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 unsigned int maxwin = l->window;
943 unsigned int mtu = l->mtu;
944 u16 ack = l->rcv_nxt - 1;
945 u16 seqno = l->snd_nxt;
946 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
947 struct sk_buff_head *transmq = &l->transmq;
948 struct sk_buff_head *backlogq = &l->backlogq;
949 struct sk_buff *skb, *_skb, **tskb;
950 int pkt_cnt = skb_queue_len(list);
951 struct tipc_msg *hdr;
958 hdr = buf_msg(skb_peek(list));
959 if (unlikely(msg_size(hdr) > mtu)) {
960 pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
961 skb_queue_len(list), msg_user(hdr),
962 msg_type(hdr), msg_size(hdr), mtu);
963 __skb_queue_purge(list);
967 imp = msg_importance(hdr);
968 /* Allow oversubscription of one data msg per source at congestion */
969 if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
970 if (imp == TIPC_SYSTEM_IMPORTANCE) {
971 pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
974 rc = link_schedule_user(l, hdr);
978 l->stats.sent_fragmented++;
979 l->stats.sent_fragments += pkt_cnt;
982 /* Prepare each packet for sending, and add to relevant queue: */
983 while (skb_queue_len(list)) {
984 skb = skb_peek(list);
986 msg_set_seqno(hdr, seqno);
987 msg_set_ack(hdr, ack);
988 msg_set_bcast_ack(hdr, bc_ack);
990 if (likely(skb_queue_len(transmq) < maxwin)) {
991 _skb = skb_clone(skb, GFP_ATOMIC);
993 __skb_queue_purge(list);
997 __skb_queue_tail(transmq, skb);
998 /* next retransmit attempt */
999 if (link_is_bc_sndlink(l))
1000 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1001 __skb_queue_tail(xmitq, _skb);
1002 TIPC_SKB_CB(skb)->ackers = l->ackers;
1004 l->stats.sent_pkts++;
1008 tskb = &l->backlog[imp].target_bskb;
1009 if (tipc_msg_bundle(*tskb, hdr, mtu)) {
1010 kfree_skb(__skb_dequeue(list));
1011 l->stats.sent_bundled++;
1014 if (tipc_msg_make_bundle(tskb, hdr, mtu, l->addr)) {
1015 kfree_skb(__skb_dequeue(list));
1016 __skb_queue_tail(backlogq, *tskb);
1017 l->backlog[imp].len++;
1018 l->stats.sent_bundled++;
1019 l->stats.sent_bundles++;
1022 l->backlog[imp].target_bskb = NULL;
1023 l->backlog[imp].len += skb_queue_len(list);
1024 skb_queue_splice_tail_init(list, backlogq);
1030 static void tipc_link_advance_backlog(struct tipc_link *l,
1031 struct sk_buff_head *xmitq)
1033 struct sk_buff *skb, *_skb;
1034 struct tipc_msg *hdr;
1035 u16 seqno = l->snd_nxt;
1036 u16 ack = l->rcv_nxt - 1;
1037 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1040 while (skb_queue_len(&l->transmq) < l->window) {
1041 skb = skb_peek(&l->backlogq);
1044 _skb = skb_clone(skb, GFP_ATOMIC);
1047 __skb_dequeue(&l->backlogq);
1049 imp = msg_importance(hdr);
1050 l->backlog[imp].len--;
1051 if (unlikely(skb == l->backlog[imp].target_bskb))
1052 l->backlog[imp].target_bskb = NULL;
1053 __skb_queue_tail(&l->transmq, skb);
1054 /* next retransmit attempt */
1055 if (link_is_bc_sndlink(l))
1056 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1058 __skb_queue_tail(xmitq, _skb);
1059 TIPC_SKB_CB(skb)->ackers = l->ackers;
1060 msg_set_seqno(hdr, seqno);
1061 msg_set_ack(hdr, ack);
1062 msg_set_bcast_ack(hdr, bc_ack);
1064 l->stats.sent_pkts++;
1071 * link_retransmit_failure() - Detect repeated retransmit failures
1072 * @l: tipc link sender
1073 * @r: tipc link receiver (= l in case of unicast)
1074 * @rc: returned code
1076 * Return: true if the repeated retransmit failures happens, otherwise
1079 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1082 struct sk_buff *skb = skb_peek(&l->transmq);
1083 struct tipc_msg *hdr;
1088 if (!TIPC_SKB_CB(skb)->retr_cnt)
1091 if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1092 msecs_to_jiffies(r->tolerance * 10)))
1096 if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1099 pr_warn("Retransmission failure on link <%s>\n", l->name);
1100 link_print(l, "State of link ");
1101 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1102 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1103 pr_info("sqno %u, prev: %x, dest: %x\n",
1104 msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1105 pr_info("retr_stamp %d, retr_cnt %d\n",
1106 jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1107 TIPC_SKB_CB(skb)->retr_cnt);
1109 trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1110 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1111 trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1113 if (link_is_bc_sndlink(l)) {
1114 r->state = LINK_RESET;
1115 *rc = TIPC_LINK_DOWN_EVT;
1117 *rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1123 /* tipc_link_bc_retrans() - retransmit zero or more packets
1124 * @l: the link to transmit on
1125 * @r: the receiving link ordering the retransmit. Same as l if unicast
1126 * @from: retransmit from (inclusive) this sequence number
1127 * @to: retransmit to (inclusive) this sequence number
1128 * xmitq: queue for accumulating the retransmitted packets
1130 static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r,
1131 u16 from, u16 to, struct sk_buff_head *xmitq)
1133 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1134 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1135 u16 ack = l->rcv_nxt - 1;
1136 struct tipc_msg *hdr;
1144 trace_tipc_link_retrans(r, from, to, &l->transmq);
1146 if (link_retransmit_failure(l, r, &rc))
1149 skb_queue_walk(&l->transmq, skb) {
1151 if (less(msg_seqno(hdr), from))
1153 if (more(msg_seqno(hdr), to))
1156 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1158 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1159 _skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE, GFP_ATOMIC);
1162 hdr = buf_msg(_skb);
1163 msg_set_ack(hdr, ack);
1164 msg_set_bcast_ack(hdr, bc_ack);
1165 _skb->priority = TC_PRIO_CONTROL;
1166 __skb_queue_tail(xmitq, _skb);
1167 l->stats.retransmitted++;
1169 /* Increase actual retrans counter & mark first time */
1170 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1171 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1176 /* tipc_data_input - deliver data and name distr msgs to upper layer
1178 * Consumes buffer if message is of right type
1179 * Node lock must be held
1181 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1182 struct sk_buff_head *inputq)
1184 struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1185 struct tipc_msg *hdr = buf_msg(skb);
1187 switch (msg_user(hdr)) {
1188 case TIPC_LOW_IMPORTANCE:
1189 case TIPC_MEDIUM_IMPORTANCE:
1190 case TIPC_HIGH_IMPORTANCE:
1191 case TIPC_CRITICAL_IMPORTANCE:
1192 if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1193 skb_queue_tail(mc_inputq, skb);
1198 skb_queue_tail(inputq, skb);
1200 case GROUP_PROTOCOL:
1201 skb_queue_tail(mc_inputq, skb);
1203 case NAME_DISTRIBUTOR:
1204 l->bc_rcvlink->state = LINK_ESTABLISHED;
1205 skb_queue_tail(l->namedq, skb);
1208 case TUNNEL_PROTOCOL:
1209 case MSG_FRAGMENTER:
1210 case BCAST_PROTOCOL:
1213 pr_warn("Dropping received illegal msg type\n");
1219 /* tipc_link_input - process packet that has passed link protocol check
1223 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1224 struct sk_buff_head *inputq,
1225 struct sk_buff **reasm_skb)
1227 struct tipc_msg *hdr = buf_msg(skb);
1228 struct sk_buff *iskb;
1229 struct sk_buff_head tmpq;
1230 int usr = msg_user(hdr);
1233 if (usr == MSG_BUNDLER) {
1234 skb_queue_head_init(&tmpq);
1235 l->stats.recv_bundles++;
1236 l->stats.recv_bundled += msg_msgcnt(hdr);
1237 while (tipc_msg_extract(skb, &iskb, &pos))
1238 tipc_data_input(l, iskb, &tmpq);
1239 tipc_skb_queue_splice_tail(&tmpq, inputq);
1241 } else if (usr == MSG_FRAGMENTER) {
1242 l->stats.recv_fragments++;
1243 if (tipc_buf_append(reasm_skb, &skb)) {
1244 l->stats.recv_fragmented++;
1245 tipc_data_input(l, skb, inputq);
1246 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1247 pr_warn_ratelimited("Unable to build fragment list\n");
1248 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1251 } else if (usr == BCAST_PROTOCOL) {
1252 tipc_bcast_lock(l->net);
1253 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1254 tipc_bcast_unlock(l->net);
1261 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1262 * inner message along with the ones in the old link's
1265 * @skb: TUNNEL_PROTOCOL message
1266 * @inputq: queue to put messages ready for delivery
1268 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1269 struct sk_buff_head *inputq)
1271 struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1272 struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1273 struct sk_buff_head *fdefq = &l->failover_deferdq;
1274 struct tipc_msg *hdr = buf_msg(skb);
1275 struct sk_buff *iskb;
1280 if (msg_type(hdr) == SYNCH_MSG) {
1285 /* Not a fragment? */
1286 if (likely(!msg_nof_fragms(hdr))) {
1287 if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1288 pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1289 skb_queue_len(fdefq));
1294 /* Set fragment type for buf_append */
1295 if (msg_fragm_no(hdr) == 1)
1296 msg_set_type(hdr, FIRST_FRAGMENT);
1297 else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1298 msg_set_type(hdr, FRAGMENT);
1300 msg_set_type(hdr, LAST_FRAGMENT);
1302 if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1303 /* Successful but non-complete reassembly? */
1304 if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1306 pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1307 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1313 seqno = buf_seqno(iskb);
1314 if (unlikely(less(seqno, l->drop_point))) {
1318 if (unlikely(seqno != l->drop_point)) {
1319 __tipc_skb_queue_sorted(fdefq, seqno, iskb);
1324 if (!tipc_data_input(l, iskb, inputq))
1325 rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1328 } while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1333 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1335 bool released = false;
1336 struct sk_buff *skb, *tmp;
1338 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1339 if (more(buf_seqno(skb), acked))
1341 __skb_unlink(skb, &l->transmq);
1348 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1349 * @l: tipc link that data have come with gaps in sequence if any
1350 * @data: data buffer to store the Gap ACK blocks after built
1352 * returns the actual allocated memory size
1354 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data)
1356 struct sk_buff *skb = skb_peek(&l->deferdq);
1357 struct tipc_gap_ack_blks *ga = data;
1358 u16 len, expect, seqno = 0;
1364 expect = buf_seqno(skb);
1365 skb_queue_walk(&l->deferdq, skb) {
1366 seqno = buf_seqno(skb);
1367 if (unlikely(more(seqno, expect))) {
1368 ga->gacks[n].ack = htons(expect - 1);
1369 ga->gacks[n].gap = htons(seqno - expect);
1370 if (++n >= MAX_GAP_ACK_BLKS) {
1371 pr_info_ratelimited("Too few Gap ACK blocks!\n");
1374 } else if (unlikely(less(seqno, expect))) {
1375 pr_warn("Unexpected skb in deferdq!\n");
1382 ga->gacks[n].ack = htons(seqno);
1383 ga->gacks[n].gap = 0;
1387 len = tipc_gap_ack_blks_sz(n);
1388 ga->len = htons(len);
1393 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1394 * acked packets, also doing retransmissions if
1396 * @l: tipc link with transmq queue to be advanced
1397 * @acked: seqno of last packet acked by peer without any gaps before
1398 * @gap: # of gap packets
1399 * @ga: buffer pointer to Gap ACK blocks from peer
1400 * @xmitq: queue for accumulating the retransmitted packets if any
1402 * In case of a repeated retransmit failures, the call will return shortly
1403 * with a returned code (e.g. TIPC_LINK_DOWN_EVT)
1405 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
1406 struct tipc_gap_ack_blks *ga,
1407 struct sk_buff_head *xmitq)
1409 struct sk_buff *skb, *_skb, *tmp;
1410 struct tipc_msg *hdr;
1411 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1412 u16 ack = l->rcv_nxt - 1;
1413 bool passed = false;
1417 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1418 seqno = buf_seqno(skb);
1421 if (less_eq(seqno, acked)) {
1423 __skb_unlink(skb, &l->transmq);
1425 } else if (less_eq(seqno, acked + gap)) {
1426 /* First, check if repeated retrans failures occurs? */
1427 if (!passed && link_retransmit_failure(l, l, &rc))
1431 /* retransmit skb if unrestricted*/
1432 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1434 TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
1435 _skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE,
1439 hdr = buf_msg(_skb);
1440 msg_set_ack(hdr, ack);
1441 msg_set_bcast_ack(hdr, bc_ack);
1442 _skb->priority = TC_PRIO_CONTROL;
1443 __skb_queue_tail(xmitq, _skb);
1444 l->stats.retransmitted++;
1446 /* Increase actual retrans counter & mark first time */
1447 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1448 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1450 /* retry with Gap ACK blocks if any */
1451 if (!ga || n >= ga->gack_cnt)
1453 acked = ntohs(ga->gacks[n].ack);
1454 gap = ntohs(ga->gacks[n].gap);
1463 /* tipc_link_build_state_msg: prepare link state message for transmission
1465 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1466 * risk of ack storms towards the sender
1468 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1473 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1474 if (link_is_bc_rcvlink(l)) {
1475 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1479 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1480 l->snd_nxt = l->rcv_nxt;
1481 return TIPC_LINK_SND_STATE;
1486 l->stats.sent_acks++;
1487 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1491 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1493 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1495 int mtyp = RESET_MSG;
1496 struct sk_buff *skb;
1498 if (l->state == LINK_ESTABLISHING)
1499 mtyp = ACTIVATE_MSG;
1501 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1503 /* Inform peer that this endpoint is going down if applicable */
1504 skb = skb_peek_tail(xmitq);
1505 if (skb && (l->state == LINK_RESET))
1506 msg_set_peer_stopping(buf_msg(skb), 1);
1509 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1510 * Note that sending of broadcast NACK is coordinated among nodes, to
1511 * reduce the risk of NACK storms towards the sender
1513 static int tipc_link_build_nack_msg(struct tipc_link *l,
1514 struct sk_buff_head *xmitq)
1516 u32 def_cnt = ++l->stats.deferred_recv;
1517 u32 defq_len = skb_queue_len(&l->deferdq);
1520 if (link_is_bc_rcvlink(l)) {
1521 match1 = def_cnt & 0xf;
1522 match2 = tipc_own_addr(l->net) & 0xf;
1523 if (match1 == match2)
1524 return TIPC_LINK_SND_STATE;
1528 if (defq_len >= 3 && !((defq_len - 3) % 16))
1529 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1533 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1534 * @l: the link that should handle the message
1536 * @xmitq: queue to place packets to be sent after this call
1538 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1539 struct sk_buff_head *xmitq)
1541 struct sk_buff_head *defq = &l->deferdq;
1542 struct tipc_msg *hdr = buf_msg(skb);
1543 u16 seqno, rcv_nxt, win_lim;
1546 /* Verify and update link state */
1547 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1548 return tipc_link_proto_rcv(l, skb, xmitq);
1550 /* Don't send probe at next timeout expiration */
1551 l->silent_intv_cnt = 0;
1555 seqno = msg_seqno(hdr);
1556 rcv_nxt = l->rcv_nxt;
1557 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1559 if (unlikely(!link_is_up(l))) {
1560 if (l->state == LINK_ESTABLISHING)
1561 rc = TIPC_LINK_UP_EVT;
1565 /* Drop if outside receive window */
1566 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1567 l->stats.duplicates++;
1571 /* Forward queues and wake up waiting users */
1572 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1573 tipc_link_advance_backlog(l, xmitq);
1574 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1575 link_prepare_wakeup(l);
1578 /* Defer delivery if sequence gap */
1579 if (unlikely(seqno != rcv_nxt)) {
1580 __tipc_skb_queue_sorted(defq, seqno, skb);
1581 rc |= tipc_link_build_nack_msg(l, xmitq);
1585 /* Deliver packet */
1587 l->stats.recv_pkts++;
1589 if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1590 rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1591 else if (!tipc_data_input(l, skb, l->inputq))
1592 rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1593 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1594 rc |= tipc_link_build_state_msg(l, xmitq);
1595 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1597 } while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1605 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1606 bool probe_reply, u16 rcvgap,
1607 int tolerance, int priority,
1608 struct sk_buff_head *xmitq)
1610 struct tipc_link *bcl = l->bc_rcvlink;
1611 struct sk_buff *skb;
1612 struct tipc_msg *hdr;
1613 struct sk_buff_head *dfq = &l->deferdq;
1614 bool node_up = link_is_up(bcl);
1615 struct tipc_mon_state *mstate = &l->mon_state;
1620 /* Don't send protocol message during reset or link failover */
1621 if (tipc_link_is_blocked(l))
1624 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1627 if (!skb_queue_empty(dfq))
1628 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1630 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1631 tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1632 l->addr, tipc_own_addr(l->net), 0, 0, 0);
1637 data = msg_data(hdr);
1638 msg_set_session(hdr, l->session);
1639 msg_set_bearer_id(hdr, l->bearer_id);
1640 msg_set_net_plane(hdr, l->net_plane);
1641 msg_set_next_sent(hdr, l->snd_nxt);
1642 msg_set_ack(hdr, l->rcv_nxt - 1);
1643 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1644 msg_set_bc_ack_invalid(hdr, !node_up);
1645 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1646 msg_set_link_tolerance(hdr, tolerance);
1647 msg_set_linkprio(hdr, priority);
1648 msg_set_redundant_link(hdr, node_up);
1649 msg_set_seq_gap(hdr, 0);
1650 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1652 if (mtyp == STATE_MSG) {
1653 if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1654 msg_set_seqno(hdr, l->snd_nxt_state++);
1655 msg_set_seq_gap(hdr, rcvgap);
1656 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1657 msg_set_probe(hdr, probe);
1658 msg_set_is_keepalive(hdr, probe || probe_reply);
1659 if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1660 glen = tipc_build_gap_ack_blks(l, data);
1661 tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1662 msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1663 skb_trim(skb, INT_H_SIZE + glen + dlen);
1664 l->stats.sent_states++;
1667 /* RESET_MSG or ACTIVATE_MSG */
1668 if (mtyp == ACTIVATE_MSG) {
1669 msg_set_dest_session_valid(hdr, 1);
1670 msg_set_dest_session(hdr, l->peer_session);
1672 msg_set_max_pkt(hdr, l->advertised_mtu);
1673 strcpy(data, l->if_name);
1674 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1675 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1678 l->stats.sent_probes++;
1680 l->stats.sent_nacks++;
1681 skb->priority = TC_PRIO_CONTROL;
1682 __skb_queue_tail(xmitq, skb);
1683 trace_tipc_proto_build(skb, false, l->name);
1686 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1687 struct sk_buff_head *xmitq)
1689 u32 onode = tipc_own_addr(l->net);
1690 struct tipc_msg *hdr, *ihdr;
1691 struct sk_buff_head tnlq;
1692 struct sk_buff *skb;
1693 u32 dnode = l->addr;
1695 __skb_queue_head_init(&tnlq);
1696 skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1697 INT_H_SIZE, BASIC_H_SIZE,
1698 dnode, onode, 0, 0, 0);
1700 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1705 msg_set_msgcnt(hdr, 1);
1706 msg_set_bearer_id(hdr, l->peer_bearer_id);
1708 ihdr = (struct tipc_msg *)msg_data(hdr);
1709 tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1710 BASIC_H_SIZE, dnode);
1711 msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1712 __skb_queue_tail(&tnlq, skb);
1713 tipc_link_xmit(l, &tnlq, xmitq);
1716 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1717 * with contents of the link's transmit and backlog queues.
1719 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1720 int mtyp, struct sk_buff_head *xmitq)
1722 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1723 struct sk_buff *skb, *tnlskb;
1724 struct tipc_msg *hdr, tnlhdr;
1725 struct sk_buff_head *queue = &l->transmq;
1726 struct sk_buff_head tmpxq, tnlq, frags;
1727 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1728 bool pktcnt_need_update = false;
1735 __skb_queue_head_init(&tnlq);
1736 __skb_queue_head_init(&tmpxq);
1737 __skb_queue_head_init(&frags);
1739 /* At least one packet required for safe algorithm => add dummy */
1740 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1741 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1742 0, 0, TIPC_ERR_NO_PORT);
1744 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1747 __skb_queue_tail(&tnlq, skb);
1748 tipc_link_xmit(l, &tnlq, &tmpxq);
1749 __skb_queue_purge(&tmpxq);
1752 * From now on, send only one single ("dummy") SYNCH message
1753 * to peer. The SYNCH message does not contain any data, just
1754 * a header conveying the synch point to the peer.
1756 if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1757 tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1758 INT_H_SIZE, 0, l->addr,
1759 tipc_own_addr(l->net),
1762 pr_warn("%sunable to create dummy SYNCH_MSG\n",
1767 hdr = buf_msg(tnlskb);
1768 syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
1769 msg_set_syncpt(hdr, syncpt);
1770 msg_set_bearer_id(hdr, l->peer_bearer_id);
1771 __skb_queue_tail(&tnlq, tnlskb);
1772 tipc_link_xmit(tnl, &tnlq, xmitq);
1776 /* Initialize reusable tunnel packet header */
1777 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1778 mtyp, INT_H_SIZE, l->addr);
1779 if (mtyp == SYNCH_MSG)
1780 pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
1782 pktcnt = skb_queue_len(&l->transmq);
1783 pktcnt += skb_queue_len(&l->backlogq);
1784 msg_set_msgcnt(&tnlhdr, pktcnt);
1785 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1787 /* Wrap each packet into a tunnel packet */
1788 skb_queue_walk(queue, skb) {
1790 if (queue == &l->backlogq)
1791 msg_set_seqno(hdr, seqno++);
1792 pktlen = msg_size(hdr);
1794 /* Tunnel link MTU is not large enough? This could be
1796 * 1) Link MTU has just changed or set differently;
1797 * 2) Or FAILOVER on the top of a SYNCH message
1799 * The 2nd case should not happen if peer supports
1800 * TIPC_TUNNEL_ENHANCED
1802 if (pktlen > tnl->mtu - INT_H_SIZE) {
1803 if (mtyp == FAILOVER_MSG &&
1804 (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1805 rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
1808 pr_warn("%sunable to frag msg: rc %d\n",
1812 pktcnt += skb_queue_len(&frags) - 1;
1813 pktcnt_need_update = true;
1814 skb_queue_splice_tail_init(&frags, &tnlq);
1817 /* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
1818 * => Just warn it and return!
1820 pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
1821 link_co_err, msg_user(hdr),
1822 msg_type(hdr), msg_size(hdr));
1826 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1827 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1829 pr_warn("%sunable to send packet\n", link_co_err);
1832 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1833 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1834 __skb_queue_tail(&tnlq, tnlskb);
1836 if (queue != &l->backlogq) {
1837 queue = &l->backlogq;
1841 if (pktcnt_need_update)
1842 skb_queue_walk(&tnlq, skb) {
1844 msg_set_msgcnt(hdr, pktcnt);
1847 tipc_link_xmit(tnl, &tnlq, xmitq);
1849 if (mtyp == FAILOVER_MSG) {
1850 tnl->drop_point = l->rcv_nxt;
1851 tnl->failover_reasm_skb = l->reasm_buf;
1852 l->reasm_buf = NULL;
1854 /* Failover the link's deferdq */
1855 if (unlikely(!skb_queue_empty(fdefq))) {
1856 pr_warn("Link failover deferdq not empty: %d!\n",
1857 skb_queue_len(fdefq));
1858 __skb_queue_purge(fdefq);
1860 skb_queue_splice_init(&l->deferdq, fdefq);
1865 * tipc_link_failover_prepare() - prepare tnl for link failover
1867 * This is a special version of the precursor - tipc_link_tnl_prepare(),
1868 * see the tipc_node_link_failover() for details
1872 * @xmitq: queue for messages to be xmited
1874 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
1875 struct sk_buff_head *xmitq)
1877 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1879 tipc_link_create_dummy_tnl_msg(tnl, xmitq);
1881 /* This failover link enpoint was never established before,
1882 * so it has not received anything from peer.
1883 * Otherwise, it must be a normal failover situation or the
1884 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
1885 * would have to start over from scratch instead.
1887 tnl->drop_point = 1;
1888 tnl->failover_reasm_skb = NULL;
1890 /* Initiate the link's failover deferdq */
1891 if (unlikely(!skb_queue_empty(fdefq))) {
1892 pr_warn("Link failover deferdq not empty: %d!\n",
1893 skb_queue_len(fdefq));
1894 __skb_queue_purge(fdefq);
1898 /* tipc_link_validate_msg(): validate message against current link state
1899 * Returns true if message should be accepted, otherwise false
1901 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
1903 u16 curr_session = l->peer_session;
1904 u16 session = msg_session(hdr);
1905 int mtyp = msg_type(hdr);
1907 if (msg_user(hdr) != LINK_PROTOCOL)
1914 /* Accept only RESET with new session number */
1915 return more(session, curr_session);
1919 /* Accept only ACTIVATE with new or current session number */
1920 return !less(session, curr_session);
1922 /* Accept only STATE with current session number */
1925 if (session != curr_session)
1927 /* Extra sanity check */
1928 if (!link_is_up(l) && msg_ack(hdr))
1930 if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
1932 /* Accept only STATE with new sequence number */
1933 return !less(msg_seqno(hdr), l->rcv_nxt_state);
1939 /* tipc_link_proto_rcv(): receive link level protocol message :
1940 * Note that network plane id propagates through the network, and may
1941 * change at any time. The node with lowest numerical id determines
1944 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1945 struct sk_buff_head *xmitq)
1947 struct tipc_msg *hdr = buf_msg(skb);
1948 struct tipc_gap_ack_blks *ga = NULL;
1950 u16 ack = msg_ack(hdr);
1951 u16 gap = msg_seq_gap(hdr);
1952 u16 peers_snd_nxt = msg_next_sent(hdr);
1953 u16 peers_tol = msg_link_tolerance(hdr);
1954 u16 peers_prio = msg_linkprio(hdr);
1955 u16 rcv_nxt = l->rcv_nxt;
1956 u32 dlen = msg_data_sz(hdr), glen = 0;
1957 int mtyp = msg_type(hdr);
1958 bool reply = msg_probe(hdr);
1963 trace_tipc_proto_rcv(skb, false, l->name);
1968 if (tipc_link_is_blocked(l) || !xmitq)
1971 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1972 l->net_plane = msg_net_plane(hdr);
1974 if (skb_linearize(skb))
1978 data = msg_data(hdr);
1980 if (!tipc_link_validate_msg(l, hdr)) {
1981 trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
1982 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
1989 /* Complete own link name with peer's interface name */
1990 if_name = strrchr(l->name, ':') + 1;
1991 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1993 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1995 strncpy(if_name, data, TIPC_MAX_IF_NAME);
1997 /* Update own tolerance if peer indicates a non-zero value */
1998 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1999 l->tolerance = peers_tol;
2000 l->bc_rcvlink->tolerance = peers_tol;
2002 /* Update own priority if peer's priority is higher */
2003 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
2004 l->priority = peers_prio;
2006 /* If peer is going down we want full re-establish cycle */
2007 if (msg_peer_stopping(hdr)) {
2008 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2012 /* If this endpoint was re-created while peer was ESTABLISHING
2013 * it doesn't know current session number. Force re-synch.
2015 if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2016 l->session != msg_dest_session(hdr)) {
2017 if (less(l->session, msg_dest_session(hdr)))
2018 l->session = msg_dest_session(hdr) + 1;
2022 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2023 if (mtyp == RESET_MSG || !link_is_up(l))
2024 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2026 /* ACTIVATE_MSG takes up link if it was already locally reset */
2027 if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2028 rc = TIPC_LINK_UP_EVT;
2030 l->peer_session = msg_session(hdr);
2031 l->in_session = true;
2032 l->peer_bearer_id = msg_bearer_id(hdr);
2033 if (l->mtu > msg_max_pkt(hdr))
2034 l->mtu = msg_max_pkt(hdr);
2038 l->rcv_nxt_state = msg_seqno(hdr) + 1;
2040 /* Update own tolerance if peer indicates a non-zero value */
2041 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2042 l->tolerance = peers_tol;
2043 l->bc_rcvlink->tolerance = peers_tol;
2045 /* Update own prio if peer indicates a different value */
2046 if ((peers_prio != l->priority) &&
2047 in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2048 l->priority = peers_prio;
2049 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2052 l->silent_intv_cnt = 0;
2053 l->stats.recv_states++;
2055 l->stats.recv_probes++;
2057 if (!link_is_up(l)) {
2058 if (l->state == LINK_ESTABLISHING)
2059 rc = TIPC_LINK_UP_EVT;
2063 /* Receive Gap ACK blocks from peer if any */
2064 if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
2065 ga = (struct tipc_gap_ack_blks *)data;
2066 glen = ntohs(ga->len);
2067 /* sanity check: if failed, ignore Gap ACK blocks */
2068 if (glen != tipc_gap_ack_blks_sz(ga->gack_cnt))
2073 tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2074 &l->mon_state, l->bearer_id);
2076 /* Send NACK if peer has sent pkts we haven't received yet */
2077 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
2078 rcvgap = peers_snd_nxt - l->rcv_nxt;
2079 if (rcvgap || reply)
2080 tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2081 rcvgap, 0, 0, xmitq);
2083 rc |= tipc_link_advance_transmq(l, ack, gap, ga, xmitq);
2085 /* If NACK, retransmit will now start at right position */
2087 l->stats.recv_nacks++;
2089 tipc_link_advance_backlog(l, xmitq);
2090 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2091 link_prepare_wakeup(l);
2098 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2100 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2102 struct sk_buff_head *xmitq)
2104 struct sk_buff *skb;
2105 struct tipc_msg *hdr;
2106 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2107 u16 ack = l->rcv_nxt - 1;
2108 u16 gap_to = peers_snd_nxt - 1;
2110 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2111 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2115 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2116 msg_set_bcast_ack(hdr, ack);
2117 msg_set_bcgap_after(hdr, ack);
2119 gap_to = buf_seqno(dfrd_skb) - 1;
2120 msg_set_bcgap_to(hdr, gap_to);
2121 msg_set_non_seq(hdr, bcast);
2122 __skb_queue_tail(xmitq, skb);
2126 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2128 * Give a newly added peer node the sequence number where it should
2129 * start receiving and acking broadcast packets.
2131 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2132 struct sk_buff_head *xmitq)
2134 struct sk_buff_head list;
2136 __skb_queue_head_init(&list);
2137 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2139 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2140 tipc_link_xmit(l, &list, xmitq);
2143 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2145 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2147 int mtyp = msg_type(hdr);
2148 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2153 if (msg_user(hdr) == BCAST_PROTOCOL) {
2154 l->rcv_nxt = peers_snd_nxt;
2155 l->state = LINK_ESTABLISHED;
2159 if (l->peer_caps & TIPC_BCAST_SYNCH)
2162 if (msg_peer_node_is_up(hdr))
2165 /* Compatibility: accept older, less safe initial synch data */
2166 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2167 l->rcv_nxt = peers_snd_nxt;
2170 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2172 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2173 struct sk_buff_head *xmitq)
2175 struct tipc_link *snd_l = l->bc_sndlink;
2176 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2177 u16 from = msg_bcast_ack(hdr) + 1;
2178 u16 to = from + msg_bc_gap(hdr) - 1;
2184 if (!msg_peer_node_is_up(hdr))
2187 /* Open when peer ackowledges our bcast init msg (pkt #1) */
2189 l->bc_peer_is_up = true;
2191 if (!l->bc_peer_is_up)
2194 l->stats.recv_nacks++;
2196 /* Ignore if peers_snd_nxt goes beyond receive window */
2197 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2200 rc = tipc_link_bc_retrans(snd_l, l, from, to, xmitq);
2202 l->snd_nxt = peers_snd_nxt;
2203 if (link_bc_rcv_gap(l))
2204 rc |= TIPC_LINK_SND_STATE;
2206 /* Return now if sender supports nack via STATE messages */
2207 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2210 /* Otherwise, be backwards compatible */
2212 if (!more(peers_snd_nxt, l->rcv_nxt)) {
2213 l->nack_state = BC_NACK_SND_CONDITIONAL;
2217 /* Don't NACK if one was recently sent or peeked */
2218 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2219 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2223 /* Conditionally delay NACK sending until next synch rcv */
2224 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2225 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2226 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2230 /* Send NACK now but suppress next one */
2231 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2232 l->nack_state = BC_NACK_SND_SUPPRESS;
2236 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
2237 struct sk_buff_head *xmitq)
2239 struct sk_buff *skb, *tmp;
2240 struct tipc_link *snd_l = l->bc_sndlink;
2242 if (!link_is_up(l) || !l->bc_peer_is_up)
2245 if (!more(acked, l->acked))
2248 trace_tipc_link_bc_ack(l, l->acked, acked, &snd_l->transmq);
2249 /* Skip over packets peer has already acked */
2250 skb_queue_walk(&snd_l->transmq, skb) {
2251 if (more(buf_seqno(skb), l->acked))
2255 /* Update/release the packets peer is acking now */
2256 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
2257 if (more(buf_seqno(skb), acked))
2259 if (!--TIPC_SKB_CB(skb)->ackers) {
2260 __skb_unlink(skb, &snd_l->transmq);
2265 tipc_link_advance_backlog(snd_l, xmitq);
2266 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
2267 link_prepare_wakeup(snd_l);
2270 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2271 * This function is here for backwards compatibility, since
2272 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2274 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2275 struct sk_buff_head *xmitq)
2277 struct tipc_msg *hdr = buf_msg(skb);
2278 u32 dnode = msg_destnode(hdr);
2279 int mtyp = msg_type(hdr);
2280 u16 acked = msg_bcast_ack(hdr);
2281 u16 from = acked + 1;
2282 u16 to = msg_bcgap_to(hdr);
2283 u16 peers_snd_nxt = to + 1;
2288 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2291 if (mtyp != STATE_MSG)
2294 if (dnode == tipc_own_addr(l->net)) {
2295 tipc_link_bc_ack_rcv(l, acked, xmitq);
2296 rc = tipc_link_bc_retrans(l->bc_sndlink, l, from, to, xmitq);
2297 l->stats.recv_nacks++;
2301 /* Msg for other node => suppress own NACK at next sync if applicable */
2302 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2303 l->nack_state = BC_NACK_SND_SUPPRESS;
2308 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
2310 int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2313 l->backlog[TIPC_LOW_IMPORTANCE].limit = max_t(u16, 50, win);
2314 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = max_t(u16, 100, win * 2);
2315 l->backlog[TIPC_HIGH_IMPORTANCE].limit = max_t(u16, 150, win * 3);
2316 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
2317 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
2321 * link_reset_stats - reset link statistics
2322 * @l: pointer to link
2324 void tipc_link_reset_stats(struct tipc_link *l)
2326 memset(&l->stats, 0, sizeof(l->stats));
2329 static void link_print(struct tipc_link *l, const char *str)
2331 struct sk_buff *hskb = skb_peek(&l->transmq);
2332 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2333 u16 tail = l->snd_nxt - 1;
2335 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2336 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2337 skb_queue_len(&l->transmq), head, tail,
2338 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2341 /* Parse and validate nested (link) properties valid for media, bearer and link
2343 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2347 err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2348 tipc_nl_prop_policy, NULL);
2352 if (props[TIPC_NLA_PROP_PRIO]) {
2355 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2356 if (prio > TIPC_MAX_LINK_PRI)
2360 if (props[TIPC_NLA_PROP_TOL]) {
2363 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2364 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2368 if (props[TIPC_NLA_PROP_WIN]) {
2371 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2372 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
2379 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2382 struct nlattr *stats;
2389 struct nla_map map[] = {
2390 {TIPC_NLA_STATS_RX_INFO, 0},
2391 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2392 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2393 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2394 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2395 {TIPC_NLA_STATS_TX_INFO, 0},
2396 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2397 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2398 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2399 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2400 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2401 s->msg_length_counts : 1},
2402 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2403 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2404 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2405 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2406 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2407 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2408 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2409 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2410 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2411 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
2412 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2413 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2414 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2415 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
2416 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2417 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2418 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2419 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2420 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2421 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2422 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2423 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2424 (s->accu_queue_sz / s->queue_sz_counts) : 0}
2427 stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2431 for (i = 0; i < ARRAY_SIZE(map); i++)
2432 if (nla_put_u32(skb, map[i].key, map[i].val))
2435 nla_nest_end(skb, stats);
2439 nla_nest_cancel(skb, stats);
2444 /* Caller should hold appropriate locks to protect the link */
2445 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2446 struct tipc_link *link, int nlflags)
2448 u32 self = tipc_own_addr(net);
2449 struct nlattr *attrs;
2450 struct nlattr *prop;
2454 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2455 nlflags, TIPC_NL_LINK_GET);
2459 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2463 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2465 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2467 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2469 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2471 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2474 if (tipc_link_is_up(link))
2475 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2478 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2481 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2484 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2486 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2488 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2491 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2493 nla_nest_end(msg->skb, prop);
2495 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2499 nla_nest_end(msg->skb, attrs);
2500 genlmsg_end(msg->skb, hdr);
2505 nla_nest_cancel(msg->skb, prop);
2507 nla_nest_cancel(msg->skb, attrs);
2509 genlmsg_cancel(msg->skb, hdr);
2514 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2515 struct tipc_stats *stats)
2518 struct nlattr *nest;
2525 struct nla_map map[] = {
2526 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2527 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2528 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2529 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2530 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2531 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2532 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2533 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2534 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2535 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2536 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2537 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2538 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2539 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2540 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2541 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2542 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2543 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2544 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2545 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2548 nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2552 for (i = 0; i < ARRAY_SIZE(map); i++)
2553 if (nla_put_u32(skb, map[i].key, map[i].val))
2556 nla_nest_end(skb, nest);
2560 nla_nest_cancel(skb, nest);
2565 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2569 struct nlattr *attrs;
2570 struct nlattr *prop;
2571 struct tipc_net *tn = net_generic(net, tipc_net_id);
2572 u32 bc_mode = tipc_bcast_get_broadcast_mode(net);
2573 u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2574 struct tipc_link *bcl = tn->bcl;
2579 tipc_bcast_lock(net);
2581 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2582 NLM_F_MULTI, TIPC_NL_LINK_GET);
2584 tipc_bcast_unlock(net);
2588 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2592 /* The broadcast link is always up */
2593 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2596 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2598 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2600 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2602 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2605 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2608 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2610 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2612 if (bc_mode & BCLINK_MODE_SEL)
2613 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2616 nla_nest_end(msg->skb, prop);
2618 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2622 tipc_bcast_unlock(net);
2623 nla_nest_end(msg->skb, attrs);
2624 genlmsg_end(msg->skb, hdr);
2629 nla_nest_cancel(msg->skb, prop);
2631 nla_nest_cancel(msg->skb, attrs);
2633 tipc_bcast_unlock(net);
2634 genlmsg_cancel(msg->skb, hdr);
2639 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2640 struct sk_buff_head *xmitq)
2644 l->bc_rcvlink->tolerance = tol;
2646 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2649 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2650 struct sk_buff_head *xmitq)
2653 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2656 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2658 l->abort_limit = limit;
2661 char *tipc_link_name_ext(struct tipc_link *l, char *buf)
2664 scnprintf(buf, TIPC_MAX_LINK_NAME, "null");
2665 else if (link_is_bc_sndlink(l))
2666 scnprintf(buf, TIPC_MAX_LINK_NAME, "broadcast-sender");
2667 else if (link_is_bc_rcvlink(l))
2668 scnprintf(buf, TIPC_MAX_LINK_NAME,
2669 "broadcast-receiver, peer %x", l->addr);
2671 memcpy(buf, l->name, TIPC_MAX_LINK_NAME);
2677 * tipc_link_dump - dump TIPC link data
2678 * @l: tipc link to be dumped
2679 * @dqueues: bitmask to decide if any link queue to be dumped?
2680 * - TIPC_DUMP_NONE: don't dump link queues
2681 * - TIPC_DUMP_TRANSMQ: dump link transmq queue
2682 * - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2683 * - TIPC_DUMP_DEFERDQ: dump link deferd queue
2684 * - TIPC_DUMP_INPUTQ: dump link input queue
2685 * - TIPC_DUMP_WAKEUP: dump link wakeup queue
2686 * - TIPC_DUMP_ALL: dump all the link queues above
2687 * @buf: returned buffer of dump data in format
2689 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2692 size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2693 struct sk_buff_head *list;
2694 struct sk_buff *hskb, *tskb;
2698 i += scnprintf(buf, sz, "link data: (null)\n");
2702 i += scnprintf(buf, sz, "link data: %x", l->addr);
2703 i += scnprintf(buf + i, sz - i, " %x", l->state);
2704 i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2705 i += scnprintf(buf + i, sz - i, " %u", l->session);
2706 i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2707 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2708 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2709 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2710 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2711 i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2712 i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2713 i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2714 i += scnprintf(buf + i, sz - i, " %u", 0);
2715 i += scnprintf(buf + i, sz - i, " %u", 0);
2716 i += scnprintf(buf + i, sz - i, " %u", l->acked);
2719 len = skb_queue_len(list);
2720 hskb = skb_peek(list);
2721 tskb = skb_peek_tail(list);
2722 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2723 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2724 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2727 len = skb_queue_len(list);
2728 hskb = skb_peek(list);
2729 tskb = skb_peek_tail(list);
2730 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2731 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2732 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2734 list = &l->backlogq;
2735 len = skb_queue_len(list);
2736 hskb = skb_peek(list);
2737 tskb = skb_peek_tail(list);
2738 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2739 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2740 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2743 len = skb_queue_len(list);
2744 hskb = skb_peek(list);
2745 tskb = skb_peek_tail(list);
2746 i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2747 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2748 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2750 if (dqueues & TIPC_DUMP_TRANSMQ) {
2751 i += scnprintf(buf + i, sz - i, "transmq: ");
2752 i += tipc_list_dump(&l->transmq, false, buf + i);
2754 if (dqueues & TIPC_DUMP_BACKLOGQ) {
2755 i += scnprintf(buf + i, sz - i,
2756 "backlogq: <%u %u %u %u %u>, ",
2757 l->backlog[TIPC_LOW_IMPORTANCE].len,
2758 l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2759 l->backlog[TIPC_HIGH_IMPORTANCE].len,
2760 l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2761 l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2762 i += tipc_list_dump(&l->backlogq, false, buf + i);
2764 if (dqueues & TIPC_DUMP_DEFERDQ) {
2765 i += scnprintf(buf + i, sz - i, "deferdq: ");
2766 i += tipc_list_dump(&l->deferdq, false, buf + i);
2768 if (dqueues & TIPC_DUMP_INPUTQ) {
2769 i += scnprintf(buf + i, sz - i, "inputq: ");
2770 i += tipc_list_dump(l->inputq, false, buf + i);
2772 if (dqueues & TIPC_DUMP_WAKEUP) {
2773 i += scnprintf(buf + i, sz - i, "wakeup: ");
2774 i += tipc_list_dump(&l->wakeupq, false, buf + i);