2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5 * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
42 #include "name_distr.h"
47 #include <linux/pkt_sched.h>
68 u32 link_congs; /* # port sends blocked by congestion */
71 u32 max_queue_sz; /* send queue size high water mark */
72 u32 accu_queue_sz; /* used for send queue size profiling */
73 u32 queue_sz_counts; /* used for send queue size profiling */
74 u32 msg_length_counts; /* used for message length profiling */
75 u32 msg_lengths_total; /* used for message length profiling */
76 u32 msg_length_profile[7]; /* used for msg. length profiling */
80 * struct tipc_link - TIPC link data structure
81 * @addr: network address of link's peer node
82 * @name: link name character string
83 * @media_addr: media address to use when sending messages over link
85 * @net: pointer to namespace struct
86 * @refcnt: reference counter for permanent references (owner node & timer)
87 * @peer_session: link session # being used by peer end of link
88 * @peer_bearer_id: bearer id used by link's peer endpoint
89 * @bearer_id: local bearer id used by link
90 * @tolerance: minimum link continuity loss needed to reset link [in ms]
91 * @abort_limit: # of unacknowledged continuity probes needed to reset link
92 * @state: current state of link FSM
93 * @peer_caps: bitmap describing capabilities of peer node
94 * @silent_intv_cnt: # of timer intervals without any reception from peer
95 * @proto_msg: template for control messages generated by link
96 * @pmsg: convenience pointer to "proto_msg" field
97 * @priority: current link priority
98 * @net_plane: current link network plane ('A' through 'H')
99 * @mon_state: cookie with information needed by link monitor
100 * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
101 * @exp_msg_count: # of tunnelled messages expected during link changeover
102 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
103 * @mtu: current maximum packet size for this link
104 * @advertised_mtu: advertised own mtu when link is being established
105 * @transmitq: queue for sent, non-acked messages
106 * @backlogq: queue for messages waiting to be sent
107 * @snt_nxt: next sequence number to use for outbound messages
108 * @last_retransmitted: sequence number of most recently retransmitted message
109 * @stale_count: # of identical retransmit requests made by peer
110 * @ackers: # of peers that needs to ack each packet before it can be released
111 * @acked: # last packet acked by a certain peer. Used for broadcast.
112 * @rcv_nxt: next sequence number to expect for inbound messages
113 * @deferred_queue: deferred queue saved OOS b'cast message received from node
114 * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
115 * @inputq: buffer queue for messages to be delivered upwards
116 * @namedq: buffer queue for name table messages to be delivered upwards
117 * @next_out: ptr to first unsent outbound message in queue
118 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
119 * @long_msg_seq_no: next identifier to use for outbound fragmented messages
120 * @reasm_buf: head of partially reassembled inbound message fragments
121 * @bc_rcvr: marks that this is a broadcast receiver link
122 * @stats: collects statistics regarding link activity
126 char name[TIPC_MAX_LINK_NAME];
129 /* Management and link supervision data */
140 char if_name[TIPC_MAX_IF_NAME];
143 struct tipc_mon_state mon_state;
148 struct sk_buff *failover_reasm_skb;
150 /* Max packet negotiation */
155 struct sk_buff_head transmq;
156 struct sk_buff_head backlogq;
169 struct sk_buff_head deferdq;
170 struct sk_buff_head *inputq;
171 struct sk_buff_head *namedq;
173 /* Congestion handling */
174 struct sk_buff_head wakeupq;
176 /* Fragmentation/reassembly */
177 struct sk_buff *reasm_buf;
182 struct tipc_link *bc_rcvlink;
183 struct tipc_link *bc_sndlink;
184 unsigned long prev_retr;
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_LIMIT 10 /* [ms] */
211 * Interval between NACKs when packets arrive out of order
213 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
215 /* Wildcard value for link session numbers. When it is known that
216 * peer endpoint is down, any session number must be accepted.
218 #define ANY_SESSION 0x10000
223 LINK_ESTABLISHED = 0xe,
224 LINK_ESTABLISHING = 0xe << 4,
225 LINK_RESET = 0x1 << 8,
226 LINK_RESETTING = 0x2 << 12,
227 LINK_PEER_RESET = 0xd << 16,
228 LINK_FAILINGOVER = 0xf << 20,
229 LINK_SYNCHING = 0xc << 24
232 /* Link FSM state checking routines
234 static int link_is_up(struct tipc_link *l)
236 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
239 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
240 struct sk_buff_head *xmitq);
241 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
242 bool probe_reply, u16 rcvgap,
243 int tolerance, int priority,
244 struct sk_buff_head *xmitq);
245 static void link_print(struct tipc_link *l, const char *str);
246 static int tipc_link_build_nack_msg(struct tipc_link *l,
247 struct sk_buff_head *xmitq);
248 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
249 struct sk_buff_head *xmitq);
250 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
253 * Simple non-static link routines (i.e. referenced outside this file)
255 bool tipc_link_is_up(struct tipc_link *l)
257 return link_is_up(l);
260 bool tipc_link_peer_is_down(struct tipc_link *l)
262 return l->state == LINK_PEER_RESET;
265 bool tipc_link_is_reset(struct tipc_link *l)
267 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
270 bool tipc_link_is_establishing(struct tipc_link *l)
272 return l->state == LINK_ESTABLISHING;
275 bool tipc_link_is_synching(struct tipc_link *l)
277 return l->state == LINK_SYNCHING;
280 bool tipc_link_is_failingover(struct tipc_link *l)
282 return l->state == LINK_FAILINGOVER;
285 bool tipc_link_is_blocked(struct tipc_link *l)
287 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
290 static bool link_is_bc_sndlink(struct tipc_link *l)
292 return !l->bc_sndlink;
295 static bool link_is_bc_rcvlink(struct tipc_link *l)
297 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
300 int tipc_link_is_active(struct tipc_link *l)
305 void tipc_link_set_active(struct tipc_link *l, bool active)
310 u32 tipc_link_id(struct tipc_link *l)
312 return l->peer_bearer_id << 16 | l->bearer_id;
315 int tipc_link_window(struct tipc_link *l)
320 int tipc_link_prio(struct tipc_link *l)
325 unsigned long tipc_link_tolerance(struct tipc_link *l)
330 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
335 char tipc_link_plane(struct tipc_link *l)
340 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
341 struct tipc_link *uc_l,
342 struct sk_buff_head *xmitq)
344 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
347 rcv_l->acked = snd_l->snd_nxt - 1;
348 snd_l->state = LINK_ESTABLISHED;
349 tipc_link_build_bc_init_msg(uc_l, xmitq);
352 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
353 struct tipc_link *rcv_l,
354 struct sk_buff_head *xmitq)
356 u16 ack = snd_l->snd_nxt - 1;
359 rcv_l->bc_peer_is_up = true;
360 rcv_l->state = LINK_ESTABLISHED;
361 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
362 tipc_link_reset(rcv_l);
363 rcv_l->state = LINK_RESET;
364 if (!snd_l->ackers) {
365 tipc_link_reset(snd_l);
366 snd_l->state = LINK_RESET;
367 __skb_queue_purge(xmitq);
371 int tipc_link_bc_peers(struct tipc_link *l)
376 u16 link_bc_rcv_gap(struct tipc_link *l)
378 struct sk_buff *skb = skb_peek(&l->deferdq);
381 if (more(l->snd_nxt, l->rcv_nxt))
382 gap = l->snd_nxt - l->rcv_nxt;
384 gap = buf_seqno(skb) - l->rcv_nxt;
388 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
393 int tipc_link_mtu(struct tipc_link *l)
398 u16 tipc_link_rcv_nxt(struct tipc_link *l)
403 u16 tipc_link_acked(struct tipc_link *l)
408 char *tipc_link_name(struct tipc_link *l)
414 * tipc_link_create - create a new link
415 * @n: pointer to associated node
416 * @if_name: associated interface name
417 * @bearer_id: id (index) of associated bearer
418 * @tolerance: link tolerance to be used by link
419 * @net_plane: network plane (A,B,c..) this link belongs to
420 * @mtu: mtu to be advertised by link
421 * @priority: priority to be used by link
422 * @window: send window to be used by link
423 * @session: session to be used by link
424 * @ownnode: identity of own node
425 * @peer: node id of peer node
426 * @peer_caps: bitmap describing peer node capabilities
427 * @bc_sndlink: the namespace global link used for broadcast sending
428 * @bc_rcvlink: the peer specific link used for broadcast reception
429 * @inputq: queue to put messages ready for delivery
430 * @namedq: queue to put binding table update messages ready for delivery
431 * @link: return value, pointer to put the created link
433 * Returns true if link was created, otherwise false
435 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
436 int tolerance, char net_plane, u32 mtu, int priority,
437 int window, u32 session, u32 self,
438 u32 peer, u8 *peer_id, u16 peer_caps,
439 struct tipc_link *bc_sndlink,
440 struct tipc_link *bc_rcvlink,
441 struct sk_buff_head *inputq,
442 struct sk_buff_head *namedq,
443 struct tipc_link **link)
445 char peer_str[NODE_ID_STR_LEN] = {0,};
446 char self_str[NODE_ID_STR_LEN] = {0,};
449 l = kzalloc(sizeof(*l), GFP_ATOMIC);
453 l->session = session;
455 /* Set link name for unicast links only */
457 tipc_nodeid2string(self_str, tipc_own_id(net));
458 if (strlen(self_str) > 16)
459 sprintf(self_str, "%x", self);
460 tipc_nodeid2string(peer_str, peer_id);
461 if (strlen(peer_str) > 16)
462 sprintf(peer_str, "%x", peer);
464 /* Peer i/f name will be completed by reset/activate message */
465 snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
466 self_str, if_name, peer_str);
468 strcpy(l->if_name, if_name);
470 l->peer_caps = peer_caps;
472 l->peer_session = ANY_SESSION;
473 l->bearer_id = bearer_id;
474 l->tolerance = tolerance;
475 l->net_plane = net_plane;
476 l->advertised_mtu = mtu;
478 l->priority = priority;
479 tipc_link_set_queue_limits(l, window);
481 l->bc_sndlink = bc_sndlink;
482 l->bc_rcvlink = bc_rcvlink;
485 l->state = LINK_RESETTING;
486 __skb_queue_head_init(&l->transmq);
487 __skb_queue_head_init(&l->backlogq);
488 __skb_queue_head_init(&l->deferdq);
489 skb_queue_head_init(&l->wakeupq);
490 skb_queue_head_init(l->inputq);
495 * tipc_link_bc_create - create new link to be used for broadcast
496 * @n: pointer to associated node
497 * @mtu: mtu to be used initially if no peers
498 * @window: send window to be used
499 * @inputq: queue to put messages ready for delivery
500 * @namedq: queue to put binding table update messages ready for delivery
501 * @link: return value, pointer to put the created link
503 * Returns true if link was created, otherwise false
505 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
506 int mtu, int window, u16 peer_caps,
507 struct sk_buff_head *inputq,
508 struct sk_buff_head *namedq,
509 struct tipc_link *bc_sndlink,
510 struct tipc_link **link)
514 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
515 0, ownnode, peer, NULL, peer_caps, bc_sndlink,
516 NULL, inputq, namedq, link))
520 strcpy(l->name, tipc_bclink_name);
522 l->state = LINK_RESET;
526 /* Broadcast send link is always up */
527 if (link_is_bc_sndlink(l))
528 l->state = LINK_ESTABLISHED;
530 /* Disable replicast if even a single peer doesn't support it */
531 if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
532 tipc_bcast_disable_rcast(net);
538 * tipc_link_fsm_evt - link finite state machine
539 * @l: pointer to link
540 * @evt: state machine event to be processed
542 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
549 case LINK_PEER_RESET_EVT:
550 l->state = LINK_PEER_RESET;
553 l->state = LINK_RESET;
555 case LINK_FAILURE_EVT:
556 case LINK_FAILOVER_BEGIN_EVT:
557 case LINK_ESTABLISH_EVT:
558 case LINK_FAILOVER_END_EVT:
559 case LINK_SYNCH_BEGIN_EVT:
560 case LINK_SYNCH_END_EVT:
567 case LINK_PEER_RESET_EVT:
568 l->state = LINK_ESTABLISHING;
570 case LINK_FAILOVER_BEGIN_EVT:
571 l->state = LINK_FAILINGOVER;
572 case LINK_FAILURE_EVT:
574 case LINK_ESTABLISH_EVT:
575 case LINK_FAILOVER_END_EVT:
577 case LINK_SYNCH_BEGIN_EVT:
578 case LINK_SYNCH_END_EVT:
583 case LINK_PEER_RESET:
586 l->state = LINK_ESTABLISHING;
588 case LINK_PEER_RESET_EVT:
589 case LINK_ESTABLISH_EVT:
590 case LINK_FAILURE_EVT:
592 case LINK_SYNCH_BEGIN_EVT:
593 case LINK_SYNCH_END_EVT:
594 case LINK_FAILOVER_BEGIN_EVT:
595 case LINK_FAILOVER_END_EVT:
600 case LINK_FAILINGOVER:
602 case LINK_FAILOVER_END_EVT:
603 l->state = LINK_RESET;
605 case LINK_PEER_RESET_EVT:
607 case LINK_ESTABLISH_EVT:
608 case LINK_FAILURE_EVT:
610 case LINK_FAILOVER_BEGIN_EVT:
611 case LINK_SYNCH_BEGIN_EVT:
612 case LINK_SYNCH_END_EVT:
617 case LINK_ESTABLISHING:
619 case LINK_ESTABLISH_EVT:
620 l->state = LINK_ESTABLISHED;
622 case LINK_FAILOVER_BEGIN_EVT:
623 l->state = LINK_FAILINGOVER;
626 l->state = LINK_RESET;
628 case LINK_FAILURE_EVT:
629 case LINK_PEER_RESET_EVT:
630 case LINK_SYNCH_BEGIN_EVT:
631 case LINK_FAILOVER_END_EVT:
633 case LINK_SYNCH_END_EVT:
638 case LINK_ESTABLISHED:
640 case LINK_PEER_RESET_EVT:
641 l->state = LINK_PEER_RESET;
642 rc |= TIPC_LINK_DOWN_EVT;
644 case LINK_FAILURE_EVT:
645 l->state = LINK_RESETTING;
646 rc |= TIPC_LINK_DOWN_EVT;
649 l->state = LINK_RESET;
651 case LINK_ESTABLISH_EVT:
652 case LINK_SYNCH_END_EVT:
654 case LINK_SYNCH_BEGIN_EVT:
655 l->state = LINK_SYNCHING;
657 case LINK_FAILOVER_BEGIN_EVT:
658 case LINK_FAILOVER_END_EVT:
665 case LINK_PEER_RESET_EVT:
666 l->state = LINK_PEER_RESET;
667 rc |= TIPC_LINK_DOWN_EVT;
669 case LINK_FAILURE_EVT:
670 l->state = LINK_RESETTING;
671 rc |= TIPC_LINK_DOWN_EVT;
674 l->state = LINK_RESET;
676 case LINK_ESTABLISH_EVT:
677 case LINK_SYNCH_BEGIN_EVT:
679 case LINK_SYNCH_END_EVT:
680 l->state = LINK_ESTABLISHED;
682 case LINK_FAILOVER_BEGIN_EVT:
683 case LINK_FAILOVER_END_EVT:
689 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
693 pr_err("Illegal FSM event %x in state %x on link %s\n",
694 evt, l->state, l->name);
698 /* link_profile_stats - update statistical profiling of traffic
700 static void link_profile_stats(struct tipc_link *l)
703 struct tipc_msg *msg;
706 /* Update counters used in statistical profiling of send traffic */
707 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
708 l->stats.queue_sz_counts++;
710 skb = skb_peek(&l->transmq);
714 length = msg_size(msg);
716 if (msg_user(msg) == MSG_FRAGMENTER) {
717 if (msg_type(msg) != FIRST_FRAGMENT)
719 length = msg_size(msg_get_wrapped(msg));
721 l->stats.msg_lengths_total += length;
722 l->stats.msg_length_counts++;
724 l->stats.msg_length_profile[0]++;
725 else if (length <= 256)
726 l->stats.msg_length_profile[1]++;
727 else if (length <= 1024)
728 l->stats.msg_length_profile[2]++;
729 else if (length <= 4096)
730 l->stats.msg_length_profile[3]++;
731 else if (length <= 16384)
732 l->stats.msg_length_profile[4]++;
733 else if (length <= 32768)
734 l->stats.msg_length_profile[5]++;
736 l->stats.msg_length_profile[6]++;
739 /* tipc_link_timeout - perform periodic task as instructed from node timeout
741 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
748 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
749 u16 bc_acked = l->bc_rcvlink->acked;
750 struct tipc_mon_state *mstate = &l->mon_state;
753 case LINK_ESTABLISHED:
756 link_profile_stats(l);
757 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
758 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
759 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
760 state = bc_acked != bc_snt;
761 state |= l->bc_rcvlink->rcv_unacked;
762 state |= l->rcv_unacked;
763 state |= !skb_queue_empty(&l->transmq);
764 state |= !skb_queue_empty(&l->deferdq);
765 probe = mstate->probing;
766 probe |= l->silent_intv_cnt;
767 if (probe || mstate->monitoring)
768 l->silent_intv_cnt++;
771 setup = l->rst_cnt++ <= 4;
772 setup |= !(l->rst_cnt % 16);
775 case LINK_ESTABLISHING:
779 case LINK_PEER_RESET:
781 case LINK_FAILINGOVER:
787 if (state || probe || setup)
788 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
794 * link_schedule_user - schedule a message sender for wakeup after congestion
796 * @hdr: header of message that is being sent
797 * Create pseudo msg to send back to user when congestion abates
799 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
801 u32 dnode = tipc_own_addr(l->net);
802 u32 dport = msg_origport(hdr);
805 /* Create and schedule wakeup pseudo message */
806 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
807 dnode, l->addr, dport, 0, 0);
810 msg_set_dest_droppable(buf_msg(skb), true);
811 TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
812 skb_queue_tail(&l->wakeupq, skb);
813 l->stats.link_congs++;
818 * link_prepare_wakeup - prepare users for wakeup after congestion
820 * Wake up a number of waiting users, as permitted by available space
823 void link_prepare_wakeup(struct tipc_link *l)
825 struct sk_buff *skb, *tmp;
828 skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
829 imp = TIPC_SKB_CB(skb)->chain_imp;
830 if (l->backlog[imp].len < l->backlog[imp].limit) {
831 skb_unlink(skb, &l->wakeupq);
832 skb_queue_tail(l->inputq, skb);
833 } else if (i++ > 10) {
839 void tipc_link_reset(struct tipc_link *l)
841 l->peer_session = ANY_SESSION;
843 l->mtu = l->advertised_mtu;
844 __skb_queue_purge(&l->transmq);
845 __skb_queue_purge(&l->deferdq);
846 skb_queue_splice_init(&l->wakeupq, l->inputq);
847 __skb_queue_purge(&l->backlogq);
848 l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
849 l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
850 l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
851 l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
852 l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
853 kfree_skb(l->reasm_buf);
854 kfree_skb(l->failover_reasm_skb);
856 l->failover_reasm_skb = NULL;
861 l->silent_intv_cnt = 0;
864 l->bc_peer_is_up = false;
865 memset(&l->mon_state, 0, sizeof(l->mon_state));
866 tipc_link_reset_stats(l);
870 * tipc_link_xmit(): enqueue buffer list according to queue situation
872 * @list: chain of buffers containing message
873 * @xmitq: returned list of packets to be sent by caller
875 * Consumes the buffer chain.
876 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
877 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
879 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
880 struct sk_buff_head *xmitq)
882 struct tipc_msg *hdr = buf_msg(skb_peek(list));
883 unsigned int maxwin = l->window;
884 int imp = msg_importance(hdr);
885 unsigned int mtu = l->mtu;
886 u16 ack = l->rcv_nxt - 1;
887 u16 seqno = l->snd_nxt;
888 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
889 struct sk_buff_head *transmq = &l->transmq;
890 struct sk_buff_head *backlogq = &l->backlogq;
891 struct sk_buff *skb, *_skb, *bskb;
892 int pkt_cnt = skb_queue_len(list);
895 if (unlikely(msg_size(hdr) > mtu)) {
896 skb_queue_purge(list);
900 /* Allow oversubscription of one data msg per source at congestion */
901 if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
902 if (imp == TIPC_SYSTEM_IMPORTANCE) {
903 pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
906 rc = link_schedule_user(l, hdr);
910 l->stats.sent_fragmented++;
911 l->stats.sent_fragments += pkt_cnt;
914 /* Prepare each packet for sending, and add to relevant queue: */
915 while (skb_queue_len(list)) {
916 skb = skb_peek(list);
918 msg_set_seqno(hdr, seqno);
919 msg_set_ack(hdr, ack);
920 msg_set_bcast_ack(hdr, bc_ack);
922 if (likely(skb_queue_len(transmq) < maxwin)) {
923 _skb = skb_clone(skb, GFP_ATOMIC);
925 skb_queue_purge(list);
929 __skb_queue_tail(transmq, skb);
930 __skb_queue_tail(xmitq, _skb);
931 TIPC_SKB_CB(skb)->ackers = l->ackers;
933 l->stats.sent_pkts++;
937 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
938 kfree_skb(__skb_dequeue(list));
939 l->stats.sent_bundled++;
942 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
943 kfree_skb(__skb_dequeue(list));
944 __skb_queue_tail(backlogq, bskb);
945 l->backlog[msg_importance(buf_msg(bskb))].len++;
946 l->stats.sent_bundled++;
947 l->stats.sent_bundles++;
950 l->backlog[imp].len += skb_queue_len(list);
951 skb_queue_splice_tail_init(list, backlogq);
957 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
959 struct sk_buff *skb, *_skb;
960 struct tipc_msg *hdr;
961 u16 seqno = l->snd_nxt;
962 u16 ack = l->rcv_nxt - 1;
963 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
965 while (skb_queue_len(&l->transmq) < l->window) {
966 skb = skb_peek(&l->backlogq);
969 _skb = skb_clone(skb, GFP_ATOMIC);
972 __skb_dequeue(&l->backlogq);
974 l->backlog[msg_importance(hdr)].len--;
975 __skb_queue_tail(&l->transmq, skb);
976 __skb_queue_tail(xmitq, _skb);
977 TIPC_SKB_CB(skb)->ackers = l->ackers;
978 msg_set_seqno(hdr, seqno);
979 msg_set_ack(hdr, ack);
980 msg_set_bcast_ack(hdr, bc_ack);
982 l->stats.sent_pkts++;
988 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
990 struct tipc_msg *hdr = buf_msg(skb);
992 pr_warn("Retransmission failure on link <%s>\n", l->name);
993 link_print(l, "State of link ");
994 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
995 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
996 pr_info("sqno %u, prev: %x, src: %x\n",
997 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
1000 int tipc_link_retrans(struct tipc_link *l, struct tipc_link *nacker,
1001 u16 from, u16 to, struct sk_buff_head *xmitq)
1003 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1004 struct tipc_msg *hdr;
1005 u16 ack = l->rcv_nxt - 1;
1006 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1011 /* Detect repeated retransmit failures on same packet */
1012 if (nacker->last_retransm != buf_seqno(skb)) {
1013 nacker->last_retransm = buf_seqno(skb);
1014 nacker->stale_count = 1;
1015 } else if (++nacker->stale_count > 100) {
1016 link_retransmit_failure(l, skb);
1017 nacker->stale_count = 0;
1018 if (link_is_bc_sndlink(l))
1019 return TIPC_LINK_DOWN_EVT;
1020 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1023 /* Move forward to where retransmission should start */
1024 skb_queue_walk(&l->transmq, skb) {
1025 if (!less(buf_seqno(skb), from))
1029 skb_queue_walk_from(&l->transmq, skb) {
1030 if (more(buf_seqno(skb), to))
1033 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1036 hdr = buf_msg(_skb);
1037 msg_set_ack(hdr, ack);
1038 msg_set_bcast_ack(hdr, bc_ack);
1039 _skb->priority = TC_PRIO_CONTROL;
1040 __skb_queue_tail(xmitq, _skb);
1041 l->stats.retransmitted++;
1046 /* tipc_data_input - deliver data and name distr msgs to upper layer
1048 * Consumes buffer if message is of right type
1049 * Node lock must be held
1051 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1052 struct sk_buff_head *inputq)
1054 struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1055 struct tipc_msg *hdr = buf_msg(skb);
1057 switch (msg_user(hdr)) {
1058 case TIPC_LOW_IMPORTANCE:
1059 case TIPC_MEDIUM_IMPORTANCE:
1060 case TIPC_HIGH_IMPORTANCE:
1061 case TIPC_CRITICAL_IMPORTANCE:
1062 if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1063 skb_queue_tail(mc_inputq, skb);
1067 skb_queue_tail(inputq, skb);
1069 case GROUP_PROTOCOL:
1070 skb_queue_tail(mc_inputq, skb);
1072 case NAME_DISTRIBUTOR:
1073 l->bc_rcvlink->state = LINK_ESTABLISHED;
1074 skb_queue_tail(l->namedq, skb);
1077 case TUNNEL_PROTOCOL:
1078 case MSG_FRAGMENTER:
1079 case BCAST_PROTOCOL:
1082 pr_warn("Dropping received illegal msg type\n");
1088 /* tipc_link_input - process packet that has passed link protocol check
1092 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1093 struct sk_buff_head *inputq)
1095 struct tipc_msg *hdr = buf_msg(skb);
1096 struct sk_buff **reasm_skb = &l->reasm_buf;
1097 struct sk_buff *iskb;
1098 struct sk_buff_head tmpq;
1099 int usr = msg_user(hdr);
1104 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1105 if (msg_type(hdr) == SYNCH_MSG) {
1106 __skb_queue_purge(&l->deferdq);
1109 if (!tipc_msg_extract(skb, &iskb, &ipos))
1114 if (less(msg_seqno(hdr), l->drop_point))
1116 if (tipc_data_input(l, skb, inputq))
1118 usr = msg_user(hdr);
1119 reasm_skb = &l->failover_reasm_skb;
1122 if (usr == MSG_BUNDLER) {
1123 skb_queue_head_init(&tmpq);
1124 l->stats.recv_bundles++;
1125 l->stats.recv_bundled += msg_msgcnt(hdr);
1126 while (tipc_msg_extract(skb, &iskb, &pos))
1127 tipc_data_input(l, iskb, &tmpq);
1128 tipc_skb_queue_splice_tail(&tmpq, inputq);
1130 } else if (usr == MSG_FRAGMENTER) {
1131 l->stats.recv_fragments++;
1132 if (tipc_buf_append(reasm_skb, &skb)) {
1133 l->stats.recv_fragmented++;
1134 tipc_data_input(l, skb, inputq);
1135 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1136 pr_warn_ratelimited("Unable to build fragment list\n");
1137 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1140 } else if (usr == BCAST_PROTOCOL) {
1141 tipc_bcast_lock(l->net);
1142 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1143 tipc_bcast_unlock(l->net);
1150 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1152 bool released = false;
1153 struct sk_buff *skb, *tmp;
1155 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1156 if (more(buf_seqno(skb), acked))
1158 __skb_unlink(skb, &l->transmq);
1165 /* tipc_link_build_state_msg: prepare link state message for transmission
1167 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1168 * risk of ack storms towards the sender
1170 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1175 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1176 if (link_is_bc_rcvlink(l)) {
1177 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1181 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1182 l->snd_nxt = l->rcv_nxt;
1183 return TIPC_LINK_SND_STATE;
1188 l->stats.sent_acks++;
1189 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1193 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1195 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1197 int mtyp = RESET_MSG;
1198 struct sk_buff *skb;
1200 if (l->state == LINK_ESTABLISHING)
1201 mtyp = ACTIVATE_MSG;
1203 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1205 /* Inform peer that this endpoint is going down if applicable */
1206 skb = skb_peek_tail(xmitq);
1207 if (skb && (l->state == LINK_RESET))
1208 msg_set_peer_stopping(buf_msg(skb), 1);
1211 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1212 * Note that sending of broadcast NACK is coordinated among nodes, to
1213 * reduce the risk of NACK storms towards the sender
1215 static int tipc_link_build_nack_msg(struct tipc_link *l,
1216 struct sk_buff_head *xmitq)
1218 u32 def_cnt = ++l->stats.deferred_recv;
1221 if (link_is_bc_rcvlink(l)) {
1222 match1 = def_cnt & 0xf;
1223 match2 = tipc_own_addr(l->net) & 0xf;
1224 if (match1 == match2)
1225 return TIPC_LINK_SND_STATE;
1229 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1230 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1234 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1235 * @l: the link that should handle the message
1237 * @xmitq: queue to place packets to be sent after this call
1239 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1240 struct sk_buff_head *xmitq)
1242 struct sk_buff_head *defq = &l->deferdq;
1243 struct tipc_msg *hdr;
1244 u16 seqno, rcv_nxt, win_lim;
1249 seqno = msg_seqno(hdr);
1250 rcv_nxt = l->rcv_nxt;
1251 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1253 /* Verify and update link state */
1254 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1255 return tipc_link_proto_rcv(l, skb, xmitq);
1257 if (unlikely(!link_is_up(l))) {
1258 if (l->state == LINK_ESTABLISHING)
1259 rc = TIPC_LINK_UP_EVT;
1263 /* Don't send probe at next timeout expiration */
1264 l->silent_intv_cnt = 0;
1266 /* Drop if outside receive window */
1267 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1268 l->stats.duplicates++;
1272 /* Forward queues and wake up waiting users */
1273 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1274 tipc_link_advance_backlog(l, xmitq);
1275 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1276 link_prepare_wakeup(l);
1279 /* Defer delivery if sequence gap */
1280 if (unlikely(seqno != rcv_nxt)) {
1281 __tipc_skb_queue_sorted(defq, seqno, skb);
1282 rc |= tipc_link_build_nack_msg(l, xmitq);
1286 /* Deliver packet */
1288 l->stats.recv_pkts++;
1289 if (!tipc_data_input(l, skb, l->inputq))
1290 rc |= tipc_link_input(l, skb, l->inputq);
1291 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1292 rc |= tipc_link_build_state_msg(l, xmitq);
1293 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1295 } while ((skb = __skb_dequeue(defq)));
1303 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1304 bool probe_reply, u16 rcvgap,
1305 int tolerance, int priority,
1306 struct sk_buff_head *xmitq)
1308 struct tipc_link *bcl = l->bc_rcvlink;
1309 struct sk_buff *skb;
1310 struct tipc_msg *hdr;
1311 struct sk_buff_head *dfq = &l->deferdq;
1312 bool node_up = link_is_up(bcl);
1313 struct tipc_mon_state *mstate = &l->mon_state;
1317 /* Don't send protocol message during reset or link failover */
1318 if (tipc_link_is_blocked(l))
1321 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1324 if (!skb_queue_empty(dfq))
1325 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1327 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1328 tipc_max_domain_size, l->addr,
1329 tipc_own_addr(l->net), 0, 0, 0);
1334 data = msg_data(hdr);
1335 msg_set_session(hdr, l->session);
1336 msg_set_bearer_id(hdr, l->bearer_id);
1337 msg_set_net_plane(hdr, l->net_plane);
1338 msg_set_next_sent(hdr, l->snd_nxt);
1339 msg_set_ack(hdr, l->rcv_nxt - 1);
1340 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1341 msg_set_bc_ack_invalid(hdr, !node_up);
1342 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1343 msg_set_link_tolerance(hdr, tolerance);
1344 msg_set_linkprio(hdr, priority);
1345 msg_set_redundant_link(hdr, node_up);
1346 msg_set_seq_gap(hdr, 0);
1347 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1349 if (mtyp == STATE_MSG) {
1350 msg_set_seq_gap(hdr, rcvgap);
1351 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1352 msg_set_probe(hdr, probe);
1353 msg_set_is_keepalive(hdr, probe || probe_reply);
1354 tipc_mon_prep(l->net, data, &dlen, mstate, l->bearer_id);
1355 msg_set_size(hdr, INT_H_SIZE + dlen);
1356 skb_trim(skb, INT_H_SIZE + dlen);
1357 l->stats.sent_states++;
1360 /* RESET_MSG or ACTIVATE_MSG */
1361 msg_set_max_pkt(hdr, l->advertised_mtu);
1362 strcpy(data, l->if_name);
1363 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1364 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1367 l->stats.sent_probes++;
1369 l->stats.sent_nacks++;
1370 skb->priority = TC_PRIO_CONTROL;
1371 __skb_queue_tail(xmitq, skb);
1374 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1375 * with contents of the link's transmit and backlog queues.
1377 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1378 int mtyp, struct sk_buff_head *xmitq)
1380 struct sk_buff *skb, *tnlskb;
1381 struct tipc_msg *hdr, tnlhdr;
1382 struct sk_buff_head *queue = &l->transmq;
1383 struct sk_buff_head tmpxq, tnlq;
1384 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1389 skb_queue_head_init(&tnlq);
1390 skb_queue_head_init(&tmpxq);
1392 /* At least one packet required for safe algorithm => add dummy */
1393 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1394 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1395 0, 0, TIPC_ERR_NO_PORT);
1397 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1400 skb_queue_tail(&tnlq, skb);
1401 tipc_link_xmit(l, &tnlq, &tmpxq);
1402 __skb_queue_purge(&tmpxq);
1404 /* Initialize reusable tunnel packet header */
1405 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1406 mtyp, INT_H_SIZE, l->addr);
1407 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1408 msg_set_msgcnt(&tnlhdr, pktcnt);
1409 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1411 /* Wrap each packet into a tunnel packet */
1412 skb_queue_walk(queue, skb) {
1414 if (queue == &l->backlogq)
1415 msg_set_seqno(hdr, seqno++);
1416 pktlen = msg_size(hdr);
1417 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1418 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1420 pr_warn("%sunable to send packet\n", link_co_err);
1423 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1424 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1425 __skb_queue_tail(&tnlq, tnlskb);
1427 if (queue != &l->backlogq) {
1428 queue = &l->backlogq;
1432 tipc_link_xmit(tnl, &tnlq, xmitq);
1434 if (mtyp == FAILOVER_MSG) {
1435 tnl->drop_point = l->rcv_nxt;
1436 tnl->failover_reasm_skb = l->reasm_buf;
1437 l->reasm_buf = NULL;
1441 /* tipc_link_proto_rcv(): receive link level protocol message :
1442 * Note that network plane id propagates through the network, and may
1443 * change at any time. The node with lowest numerical id determines
1446 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1447 struct sk_buff_head *xmitq)
1449 struct tipc_msg *hdr = buf_msg(skb);
1451 u16 ack = msg_ack(hdr);
1452 u16 gap = msg_seq_gap(hdr);
1453 u16 peers_snd_nxt = msg_next_sent(hdr);
1454 u16 peers_tol = msg_link_tolerance(hdr);
1455 u16 peers_prio = msg_linkprio(hdr);
1456 u16 rcv_nxt = l->rcv_nxt;
1457 u16 dlen = msg_data_sz(hdr);
1458 int mtyp = msg_type(hdr);
1459 bool reply = msg_probe(hdr);
1464 if (tipc_link_is_blocked(l) || !xmitq)
1467 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1468 l->net_plane = msg_net_plane(hdr);
1472 data = msg_data(hdr);
1477 /* Ignore duplicate RESET with old session number */
1478 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1479 (l->peer_session != ANY_SESSION))
1485 /* Complete own link name with peer's interface name */
1486 if_name = strrchr(l->name, ':') + 1;
1487 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1489 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1491 strncpy(if_name, data, TIPC_MAX_IF_NAME);
1493 /* Update own tolerance if peer indicates a non-zero value */
1494 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1495 l->tolerance = peers_tol;
1497 /* Update own priority if peer's priority is higher */
1498 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1499 l->priority = peers_prio;
1501 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1502 if (msg_peer_stopping(hdr))
1503 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1504 else if ((mtyp == RESET_MSG) || !link_is_up(l))
1505 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1507 /* ACTIVATE_MSG takes up link if it was already locally reset */
1508 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1509 rc = TIPC_LINK_UP_EVT;
1511 l->peer_session = msg_session(hdr);
1512 l->peer_bearer_id = msg_bearer_id(hdr);
1513 if (l->mtu > msg_max_pkt(hdr))
1514 l->mtu = msg_max_pkt(hdr);
1519 /* Update own tolerance if peer indicates a non-zero value */
1520 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1521 l->tolerance = peers_tol;
1523 /* Update own prio if peer indicates a different value */
1524 if ((peers_prio != l->priority) &&
1525 in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
1526 l->priority = peers_prio;
1527 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1530 l->silent_intv_cnt = 0;
1531 l->stats.recv_states++;
1533 l->stats.recv_probes++;
1535 if (!link_is_up(l)) {
1536 if (l->state == LINK_ESTABLISHING)
1537 rc = TIPC_LINK_UP_EVT;
1540 tipc_mon_rcv(l->net, data, dlen, l->addr,
1541 &l->mon_state, l->bearer_id);
1543 /* Send NACK if peer has sent pkts we haven't received yet */
1544 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1545 rcvgap = peers_snd_nxt - l->rcv_nxt;
1546 if (rcvgap || reply)
1547 tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
1548 rcvgap, 0, 0, xmitq);
1549 tipc_link_release_pkts(l, ack);
1551 /* If NACK, retransmit will now start at right position */
1553 rc = tipc_link_retrans(l, l, ack + 1, ack + gap, xmitq);
1554 l->stats.recv_nacks++;
1557 tipc_link_advance_backlog(l, xmitq);
1558 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1559 link_prepare_wakeup(l);
1566 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1568 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1570 struct sk_buff_head *xmitq)
1572 struct sk_buff *skb;
1573 struct tipc_msg *hdr;
1574 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1575 u16 ack = l->rcv_nxt - 1;
1576 u16 gap_to = peers_snd_nxt - 1;
1578 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1579 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1583 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1584 msg_set_bcast_ack(hdr, ack);
1585 msg_set_bcgap_after(hdr, ack);
1587 gap_to = buf_seqno(dfrd_skb) - 1;
1588 msg_set_bcgap_to(hdr, gap_to);
1589 msg_set_non_seq(hdr, bcast);
1590 __skb_queue_tail(xmitq, skb);
1594 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1596 * Give a newly added peer node the sequence number where it should
1597 * start receiving and acking broadcast packets.
1599 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1600 struct sk_buff_head *xmitq)
1602 struct sk_buff_head list;
1604 __skb_queue_head_init(&list);
1605 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1607 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
1608 tipc_link_xmit(l, &list, xmitq);
1611 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1613 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1615 int mtyp = msg_type(hdr);
1616 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1621 if (msg_user(hdr) == BCAST_PROTOCOL) {
1622 l->rcv_nxt = peers_snd_nxt;
1623 l->state = LINK_ESTABLISHED;
1627 if (l->peer_caps & TIPC_BCAST_SYNCH)
1630 if (msg_peer_node_is_up(hdr))
1633 /* Compatibility: accept older, less safe initial synch data */
1634 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1635 l->rcv_nxt = peers_snd_nxt;
1638 /* link_bc_retr eval()- check if the indicated range can be retransmitted now
1639 * - Adjust permitted range if there is overlap with previous retransmission
1641 static bool link_bc_retr_eval(struct tipc_link *l, u16 *from, u16 *to)
1643 unsigned long elapsed = jiffies_to_msecs(jiffies - l->prev_retr);
1645 if (less(*to, *from))
1648 /* New retransmission request */
1649 if ((elapsed > TIPC_BC_RETR_LIMIT) ||
1650 less(*to, l->prev_from) || more(*from, l->prev_to)) {
1651 l->prev_from = *from;
1653 l->prev_retr = jiffies;
1657 /* Inside range of previous retransmit */
1658 if (!less(*from, l->prev_from) && !more(*to, l->prev_to))
1661 /* Fully or partially outside previous range => exclude overlap */
1662 if (less(*from, l->prev_from)) {
1663 *to = l->prev_from - 1;
1664 l->prev_from = *from;
1666 if (more(*to, l->prev_to)) {
1667 *from = l->prev_to + 1;
1670 l->prev_retr = jiffies;
1674 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1676 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1677 struct sk_buff_head *xmitq)
1679 struct tipc_link *snd_l = l->bc_sndlink;
1680 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1681 u16 from = msg_bcast_ack(hdr) + 1;
1682 u16 to = from + msg_bc_gap(hdr) - 1;
1688 if (!msg_peer_node_is_up(hdr))
1691 /* Open when peer ackowledges our bcast init msg (pkt #1) */
1693 l->bc_peer_is_up = true;
1695 if (!l->bc_peer_is_up)
1698 l->stats.recv_nacks++;
1700 /* Ignore if peers_snd_nxt goes beyond receive window */
1701 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1704 if (link_bc_retr_eval(snd_l, &from, &to))
1705 rc = tipc_link_retrans(snd_l, l, from, to, xmitq);
1707 l->snd_nxt = peers_snd_nxt;
1708 if (link_bc_rcv_gap(l))
1709 rc |= TIPC_LINK_SND_STATE;
1711 /* Return now if sender supports nack via STATE messages */
1712 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
1715 /* Otherwise, be backwards compatible */
1717 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1718 l->nack_state = BC_NACK_SND_CONDITIONAL;
1722 /* Don't NACK if one was recently sent or peeked */
1723 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1724 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1728 /* Conditionally delay NACK sending until next synch rcv */
1729 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1730 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1731 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1735 /* Send NACK now but suppress next one */
1736 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1737 l->nack_state = BC_NACK_SND_SUPPRESS;
1741 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1742 struct sk_buff_head *xmitq)
1744 struct sk_buff *skb, *tmp;
1745 struct tipc_link *snd_l = l->bc_sndlink;
1747 if (!link_is_up(l) || !l->bc_peer_is_up)
1750 if (!more(acked, l->acked))
1753 /* Skip over packets peer has already acked */
1754 skb_queue_walk(&snd_l->transmq, skb) {
1755 if (more(buf_seqno(skb), l->acked))
1759 /* Update/release the packets peer is acking now */
1760 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1761 if (more(buf_seqno(skb), acked))
1763 if (!--TIPC_SKB_CB(skb)->ackers) {
1764 __skb_unlink(skb, &snd_l->transmq);
1769 tipc_link_advance_backlog(snd_l, xmitq);
1770 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1771 link_prepare_wakeup(snd_l);
1774 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1775 * This function is here for backwards compatibility, since
1776 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
1778 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1779 struct sk_buff_head *xmitq)
1781 struct tipc_msg *hdr = buf_msg(skb);
1782 u32 dnode = msg_destnode(hdr);
1783 int mtyp = msg_type(hdr);
1784 u16 acked = msg_bcast_ack(hdr);
1785 u16 from = acked + 1;
1786 u16 to = msg_bcgap_to(hdr);
1787 u16 peers_snd_nxt = to + 1;
1792 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1795 if (mtyp != STATE_MSG)
1798 if (dnode == tipc_own_addr(l->net)) {
1799 tipc_link_bc_ack_rcv(l, acked, xmitq);
1800 rc = tipc_link_retrans(l->bc_sndlink, l, from, to, xmitq);
1801 l->stats.recv_nacks++;
1805 /* Msg for other node => suppress own NACK at next sync if applicable */
1806 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1807 l->nack_state = BC_NACK_SND_SUPPRESS;
1812 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1814 int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
1817 l->backlog[TIPC_LOW_IMPORTANCE].limit = max_t(u16, 50, win);
1818 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = max_t(u16, 100, win * 2);
1819 l->backlog[TIPC_HIGH_IMPORTANCE].limit = max_t(u16, 150, win * 3);
1820 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
1821 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1825 * link_reset_stats - reset link statistics
1826 * @l: pointer to link
1828 void tipc_link_reset_stats(struct tipc_link *l)
1830 memset(&l->stats, 0, sizeof(l->stats));
1833 static void link_print(struct tipc_link *l, const char *str)
1835 struct sk_buff *hskb = skb_peek(&l->transmq);
1836 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1837 u16 tail = l->snd_nxt - 1;
1839 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1840 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1841 skb_queue_len(&l->transmq), head, tail,
1842 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1845 /* Parse and validate nested (link) properties valid for media, bearer and link
1847 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1851 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1852 tipc_nl_prop_policy, NULL);
1856 if (props[TIPC_NLA_PROP_PRIO]) {
1859 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1860 if (prio > TIPC_MAX_LINK_PRI)
1864 if (props[TIPC_NLA_PROP_TOL]) {
1867 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1868 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1872 if (props[TIPC_NLA_PROP_WIN]) {
1875 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1876 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1883 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1886 struct nlattr *stats;
1893 struct nla_map map[] = {
1894 {TIPC_NLA_STATS_RX_INFO, 0},
1895 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1896 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1897 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1898 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1899 {TIPC_NLA_STATS_TX_INFO, 0},
1900 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1901 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1902 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1903 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1904 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1905 s->msg_length_counts : 1},
1906 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1907 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1908 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1909 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1910 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1911 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1912 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1913 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1914 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1915 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1916 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1917 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1918 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1919 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1920 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1921 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1922 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1923 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1924 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1925 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1926 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1927 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1928 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1931 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1935 for (i = 0; i < ARRAY_SIZE(map); i++)
1936 if (nla_put_u32(skb, map[i].key, map[i].val))
1939 nla_nest_end(skb, stats);
1943 nla_nest_cancel(skb, stats);
1948 /* Caller should hold appropriate locks to protect the link */
1949 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1950 struct tipc_link *link, int nlflags)
1952 u32 self = tipc_own_addr(net);
1953 struct nlattr *attrs;
1954 struct nlattr *prop;
1958 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1959 nlflags, TIPC_NL_LINK_GET);
1963 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1967 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1969 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
1971 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1973 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
1975 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
1978 if (tipc_link_is_up(link))
1979 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1982 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
1985 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1988 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1990 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
1992 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
1995 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1997 nla_nest_end(msg->skb, prop);
1999 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2003 nla_nest_end(msg->skb, attrs);
2004 genlmsg_end(msg->skb, hdr);
2009 nla_nest_cancel(msg->skb, prop);
2011 nla_nest_cancel(msg->skb, attrs);
2013 genlmsg_cancel(msg->skb, hdr);
2018 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2019 struct tipc_stats *stats)
2022 struct nlattr *nest;
2029 struct nla_map map[] = {
2030 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2031 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2032 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2033 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2034 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2035 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2036 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2037 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2038 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2039 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2040 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2041 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2042 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2043 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2044 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2045 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2046 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2047 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2048 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2049 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2052 nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
2056 for (i = 0; i < ARRAY_SIZE(map); i++)
2057 if (nla_put_u32(skb, map[i].key, map[i].val))
2060 nla_nest_end(skb, nest);
2064 nla_nest_cancel(skb, nest);
2069 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2073 struct nlattr *attrs;
2074 struct nlattr *prop;
2075 struct tipc_net *tn = net_generic(net, tipc_net_id);
2076 struct tipc_link *bcl = tn->bcl;
2081 tipc_bcast_lock(net);
2083 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2084 NLM_F_MULTI, TIPC_NL_LINK_GET);
2086 tipc_bcast_unlock(net);
2090 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
2094 /* The broadcast link is always up */
2095 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2098 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2100 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2102 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2104 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2107 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2110 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2112 nla_nest_end(msg->skb, prop);
2114 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2118 tipc_bcast_unlock(net);
2119 nla_nest_end(msg->skb, attrs);
2120 genlmsg_end(msg->skb, hdr);
2125 nla_nest_cancel(msg->skb, prop);
2127 nla_nest_cancel(msg->skb, attrs);
2129 tipc_bcast_unlock(net);
2130 genlmsg_cancel(msg->skb, hdr);
2135 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2136 struct sk_buff_head *xmitq)
2140 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2143 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2144 struct sk_buff_head *xmitq)
2147 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2150 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2152 l->abort_limit = limit;