct tipc_msg *hdr;
1406 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1407 u16 ack = l->rcv_nxt - 1;
1408 bool passed = false;
1412 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1413 seqno = buf_seqno(skb);
1416 if (less_eq(seqno, acked)) {
1418 __skb_unlink(skb, &l->transmq);
1420 } else if (less_eq(seqno, acked + gap)) {
1421 /* First, check if repeated retrans failures occurs? */
1422 if (!passed && link_retransmit_failure(l, l, &rc))
1426 /* retransmit skb if unrestricted*/
1427 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1429 TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
1430 _skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE,
1434 hdr = buf_msg(_skb);
1435 msg_set_ack(hdr, ack);
1436 msg_set_bcast_ack(hdr, bc_ack);
1437 _skb->priority = TC_PRIO_CONTROL;
1438 __skb_queue_tail(xmitq, _skb);
1439 l->stats.retransmitted++;
1441 /* Increase actual retrans counter & mark first time */
1442 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1443 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1445 /* retry with Gap ACK blocks if any */
1446 if (!ga || n >= ga->gack_cnt)
1448 acked = ntohs(ga->gacks[n].ack);
1449 gap = ntohs(ga->gacks[n].gap);
1458 /* tipc_link_build_state_msg: prepare link state message for transmission
1460 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1461 * risk of ack storms towards the sender
1463 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1468 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1469 if (link_is_bc_rcvlink(l)) {
1470 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1474 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1475 l->snd_nxt = l->rcv_nxt;
1476 return TIPC_LINK_SND_STATE;
1481 l->stats.sent_acks++;
1482 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1486 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1488 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1490 int mtyp = RESET_MSG;
1491 struct sk_buff *skb;
1493 if (l->state == LINK_ESTABLISHING)
1494 mtyp = ACTIVATE_MSG;
1496 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1498 /* Inform peer that this endpoint is going down if applicable */
1499 skb = skb_peek_tail(xmitq);
1500 if (skb && (l->state == LINK_RESET))
1501 msg_set_peer_stopping(buf_msg(skb), 1);
1504 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1505 * Note that sending of broadcast NACK is coordinated among nodes, to
1506 * reduce the risk of NACK storms towards the sender
1508 static int tipc_link_build_nack_msg(struct tipc_link *l,
1509 struct sk_buff_head *xmitq)
1511 u32 def_cnt = ++l->stats.deferred_recv;
1512 u32 defq_len = skb_queue_len(&l->deferdq);
1515 if (link_is_bc_rcvlink(l)) {
1516 match1 = def_cnt & 0xf;
1517 match2 = tipc_own_addr(l->net) & 0xf;
1518 if (match1 == match2)
1519 return TIPC_LINK_SND_STATE;
1523 if (defq_len >= 3 && !((defq_len - 3) % 16))
1524 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1528 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1529 * @l: the link that should handle the message
1531 * @xmitq: queue to place packets to be sent after this call
1533 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1534 struct sk_buff_head *xmitq)
1536 struct sk_buff_head *defq = &l->deferdq;
1537 struct tipc_msg *hdr = buf_msg(skb);
1538 u16 seqno, rcv_nxt, win_lim;
1541 /* Verify and update link state */
1542 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1543 return tipc_link_proto_rcv(l, skb, xmitq);
1545 /* Don't send probe at next timeout expiration */
1546 l->silent_intv_cnt = 0;
1550 seqno = msg_seqno(hdr);
1551 rcv_nxt = l->rcv_nxt;
1552 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1554 if (unlikely(!link_is_up(l))) {
1555 if (l->state == LINK_ESTABLISHING)
1556 rc = TIPC_LINK_UP_EVT;
1560 /* Drop if outside receive window */
1561 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1562 l->stats.duplicates++;
1566 /* Forward queues and wake up waiting users */
1567 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1568 tipc_link_advance_backlog(l, xmitq);
1569 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1570 link_prepare_wakeup(l);
1573 /* Defer delivery if sequence gap */
1574 if (unlikely(seqno != rcv_nxt)) {
1575 __tipc_skb_queue_sorted(defq, seqno, skb);
1576 rc |= tipc_link_build_nack_msg(l, xmitq);
1580 /* Deliver packet */
1582 l->stats.recv_pkts++;
1584 if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1585 rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1586 else if (!tipc_data_input(l, skb, l->inputq))
1587 rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1588 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1589 rc |= tipc_link_build_state_msg(l, xmitq);
1590 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1592 } while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1600 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1601 bool probe_reply, u16 rcvgap,
1602 int tolerance, int priority,
1603 struct sk_buff_head *xmitq)
1605 struct tipc_link *bcl = l->bc_rcvlink;
1606 struct sk_buff *skb;
1607 struct tipc_msg *hdr;
1608 struct sk_buff_head *dfq = &l->deferdq;
1609 bool node_up = link_is_up(bcl);
1610 struct tipc_mon_state *mstate = &l->mon_state;
1615 /* Don't send protocol message during reset or link failover */
1616 if (tipc_link_is_blocked(l))
1619 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1622 if (!skb_queue_empty(dfq))
1623 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1625 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1626 tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1627 l->addr, tipc_own_addr(l->net), 0, 0, 0);
1632 data = msg_data(hdr);
1633 msg_set_session(hdr, l->session);
1634 msg_set_bearer_id(hdr, l->bearer_id);
1635 msg_set_net_plane(hdr, l->net_plane);
1636 msg_set_next_sent(hdr, l->snd_nxt);
1637 msg_set_ack(hdr, l->rcv_nxt - 1);
1638 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1639 msg_set_bc_ack_invalid(hdr, !node_up);
1640 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1641 msg_set_link_tolerance(hdr, tolerance);
1642 msg_set_linkprio(hdr, priority);
1643 msg_set_redundant_link(hdr, node_up);
1644 msg_set_seq_gap(hdr, 0);
1645 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1647 if (mtyp == STATE_MSG) {
1648 if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1649 msg_set_seqno(hdr, l->snd_nxt_state++);
1650 msg_set_seq_gap(hdr, rcvgap);
1651 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1652 msg_set_probe(hdr, probe);
1653 msg_set_is_keepalive(hdr, probe || probe_reply);
1654 if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1655 glen = tipc_build_gap_ack_blks(l, data);
1656 tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1657 msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1658 skb_trim(skb, INT_H_SIZE + glen + dlen);
1659 l->stats.sent_states++;
1662 /* RESET_MSG or ACTIVATE_MSG */
1663 if (mtyp == ACTIVATE_MSG) {
1664 msg_set_dest_session_valid(hdr, 1);
1665 msg_set_dest_session(hdr, l->peer_session);
1667 msg_set_max_pkt(hdr, l->advertised_mtu);
1668 strcpy(data, l->if_name);
1669 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1670 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1673 l->stats.sent_probes++;
1675 l->stats.sent_nacks++;
1676 skb->priority = TC_PRIO_CONTROL;
1677 __skb_queue_tail(xmitq, skb);
1678 trace_tipc_proto_build(skb, false, l->name);
1681 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1682 struct sk_buff_head *xmitq)
1684 u32 onode = tipc_own_addr(l->net);
1685 struct tipc_msg *hdr, *ihdr;
1686 struct sk_buff_head tnlq;
1687 struct sk_buff *skb;
1688 u32 dnode = l->addr;
1690 __skb_queue_head_init(&tnlq);
1691 skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1692 INT_H_SIZE, BASIC_H_SIZE,
1693 dnode, onode, 0, 0, 0);
1695 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1700 msg_set_msgcnt(hdr, 1);
1701 msg_set_bearer_id(hdr, l->peer_bearer_id);
1703 ihdr = (struct tipc_msg *)msg_data(hdr);
1704 tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1705 BASIC_H_SIZE, dnode);
1706 msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1707 __skb_queue_tail(&tnlq, skb);
1708 tipc_link_xmit(l, &tnlq, xmitq);
1711 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1712 * with contents of the link's transmit and backlog queues.
1714 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1715 int mtyp, struct sk_buff_head *xmitq)
1717 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1718 struct sk_buff *skb, *tnlskb;
1719 struct tipc_msg *hdr, tnlhdr;
1720 struct sk_buff_head *queue = &l->transmq;
1721 struct sk_buff_head tmpxq, tnlq, frags;
1722 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1723 bool pktcnt_need_update = false;
1730 __skb_queue_head_init(&tnlq);
1731 __skb_queue_head_init(&tmpxq);
1732 __skb_queue_head_init(&frags);
1734 /* At least one packet required for safe algorithm => add dummy */
1735 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1736 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1737 0, 0, TIPC_ERR_NO_PORT);
1739 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1742 __skb_queue_tail(&tnlq, skb);
1743 tipc_link_xmit(l, &tnlq, &tmpxq);
1744 __skb_queue_purge(&tmpxq);
1747 * From now on, send only one single ("dummy") SYNCH message
1748 * to peer. The SYNCH message does not contain any data, just
1749 * a header conveying the synch point to the peer.
1751 if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1752 tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1753 INT_H_SIZE, 0, l->addr,
1754 tipc_own_addr(l->net),
1757 pr_warn("%sunable to create dummy SYNCH_MSG\n",
1762 hdr = buf_msg(tnlskb);
1763 syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
1764 msg_set_syncpt(hdr, syncpt);
1765 msg_set_bearer_id(hdr, l->peer_bearer_id);
1766 __skb_queue_tail(&tnlq, tnlskb);
1767 tipc_link_xmit(tnl, &tnlq, xmitq);
1771 /* Initialize reusable tunnel packet header */
1772 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1773 mtyp, INT_H_SIZE, l->addr);
1774 if (mtyp == SYNCH_MSG)
1775 pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
1777 pktcnt = skb_queue_len(&l->transmq);
1778 pktcnt += skb_queue_len(&l->backlogq);
1779 msg_set_msgcnt(&tnlhdr, pktcnt);
1780 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1782 /* Wrap each packet into a tunnel packet */
1783 skb_queue_walk(queue, skb) {
1785 if (queue == &l->backlogq)
1786 msg_set_seqno(hdr, seqno++);
1787 pktlen = msg_size(hdr);
1789 /* Tunnel link MTU is not large enough? This could be
1791 * 1) Link MTU has just changed or set differently;
1792 * 2) Or FAILOVER on the top of a SYNCH message
1794 * The 2nd case should not happen if peer supports
1795 * TIPC_TUNNEL_ENHANCED
1797 if (pktlen > tnl->mtu - INT_H_SIZE) {
1798 if (mtyp == FAILOVER_MSG &&
1799 (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1800 rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
1803 pr_warn("%sunable to frag msg: rc %d\n",
1807 pktcnt += skb_queue_len(&frags) - 1;
1808 pktcnt_need_update = true;
1809 skb_queue_splice_tail_init(&frags, &tnlq);
1812 /* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
1813 * => Just warn it and return!
1815 pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
1816 link_co_err, msg_user(hdr),
1817 msg_type(hdr), msg_size(hdr));
1821 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1822 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1824 pr_warn("%sunable to send packet\n", link_co_err);
1827 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1828 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1829 __skb_queue_tail(&tnlq, tnlskb);
1831 if (queue != &l->backlogq) {
1832 queue = &l->backlogq;
1836 if (pktcnt_need_update)
1837 skb_queue_walk(&tnlq, skb) {
1839 msg_set_msgcnt(hdr, pktcnt);
1842 tipc_link_xmit(tnl, &tnlq, xmitq);
1844 if (mtyp == FAILOVER_MSG) {
1845 tnl->drop_point = l->rcv_nxt;
1846 tnl->failover_reasm_skb = l->reasm_buf;
1847 l->reasm_buf = NULL;
1849 /* Failover the link's deferdq */
1850 if (unlikely(!skb_queue_empty(fdefq))) {
1851 pr_warn("Link failover deferdq not empty: %d!\n",
1852 skb_queue_len(fdefq));
1853 __skb_queue_purge(fdefq);
1855 skb_queue_splice_init(&l->deferdq, fdefq);
1860 * tipc_link_failover_prepare() - prepare tnl for link failover
1862 * This is a special version of the precursor - tipc_link_tnl_prepare(),
1863 * see the tipc_node_link_failover() for details
1867 * @xmitq: queue for messages to be xmited
1869 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
1870 struct sk_buff_head *xmitq)
1872 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1874 tipc_link_create_dummy_tnl_msg(tnl, xmitq);
1876 /* This failover link enpoint was never established before,
1877 * so it has not received anything from peer.
1878 * Otherwise, it must be a normal failover situation or the
1879 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
1880 * would have to start over from scratch instead.
1882 tnl->drop_point = 1;
1883 tnl->failover_reasm_skb = NULL;
1885 /* Initiate the link's failover deferdq */
1886 if (unlikely(!skb_queue_empty(fdefq))) {
1887 pr_warn("Link failover deferdq not empty: %d!\n",
1888 skb_queue_len(fdefq));
1889 __skb_queue_purge(fdefq);
1893 /* tipc_link_validate_msg(): validate message against current link state
1894 * Returns true if message should be accepted, otherwise false
1896 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
1898 u16 curr_session = l->peer_session;
1899 u16 session = msg_session(hdr);
1900 int mtyp = msg_type(hdr);
1902 if (msg_user(hdr) != LINK_PROTOCOL)
1909 /* Accept only RESET with new session number */
1910 return more(session, curr_session);
1914 /* Accept only ACTIVATE with new or current session number */
1915 return !less(session, curr_session);
1917 /* Accept only STATE with current session number */
1920 if (session != curr_session)
1922 /* Extra sanity check */
1923 if (!link_is_up(l) && msg_ack(hdr))
1925 if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
1927 /* Accept only STATE with new sequence number */
1928 return !less(msg_seqno(hdr), l->rcv_nxt_state);
1934 /* tipc_link_proto_rcv(): receive link level protocol message :
1935 * Note that network plane id propagates through the network, and may
1936 * change at any time. The node with lowest numerical id determines
1939 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1940 struct sk_buff_head *xmitq)
1942 struct tipc_msg *hdr = buf_msg(skb);
1943 struct tipc_gap_ack_blks *ga = NULL;
1945 u16 ack = msg_ack(hdr);
1946 u16 gap = msg_seq_gap(hdr);
1947 u16 peers_snd_nxt = msg_next_sent(hdr);
1948 u16 peers_tol = msg_link_tolerance(hdr);
1949 u16 peers_prio = msg_linkprio(hdr);
1950 u16 rcv_nxt = l->rcv_nxt;
1951 u16 dlen = msg_data_sz(hdr);
1952 int mtyp = msg_type(hdr);
1953 bool reply = msg_probe(hdr);
1959 trace_tipc_proto_rcv(skb, false, l->name);
1960 if (tipc_link_is_blocked(l) || !xmitq)
1963 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1964 l->net_plane = msg_net_plane(hdr);
1968 data = msg_data(hdr);
1970 if (!tipc_link_validate_msg(l, hdr)) {
1971 trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
1972 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
1979 /* Complete own link name with peer's interface name */
1980 if_name = strrchr(l->name, ':') + 1;
1981 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1983 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1985 strncpy(if_name, data, TIPC_MAX_IF_NAME);
1987 /* Update own tolerance if peer indicates a non-zero value */
1988 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1989 l->tolerance = peers_tol;
1990 l->bc_rcvlink->tolerance = peers_tol;
1992 /* Update own priority if peer's priority is higher */
1993 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1994 l->priority = peers_prio;
1996 /* If peer is going down we want full re-establish cycle */
1997 if (msg_peer_stopping(hdr)) {
1998 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2002 /* If this endpoint was re-created while peer was ESTABLISHING
2003 * it doesn't know current session number. Force re-synch.
2005 if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2006 l->session != msg_dest_session(hdr)) {
2007 if (less(l->session, msg_dest_session(hdr)))
2008 l->session = msg_dest_session(hdr) + 1;
2012 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2013 if (mtyp == RESET_MSG || !link_is_up(l))
2014 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2016 /* ACTIVATE_MSG takes up link if it was already locally reset */
2017 if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2018 rc = TIPC_LINK_UP_EVT;
2020 l->peer_session = msg_session(hdr);
2021 l->in_session = true;
2022 l->peer_bearer_id = msg_bearer_id(hdr);
2023 if (l->mtu > msg_max_pkt(hdr))
2024 l->mtu = msg_max_pkt(hdr);
2028 l->rcv_nxt_state = msg_seqno(hdr) + 1;
2030 /* Update own tolerance if peer indicates a non-zero value */
2031 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2032 l->tolerance = peers_tol;
2033 l->bc_rcvlink->tolerance = peers_tol;
2035 /* Update own prio if peer indicates a different value */
2036 if ((peers_prio != l->priority) &&
2037 in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2038 l->priority = peers_prio;
2039 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2042 l->silent_intv_cnt = 0;
2043 l->stats.recv_states++;
2045 l->stats.recv_probes++;
2047 if (!link_is_up(l)) {
2048 if (l->state == LINK_ESTABLISHING)
2049 rc = TIPC_LINK_UP_EVT;
2053 /* Receive Gap ACK blocks from peer if any */
2054 if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
2055 ga = (struct tipc_gap_ack_blks *)data;
2056 glen = ntohs(ga->len);
2057 /* sanity check: if failed, ignore Gap ACK blocks */
2058 if (glen != tipc_gap_ack_blks_sz(ga->gack_cnt))
2062 tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2063 &l->mon_state, l->bearer_id);
2065 /* Send NACK if peer has sent pkts we haven't received yet */
2066 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
2067 rcvgap = peers_snd_nxt - l->rcv_nxt;
2068 if (rcvgap || reply)
2069 tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2070 rcvgap, 0, 0, xmitq);
2072 rc |= tipc_link_advance_transmq(l, ack, gap, ga, xmitq);
2074 /* If NACK, retransmit will now start at right position */
2076 l->stats.recv_nacks++;
2078 tipc_link_advance_backlog(l, xmitq);
2079 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2080 link_prepare_wakeup(l);
2087 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2089 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2091 struct sk_buff_head *xmitq)
2093 struct sk_buff *skb;
2094 struct tipc_msg *hdr;
2095 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2096 u16 ack = l->rcv_nxt - 1;
2097 u16 gap_to = peers_snd_nxt - 1;
2099 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2100 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2104 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2105 msg_set_bcast_ack(hdr, ack);
2106 msg_set_bcgap_after(hdr, ack);
2108 gap_to = buf_seqno(dfrd_skb) - 1;
2109 msg_set_bcgap_to(hdr, gap_to);
2110 msg_set_non_seq(hdr, bcast);
2111 __skb_queue_tail(xmitq, skb);
2115 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2117 * Give a newly added peer node the sequence number where it should
2118 * start receiving and acking broadcast packets.
2120 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2121 struct sk_buff_head *xmitq)
2123 struct sk_buff_head list;
2125 __skb_queue_head_init(&list);
2126 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2128 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2129 tipc_link_xmit(l, &list, xmitq);
2132 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2134 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2136 int mtyp = msg_type(hdr);
2137 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2142 if (msg_user(hdr) == BCAST_PROTOCOL) {
2143 l->rcv_nxt = peers_snd_nxt;
2144 l->state = LINK_ESTABLISHED;
2148 if (l->peer_caps & TIPC_BCAST_SYNCH)
2151 if (msg_peer_node_is_up(hdr))
2154 /* Compatibility: accept older, less safe initial synch data */
2155 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2156 l->rcv_nxt = peers_snd_nxt;
2159 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2161 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2162 struct sk_buff_head *xmitq)
2164 struct tipc_link *snd_l = l->bc_sndlink;
2165 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2166 u16 from = msg_bcast_ack(hdr) + 1;
2167 u16 to = from + msg_bc_gap(hdr) - 1;
2173 if (!msg_peer_node_is_up(hdr))
2176 /* Open when peer ackowledges our bcast init msg (pkt #1) */
2178 l->bc_peer_is_up = true;
2180 if (!l->bc_peer_is_up)
2183 l->stats.recv_nacks++;
2185 /* Ignore if peers_snd_nxt goes beyond receive window */
2186 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2189 rc = tipc_link_bc_retrans(snd_l, l, from, to, xmitq);
2191 l->snd_nxt = peers_snd_nxt;
2192 if (link_bc_rcv_gap(l))
2193 rc |= TIPC_LINK_SND_STATE;
2195 /* Return now if sender supports nack via STATE messages */
2196 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2199 /* Otherwise, be backwards compatible */
2201 if (!more(peers_snd_nxt, l->rcv_nxt)) {
2202 l->nack_state = BC_NACK_SND_CONDITIONAL;
2206 /* Don't NACK if one was recently sent or peeked */
2207 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2208 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2212 /* Conditionally delay NACK sending until next synch rcv */
2213 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2214 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2215 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2219 /* Send NACK now but suppress next one */
2220 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2221 l->nack_state = BC_NACK_SND_SUPPRESS;
2225 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
2226 struct sk_buff_head *xmitq)
2228 struct sk_buff *skb, *tmp;
2229 struct tipc_link *snd_l = l->bc_sndlink;
2231 if (!link_is_up(l) || !l->bc_peer_is_up)
2234 if (!more(acked, l->acked))
2237 trace_tipc_link_bc_ack(l, l->acked, acked, &snd_l->transmq);
2238 /* Skip over packets peer has already acked */
2239 skb_queue_walk(&snd_l->transmq, skb) {
2240 if (more(buf_seqno(skb), l->acked))
2244 /* Update/release the packets peer is acking now */
2245 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
2246 if (more(buf_seqno(skb), acked))
2248 if (!--TIPC_SKB_CB(skb)->ackers) {
2249 __skb_unlink(skb, &snd_l->transmq);
2254 tipc_link_advance_backlog(snd_l, xmitq);
2255 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
2256 link_prepare_wakeup(snd_l);
2259 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2260 * This function is here for backwards compatibility, since
2261 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2263 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2264 struct sk_buff_head *xmitq)
2266 struct tipc_msg *hdr = buf_msg(skb);
2267 u32 dnode = msg_destnode(hdr);
2268 int mtyp = msg_type(hdr);
2269 u16 acked = msg_bcast_ack(hdr);
2270 u16 from = acked + 1;
2271 u16 to = msg_bcgap_to(hdr);
2272 u16 peers_snd_nxt = to + 1;
2277 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2280 if (mtyp != STATE_MSG)
2283 if (dnode == tipc_own_addr(l->net)) {
2284 tipc_link_bc_ack_rcv(l, acked, xmitq);
2285 rc = tipc_link_bc_retrans(l->bc_sndlink, l, from, to, xmitq);
2286 l->stats.recv_nacks++;
2290 /* Msg for other node => suppress own NACK at next sync if applicable */
2291 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2292 l->nack_state = BC_NACK_SND_SUPPRESS;
2297 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
2299 int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2302 l->backlog[TIPC_LOW_IMPORTANCE].limit = max_t(u16, 50, win);
2303 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = max_t(u16, 100, win * 2);
2304 l->backlog[TIPC_HIGH_IMPORTANCE].limit = max_t(u16, 150, win * 3);
2305 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
2306 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
2310 * link_reset_stats - reset link statistics
2311 * @l: pointer to link
2313 void tipc_link_reset_stats(struct tipc_link *l)
2315 memset(&l->stats, 0, sizeof(l->stats));
2318 static void link_print(struct tipc_link *l, const char *str)
2320 struct sk_buff *hskb = skb_peek(&l->transmq);
2321 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2322 u16 tail = l->snd_nxt - 1;
2324 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2325 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2326 skb_queue_len(&l->transmq), head, tail,
2327 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2330 /* Parse and validate nested (link) properties valid for media, bearer and link
2332 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2336 err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2337 tipc_nl_prop_policy, NULL);
2341 if (props[TIPC_NLA_PROP_PRIO]) {
2344 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2345 if (prio > TIPC_MAX_LINK_PRI)
2349 if (props[TIPC_NLA_PROP_TOL]) {
2352 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2353 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2357 if (props[TIPC_NLA_PROP_WIN]) {
2360 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2361 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
2368 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2371 struct nlattr *stats;
2378 struct nla_map map[] = {
2379 {TIPC_NLA_STATS_RX_INFO, 0},
2380 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2381 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2382 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2383 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2384 {TIPC_NLA_STATS_TX_INFO, 0},
2385 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2386 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2387 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2388 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2389 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2390 s->msg_length_counts : 1},
2391 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2392 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2393 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2394 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2395 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2396 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2397 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2398 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2399 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2400 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
2401 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2402 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2403 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2404 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
2405 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2406 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2407 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2408 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2409 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2410 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2411 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2412 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2413 (s->accu_queue_sz / s->queue_sz_counts) : 0}
2416 stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2420 for (i = 0; i < ARRAY_SIZE(map); i++)
2421 if (nla_put_u32(skb, map[i].key, map[i].val))
2424 nla_nest_end(skb, stats);
2428 nla_nest_cancel(skb, stats);
2433 /* Caller should hold appropriate locks to protect the link */
2434 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2435 struct tipc_link *link, int nlflags)
2437 u32 self = tipc_own_addr(net);
2438 struct nlattr *attrs;
2439 struct nlattr *prop;
2443 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2444 nlflags, TIPC_NL_LINK_GET);
2448 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2452 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2454 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2456 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2458 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2460 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2463 if (tipc_link_is_up(link))
2464 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2467 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2470 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2473 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2475 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2477 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2480 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2482 nla_nest_end(msg->skb, prop);
2484 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2488 nla_nest_end(msg->skb, attrs);
2489 genlmsg_end(msg->skb, hdr);
2494 nla_nest_cancel(msg->skb, prop);
2496 nla_nest_cancel(msg->skb, attrs);
2498 genlmsg_cancel(msg->skb, hdr);
2503 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2504 struct tipc_stats *stats)
2507 struct nlattr *nest;
2514 struct nla_map map[] = {
2515 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2516 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2517 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2518 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2519 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2520 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2521 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2522 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2523 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2524 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2525 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2526 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2527 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2528 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2529 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2530 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2531 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2532 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2533 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2534 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2537 nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2541 for (i = 0; i < ARRAY_SIZE(map); i++)
2542 if (nla_put_u32(skb, map[i].key, map[i].val))
2545 nla_nest_end(skb, nest);
2549 nla_nest_cancel(skb, nest);
2554 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2558 struct nlattr *attrs;
2559 struct nlattr *prop;
2560 struct tipc_net *tn = net_generic(net, tipc_net_id);
2561 u32 bc_mode = tipc_bcast_get_broadcast_mode(net);
2562 u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2563 struct tipc_link *bcl = tn->bcl;
2568 tipc_bcast_lock(net);
2570 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2571 NLM_F_MULTI, TIPC_NL_LINK_GET);
2573 tipc_bcast_unlock(net);
2577 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2581 /* The broadcast link is always up */
2582 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2585 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2587 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2589 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2591 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2594 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2597 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2599 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2601 if (bc_mode & BCLINK_MODE_SEL)
2602 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2605 nla_nest_end(msg->skb, prop);
2607 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2611 tipc_bcast_unlock(net);
2612 nla_nest_end(msg->skb, attrs);
2613 genlmsg_end(msg->skb, hdr);
2618 nla_nest_cancel(msg->skb, prop);
2620 nla_nest_cancel(msg->skb, attrs);
2622 tipc_bcast_unlock(net);
2623 genlmsg_cancel(msg->skb, hdr);
2628 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2629 struct sk_buff_head *xmitq)
2633 l->bc_rcvlink->tolerance = tol;
2635 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2638 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2639 struct sk_buff_head *xmitq)
2642 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2645 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2647 l->abort_limit = limit;
2650 char *tipc_link_name_ext(struct tipc_link *l, char *buf)
2653 scnprintf(buf, TIPC_MAX_LINK_NAME, "null");
2654 else if (link_is_bc_sndlink(l))
2655 scnprintf(buf, TIPC_MAX_LINK_NAME, "broadcast-sender");
2656 else if (link_is_bc_rcvlink(l))
2657 scnprintf(buf, TIPC_MAX_LINK_NAME,
2658 "broadcast-receiver, peer %x", l->addr);
2660 memcpy(buf, l->name, TIPC_MAX_LINK_NAME);
2666 * tipc_link_dump - dump TIPC link data
2667 * @l: tipc link to be dumped
2668 * @dqueues: bitmask to decide if any link queue to be dumped?
2669 * - TIPC_DUMP_NONE: don't dump link queues
2670 * - TIPC_DUMP_TRANSMQ: dump link transmq queue
2671 * - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2672 * - TIPC_DUMP_DEFERDQ: dump link deferd queue
2673 * - TIPC_DUMP_INPUTQ: dump link input queue
2674 * - TIPC_DUMP_WAKEUP: dump link wakeup queue
2675 * - TIPC_DUMP_ALL: dump all the link queues above
2676 * @buf: returned buffer of dump data in format
2678 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2681 size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2682 struct sk_buff_head *list;
2683 struct sk_buff *hskb, *tskb;
2687 i += scnprintf(buf, sz, "link data: (null)\n");
2691 i += scnprintf(buf, sz, "link data: %x", l->addr);
2692 i += scnprintf(buf + i, sz - i, " %x", l->state);
2693 i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2694 i += scnprintf(buf + i, sz - i, " %u", l->session);
2695 i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2696 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2697 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2698 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2699 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2700 i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2701 i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2702 i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2703 i += scnprintf(buf + i, sz - i, " %u", 0);
2704 i += scnprintf(buf + i, sz - i, " %u", 0);
2705 i += scnprintf(buf + i, sz - i, " %u", l->acked);
2708 len = skb_queue_len(list);
2709 hskb = skb_peek(list);
2710 tskb = skb_peek_tail(list);
2711 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2712 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2713 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2716 len = skb_queue_len(list);
2717 hskb = skb_peek(list);
2718 tskb = skb_peek_tail(list);
2719 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2720 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2721 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2723 list = &l->backlogq;
2724 len = skb_queue_len(list);
2725 hskb = skb_peek(list);
2726 tskb = skb_peek_tail(list);
2727 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2728 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2729 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2732 len = skb_queue_len(list);
2733 hskb = skb_peek(list);
2734 tskb = skb_peek_tail(list);
2735 i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2736 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2737 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2739 if (dqueues & TIPC_DUMP_TRANSMQ) {
2740 i += scnprintf(buf + i, sz - i, "transmq: ");
2741 i += tipc_list_dump(&l->transmq, false, buf + i);
2743 if (dqueues & TIPC_DUMP_BACKLOGQ) {
2744 i += scnprintf(buf + i, sz - i,
2745 "backlogq: <%u %u %u %u %u>, ",
2746 l->backlog[TIPC_LOW_IMPORTANCE].len,
2747 l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2748 l->backlog[TIPC_HIGH_IMPORTANCE].len,
2749 l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2750 l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2751 i += tipc_list_dump(&l->backlogq, false, buf + i);
2753 if (dqueues & TIPC_DUMP_DEFERDQ) {
2754 i += scnprintf(buf + i, sz - i, "deferdq: ");
2755 i += tipc_list_dump(&l->deferdq, false, buf + i);
2757 if (dqueues & TIPC_DUMP_INPUTQ) {
2758 i += scnprintf(buf + i, sz - i, "inputq: ");
2759 i += tipc_list_dump(l->inputq, false, buf + i);
2761 if (dqueues & TIPC_DUMP_WAKEUP) {
2762 i += scnprintf(buf + i, sz - i, "wakeup: ");
2763 i += tipc_list_dump(&l->wakeupq, false, buf + i);