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 u16 rcvgap, int tolerance, int priority,
243 struct sk_buff_head *xmitq);
244 static void link_print(struct tipc_link *l, const char *str);
245 static int tipc_link_build_nack_msg(struct tipc_link *l,
246 struct sk_buff_head *xmitq);
247 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
248 struct sk_buff_head *xmitq);
249 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
252 * Simple non-static link routines (i.e. referenced outside this file)
254 bool tipc_link_is_up(struct tipc_link *l)
256 return link_is_up(l);
259 bool tipc_link_peer_is_down(struct tipc_link *l)
261 return l->state == LINK_PEER_RESET;
264 bool tipc_link_is_reset(struct tipc_link *l)
266 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
269 bool tipc_link_is_establishing(struct tipc_link *l)
271 return l->state == LINK_ESTABLISHING;
274 bool tipc_link_is_synching(struct tipc_link *l)
276 return l->state == LINK_SYNCHING;
279 bool tipc_link_is_failingover(struct tipc_link *l)
281 return l->state == LINK_FAILINGOVER;
284 bool tipc_link_is_blocked(struct tipc_link *l)
286 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
289 static bool link_is_bc_sndlink(struct tipc_link *l)
291 return !l->bc_sndlink;
294 static bool link_is_bc_rcvlink(struct tipc_link *l)
296 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
299 int tipc_link_is_active(struct tipc_link *l)
304 void tipc_link_set_active(struct tipc_link *l, bool active)
309 u32 tipc_link_id(struct tipc_link *l)
311 return l->peer_bearer_id << 16 | l->bearer_id;
314 int tipc_link_window(struct tipc_link *l)
319 int tipc_link_prio(struct tipc_link *l)
324 unsigned long tipc_link_tolerance(struct tipc_link *l)
329 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
334 char tipc_link_plane(struct tipc_link *l)
339 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
340 struct tipc_link *uc_l,
341 struct sk_buff_head *xmitq)
343 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
346 rcv_l->acked = snd_l->snd_nxt - 1;
347 snd_l->state = LINK_ESTABLISHED;
348 tipc_link_build_bc_init_msg(uc_l, xmitq);
351 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
352 struct tipc_link *rcv_l,
353 struct sk_buff_head *xmitq)
355 u16 ack = snd_l->snd_nxt - 1;
358 rcv_l->bc_peer_is_up = true;
359 rcv_l->state = LINK_ESTABLISHED;
360 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
361 tipc_link_reset(rcv_l);
362 rcv_l->state = LINK_RESET;
363 if (!snd_l->ackers) {
364 tipc_link_reset(snd_l);
365 snd_l->state = LINK_RESET;
366 __skb_queue_purge(xmitq);
370 int tipc_link_bc_peers(struct tipc_link *l)
375 u16 link_bc_rcv_gap(struct tipc_link *l)
377 struct sk_buff *skb = skb_peek(&l->deferdq);
380 if (more(l->snd_nxt, l->rcv_nxt))
381 gap = l->snd_nxt - l->rcv_nxt;
383 gap = buf_seqno(skb) - l->rcv_nxt;
387 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
392 int tipc_link_mtu(struct tipc_link *l)
397 u16 tipc_link_rcv_nxt(struct tipc_link *l)
402 u16 tipc_link_acked(struct tipc_link *l)
407 char *tipc_link_name(struct tipc_link *l)
413 * tipc_link_create - create a new link
414 * @n: pointer to associated node
415 * @if_name: associated interface name
416 * @bearer_id: id (index) of associated bearer
417 * @tolerance: link tolerance to be used by link
418 * @net_plane: network plane (A,B,c..) this link belongs to
419 * @mtu: mtu to be advertised by link
420 * @priority: priority to be used by link
421 * @window: send window to be used by link
422 * @session: session to be used by link
423 * @ownnode: identity of own node
424 * @peer: node id of peer node
425 * @peer_caps: bitmap describing peer node capabilities
426 * @bc_sndlink: the namespace global link used for broadcast sending
427 * @bc_rcvlink: the peer specific link used for broadcast reception
428 * @inputq: queue to put messages ready for delivery
429 * @namedq: queue to put binding table update messages ready for delivery
430 * @link: return value, pointer to put the created link
432 * Returns true if link was created, otherwise false
434 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
435 int tolerance, char net_plane, u32 mtu, int priority,
436 int window, u32 session, u32 ownnode, u32 peer,
438 struct tipc_link *bc_sndlink,
439 struct tipc_link *bc_rcvlink,
440 struct sk_buff_head *inputq,
441 struct sk_buff_head *namedq,
442 struct tipc_link **link)
446 l = kzalloc(sizeof(*l), GFP_ATOMIC);
450 l->session = session;
452 /* Note: peer i/f name is completed by reset/activate message */
453 sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
454 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
455 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
456 strcpy(l->if_name, if_name);
458 l->peer_caps = peer_caps;
460 l->peer_session = ANY_SESSION;
461 l->bearer_id = bearer_id;
462 l->tolerance = tolerance;
463 l->net_plane = net_plane;
464 l->advertised_mtu = mtu;
466 l->priority = priority;
467 tipc_link_set_queue_limits(l, window);
469 l->bc_sndlink = bc_sndlink;
470 l->bc_rcvlink = bc_rcvlink;
473 l->state = LINK_RESETTING;
474 __skb_queue_head_init(&l->transmq);
475 __skb_queue_head_init(&l->backlogq);
476 __skb_queue_head_init(&l->deferdq);
477 skb_queue_head_init(&l->wakeupq);
478 skb_queue_head_init(l->inputq);
483 * tipc_link_bc_create - create new link to be used for broadcast
484 * @n: pointer to associated node
485 * @mtu: mtu to be used
486 * @window: send window to be used
487 * @inputq: queue to put messages ready for delivery
488 * @namedq: queue to put binding table update messages ready for delivery
489 * @link: return value, pointer to put the created link
491 * Returns true if link was created, otherwise false
493 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
494 int mtu, int window, u16 peer_caps,
495 struct sk_buff_head *inputq,
496 struct sk_buff_head *namedq,
497 struct tipc_link *bc_sndlink,
498 struct tipc_link **link)
502 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
503 0, ownnode, peer, peer_caps, bc_sndlink,
504 NULL, inputq, namedq, link))
508 strcpy(l->name, tipc_bclink_name);
510 l->state = LINK_RESET;
514 /* Broadcast send link is always up */
515 if (link_is_bc_sndlink(l))
516 l->state = LINK_ESTABLISHED;
518 /* Disable replicast if even a single peer doesn't support it */
519 if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
520 tipc_bcast_disable_rcast(net);
526 * tipc_link_fsm_evt - link finite state machine
527 * @l: pointer to link
528 * @evt: state machine event to be processed
530 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
537 case LINK_PEER_RESET_EVT:
538 l->state = LINK_PEER_RESET;
541 l->state = LINK_RESET;
543 case LINK_FAILURE_EVT:
544 case LINK_FAILOVER_BEGIN_EVT:
545 case LINK_ESTABLISH_EVT:
546 case LINK_FAILOVER_END_EVT:
547 case LINK_SYNCH_BEGIN_EVT:
548 case LINK_SYNCH_END_EVT:
555 case LINK_PEER_RESET_EVT:
556 l->state = LINK_ESTABLISHING;
558 case LINK_FAILOVER_BEGIN_EVT:
559 l->state = LINK_FAILINGOVER;
560 case LINK_FAILURE_EVT:
562 case LINK_ESTABLISH_EVT:
563 case LINK_FAILOVER_END_EVT:
565 case LINK_SYNCH_BEGIN_EVT:
566 case LINK_SYNCH_END_EVT:
571 case LINK_PEER_RESET:
574 l->state = LINK_ESTABLISHING;
576 case LINK_PEER_RESET_EVT:
577 case LINK_ESTABLISH_EVT:
578 case LINK_FAILURE_EVT:
580 case LINK_SYNCH_BEGIN_EVT:
581 case LINK_SYNCH_END_EVT:
582 case LINK_FAILOVER_BEGIN_EVT:
583 case LINK_FAILOVER_END_EVT:
588 case LINK_FAILINGOVER:
590 case LINK_FAILOVER_END_EVT:
591 l->state = LINK_RESET;
593 case LINK_PEER_RESET_EVT:
595 case LINK_ESTABLISH_EVT:
596 case LINK_FAILURE_EVT:
598 case LINK_FAILOVER_BEGIN_EVT:
599 case LINK_SYNCH_BEGIN_EVT:
600 case LINK_SYNCH_END_EVT:
605 case LINK_ESTABLISHING:
607 case LINK_ESTABLISH_EVT:
608 l->state = LINK_ESTABLISHED;
610 case LINK_FAILOVER_BEGIN_EVT:
611 l->state = LINK_FAILINGOVER;
614 l->state = LINK_RESET;
616 case LINK_FAILURE_EVT:
617 case LINK_PEER_RESET_EVT:
618 case LINK_SYNCH_BEGIN_EVT:
619 case LINK_FAILOVER_END_EVT:
621 case LINK_SYNCH_END_EVT:
626 case LINK_ESTABLISHED:
628 case LINK_PEER_RESET_EVT:
629 l->state = LINK_PEER_RESET;
630 rc |= TIPC_LINK_DOWN_EVT;
632 case LINK_FAILURE_EVT:
633 l->state = LINK_RESETTING;
634 rc |= TIPC_LINK_DOWN_EVT;
637 l->state = LINK_RESET;
639 case LINK_ESTABLISH_EVT:
640 case LINK_SYNCH_END_EVT:
642 case LINK_SYNCH_BEGIN_EVT:
643 l->state = LINK_SYNCHING;
645 case LINK_FAILOVER_BEGIN_EVT:
646 case LINK_FAILOVER_END_EVT:
653 case LINK_PEER_RESET_EVT:
654 l->state = LINK_PEER_RESET;
655 rc |= TIPC_LINK_DOWN_EVT;
657 case LINK_FAILURE_EVT:
658 l->state = LINK_RESETTING;
659 rc |= TIPC_LINK_DOWN_EVT;
662 l->state = LINK_RESET;
664 case LINK_ESTABLISH_EVT:
665 case LINK_SYNCH_BEGIN_EVT:
667 case LINK_SYNCH_END_EVT:
668 l->state = LINK_ESTABLISHED;
670 case LINK_FAILOVER_BEGIN_EVT:
671 case LINK_FAILOVER_END_EVT:
677 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
681 pr_err("Illegal FSM event %x in state %x on link %s\n",
682 evt, l->state, l->name);
686 /* link_profile_stats - update statistical profiling of traffic
688 static void link_profile_stats(struct tipc_link *l)
691 struct tipc_msg *msg;
694 /* Update counters used in statistical profiling of send traffic */
695 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
696 l->stats.queue_sz_counts++;
698 skb = skb_peek(&l->transmq);
702 length = msg_size(msg);
704 if (msg_user(msg) == MSG_FRAGMENTER) {
705 if (msg_type(msg) != FIRST_FRAGMENT)
707 length = msg_size(msg_get_wrapped(msg));
709 l->stats.msg_lengths_total += length;
710 l->stats.msg_length_counts++;
712 l->stats.msg_length_profile[0]++;
713 else if (length <= 256)
714 l->stats.msg_length_profile[1]++;
715 else if (length <= 1024)
716 l->stats.msg_length_profile[2]++;
717 else if (length <= 4096)
718 l->stats.msg_length_profile[3]++;
719 else if (length <= 16384)
720 l->stats.msg_length_profile[4]++;
721 else if (length <= 32768)
722 l->stats.msg_length_profile[5]++;
724 l->stats.msg_length_profile[6]++;
727 /* tipc_link_timeout - perform periodic task as instructed from node timeout
729 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
736 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
737 u16 bc_acked = l->bc_rcvlink->acked;
738 struct tipc_mon_state *mstate = &l->mon_state;
741 case LINK_ESTABLISHED:
744 link_profile_stats(l);
745 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
746 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
747 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
748 state = bc_acked != bc_snt;
749 state |= l->bc_rcvlink->rcv_unacked;
750 state |= l->rcv_unacked;
751 state |= !skb_queue_empty(&l->transmq);
752 state |= !skb_queue_empty(&l->deferdq);
753 probe = mstate->probing;
754 probe |= l->silent_intv_cnt;
755 if (probe || mstate->monitoring)
756 l->silent_intv_cnt++;
759 setup = l->rst_cnt++ <= 4;
760 setup |= !(l->rst_cnt % 16);
763 case LINK_ESTABLISHING:
767 case LINK_PEER_RESET:
769 case LINK_FAILINGOVER:
775 if (state || probe || setup)
776 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, xmitq);
782 * link_schedule_user - schedule a message sender for wakeup after congestion
784 * @hdr: header of message that is being sent
785 * Create pseudo msg to send back to user when congestion abates
787 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
789 u32 dnode = tipc_own_addr(l->net);
790 u32 dport = msg_origport(hdr);
793 /* Create and schedule wakeup pseudo message */
794 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
795 dnode, l->addr, dport, 0, 0);
798 msg_set_dest_droppable(buf_msg(skb), true);
799 TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
800 skb_queue_tail(&l->wakeupq, skb);
801 l->stats.link_congs++;
806 * link_prepare_wakeup - prepare users for wakeup after congestion
808 * Wake up a number of waiting users, as permitted by available space
811 void link_prepare_wakeup(struct tipc_link *l)
813 struct sk_buff *skb, *tmp;
816 skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
817 imp = TIPC_SKB_CB(skb)->chain_imp;
818 if (l->backlog[imp].len < l->backlog[imp].limit) {
819 skb_unlink(skb, &l->wakeupq);
820 skb_queue_tail(l->inputq, skb);
821 } else if (i++ > 10) {
827 void tipc_link_reset(struct tipc_link *l)
829 l->peer_session = ANY_SESSION;
831 l->mtu = l->advertised_mtu;
832 __skb_queue_purge(&l->transmq);
833 __skb_queue_purge(&l->deferdq);
834 skb_queue_splice_init(&l->wakeupq, l->inputq);
835 __skb_queue_purge(&l->backlogq);
836 l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
837 l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
838 l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
839 l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
840 l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
841 kfree_skb(l->reasm_buf);
842 kfree_skb(l->failover_reasm_skb);
844 l->failover_reasm_skb = NULL;
849 l->silent_intv_cnt = 0;
852 l->bc_peer_is_up = false;
853 memset(&l->mon_state, 0, sizeof(l->mon_state));
854 tipc_link_reset_stats(l);
858 * tipc_link_xmit(): enqueue buffer list according to queue situation
860 * @list: chain of buffers containing message
861 * @xmitq: returned list of packets to be sent by caller
863 * Consumes the buffer chain.
864 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
865 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
867 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
868 struct sk_buff_head *xmitq)
870 struct tipc_msg *hdr = buf_msg(skb_peek(list));
871 unsigned int maxwin = l->window;
872 int imp = msg_importance(hdr);
873 unsigned int mtu = l->mtu;
874 u16 ack = l->rcv_nxt - 1;
875 u16 seqno = l->snd_nxt;
876 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
877 struct sk_buff_head *transmq = &l->transmq;
878 struct sk_buff_head *backlogq = &l->backlogq;
879 struct sk_buff *skb, *_skb, *bskb;
880 int pkt_cnt = skb_queue_len(list);
883 if (unlikely(msg_size(hdr) > mtu)) {
884 skb_queue_purge(list);
888 /* Allow oversubscription of one data msg per source at congestion */
889 if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
890 if (imp == TIPC_SYSTEM_IMPORTANCE) {
891 pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
894 rc = link_schedule_user(l, hdr);
898 l->stats.sent_fragmented++;
899 l->stats.sent_fragments += pkt_cnt;
902 /* Prepare each packet for sending, and add to relevant queue: */
903 while (skb_queue_len(list)) {
904 skb = skb_peek(list);
906 msg_set_seqno(hdr, seqno);
907 msg_set_ack(hdr, ack);
908 msg_set_bcast_ack(hdr, bc_ack);
910 if (likely(skb_queue_len(transmq) < maxwin)) {
911 _skb = skb_clone(skb, GFP_ATOMIC);
913 skb_queue_purge(list);
917 __skb_queue_tail(transmq, skb);
918 __skb_queue_tail(xmitq, _skb);
919 TIPC_SKB_CB(skb)->ackers = l->ackers;
921 l->stats.sent_pkts++;
925 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
926 kfree_skb(__skb_dequeue(list));
927 l->stats.sent_bundled++;
930 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
931 kfree_skb(__skb_dequeue(list));
932 __skb_queue_tail(backlogq, bskb);
933 l->backlog[msg_importance(buf_msg(bskb))].len++;
934 l->stats.sent_bundled++;
935 l->stats.sent_bundles++;
938 l->backlog[imp].len += skb_queue_len(list);
939 skb_queue_splice_tail_init(list, backlogq);
945 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
947 struct sk_buff *skb, *_skb;
948 struct tipc_msg *hdr;
949 u16 seqno = l->snd_nxt;
950 u16 ack = l->rcv_nxt - 1;
951 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
953 while (skb_queue_len(&l->transmq) < l->window) {
954 skb = skb_peek(&l->backlogq);
957 _skb = skb_clone(skb, GFP_ATOMIC);
960 __skb_dequeue(&l->backlogq);
962 l->backlog[msg_importance(hdr)].len--;
963 __skb_queue_tail(&l->transmq, skb);
964 __skb_queue_tail(xmitq, _skb);
965 TIPC_SKB_CB(skb)->ackers = l->ackers;
966 msg_set_seqno(hdr, seqno);
967 msg_set_ack(hdr, ack);
968 msg_set_bcast_ack(hdr, bc_ack);
970 l->stats.sent_pkts++;
976 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
978 struct tipc_msg *hdr = buf_msg(skb);
980 pr_warn("Retransmission failure on link <%s>\n", l->name);
981 link_print(l, "Resetting link ");
982 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
983 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
984 pr_info("sqno %u, prev: %x, src: %x\n",
985 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
988 int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
989 struct sk_buff_head *xmitq)
991 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
992 struct tipc_msg *hdr;
993 u16 ack = l->rcv_nxt - 1;
994 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
999 /* Detect repeated retransmit failures on same packet */
1000 if (likely(l->last_retransm != buf_seqno(skb))) {
1001 l->last_retransm = buf_seqno(skb);
1003 } else if (++l->stale_count > 100) {
1004 link_retransmit_failure(l, skb);
1005 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1008 /* Move forward to where retransmission should start */
1009 skb_queue_walk(&l->transmq, skb) {
1010 if (!less(buf_seqno(skb), from))
1014 skb_queue_walk_from(&l->transmq, skb) {
1015 if (more(buf_seqno(skb), to))
1018 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1021 hdr = buf_msg(_skb);
1022 msg_set_ack(hdr, ack);
1023 msg_set_bcast_ack(hdr, bc_ack);
1024 _skb->priority = TC_PRIO_CONTROL;
1025 __skb_queue_tail(xmitq, _skb);
1026 l->stats.retransmitted++;
1031 /* tipc_data_input - deliver data and name distr msgs to upper layer
1033 * Consumes buffer if message is of right type
1034 * Node lock must be held
1036 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1037 struct sk_buff_head *inputq)
1039 struct tipc_msg *hdr = buf_msg(skb);
1041 switch (msg_user(hdr)) {
1042 case TIPC_LOW_IMPORTANCE:
1043 case TIPC_MEDIUM_IMPORTANCE:
1044 case TIPC_HIGH_IMPORTANCE:
1045 case TIPC_CRITICAL_IMPORTANCE:
1046 if (unlikely(msg_type(hdr) == TIPC_MCAST_MSG)) {
1047 skb_queue_tail(l->bc_rcvlink->inputq, skb);
1051 skb_queue_tail(inputq, skb);
1053 case NAME_DISTRIBUTOR:
1054 l->bc_rcvlink->state = LINK_ESTABLISHED;
1055 skb_queue_tail(l->namedq, skb);
1058 case TUNNEL_PROTOCOL:
1059 case MSG_FRAGMENTER:
1060 case BCAST_PROTOCOL:
1063 pr_warn("Dropping received illegal msg type\n");
1069 /* tipc_link_input - process packet that has passed link protocol check
1073 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1074 struct sk_buff_head *inputq)
1076 struct tipc_msg *hdr = buf_msg(skb);
1077 struct sk_buff **reasm_skb = &l->reasm_buf;
1078 struct sk_buff *iskb;
1079 struct sk_buff_head tmpq;
1080 int usr = msg_user(hdr);
1085 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1086 if (msg_type(hdr) == SYNCH_MSG) {
1087 __skb_queue_purge(&l->deferdq);
1090 if (!tipc_msg_extract(skb, &iskb, &ipos))
1095 if (less(msg_seqno(hdr), l->drop_point))
1097 if (tipc_data_input(l, skb, inputq))
1099 usr = msg_user(hdr);
1100 reasm_skb = &l->failover_reasm_skb;
1103 if (usr == MSG_BUNDLER) {
1104 skb_queue_head_init(&tmpq);
1105 l->stats.recv_bundles++;
1106 l->stats.recv_bundled += msg_msgcnt(hdr);
1107 while (tipc_msg_extract(skb, &iskb, &pos))
1108 tipc_data_input(l, iskb, &tmpq);
1109 tipc_skb_queue_splice_tail(&tmpq, inputq);
1111 } else if (usr == MSG_FRAGMENTER) {
1112 l->stats.recv_fragments++;
1113 if (tipc_buf_append(reasm_skb, &skb)) {
1114 l->stats.recv_fragmented++;
1115 tipc_data_input(l, skb, inputq);
1116 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1117 pr_warn_ratelimited("Unable to build fragment list\n");
1118 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1121 } else if (usr == BCAST_PROTOCOL) {
1122 tipc_bcast_lock(l->net);
1123 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1124 tipc_bcast_unlock(l->net);
1131 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1133 bool released = false;
1134 struct sk_buff *skb, *tmp;
1136 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1137 if (more(buf_seqno(skb), acked))
1139 __skb_unlink(skb, &l->transmq);
1146 /* tipc_link_build_state_msg: prepare link state message for transmission
1148 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1149 * risk of ack storms towards the sender
1151 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1156 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1157 if (link_is_bc_rcvlink(l)) {
1158 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1162 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1163 l->snd_nxt = l->rcv_nxt;
1164 return TIPC_LINK_SND_STATE;
1169 l->stats.sent_acks++;
1170 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1174 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1176 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1178 int mtyp = RESET_MSG;
1179 struct sk_buff *skb;
1181 if (l->state == LINK_ESTABLISHING)
1182 mtyp = ACTIVATE_MSG;
1184 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1186 /* Inform peer that this endpoint is going down if applicable */
1187 skb = skb_peek_tail(xmitq);
1188 if (skb && (l->state == LINK_RESET))
1189 msg_set_peer_stopping(buf_msg(skb), 1);
1192 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1193 * Note that sending of broadcast NACK is coordinated among nodes, to
1194 * reduce the risk of NACK storms towards the sender
1196 static int tipc_link_build_nack_msg(struct tipc_link *l,
1197 struct sk_buff_head *xmitq)
1199 u32 def_cnt = ++l->stats.deferred_recv;
1202 if (link_is_bc_rcvlink(l)) {
1203 match1 = def_cnt & 0xf;
1204 match2 = tipc_own_addr(l->net) & 0xf;
1205 if (match1 == match2)
1206 return TIPC_LINK_SND_STATE;
1210 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1211 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1215 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1216 * @l: the link that should handle the message
1218 * @xmitq: queue to place packets to be sent after this call
1220 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1221 struct sk_buff_head *xmitq)
1223 struct sk_buff_head *defq = &l->deferdq;
1224 struct tipc_msg *hdr;
1225 u16 seqno, rcv_nxt, win_lim;
1230 seqno = msg_seqno(hdr);
1231 rcv_nxt = l->rcv_nxt;
1232 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1234 /* Verify and update link state */
1235 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1236 return tipc_link_proto_rcv(l, skb, xmitq);
1238 if (unlikely(!link_is_up(l))) {
1239 if (l->state == LINK_ESTABLISHING)
1240 rc = TIPC_LINK_UP_EVT;
1244 /* Don't send probe at next timeout expiration */
1245 l->silent_intv_cnt = 0;
1247 /* Drop if outside receive window */
1248 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1249 l->stats.duplicates++;
1253 /* Forward queues and wake up waiting users */
1254 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1255 tipc_link_advance_backlog(l, xmitq);
1256 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1257 link_prepare_wakeup(l);
1260 /* Defer delivery if sequence gap */
1261 if (unlikely(seqno != rcv_nxt)) {
1262 __tipc_skb_queue_sorted(defq, seqno, skb);
1263 rc |= tipc_link_build_nack_msg(l, xmitq);
1267 /* Deliver packet */
1269 l->stats.recv_pkts++;
1270 if (!tipc_data_input(l, skb, l->inputq))
1271 rc |= tipc_link_input(l, skb, l->inputq);
1272 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1273 rc |= tipc_link_build_state_msg(l, xmitq);
1274 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1276 } while ((skb = __skb_dequeue(defq)));
1284 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1285 u16 rcvgap, int tolerance, int priority,
1286 struct sk_buff_head *xmitq)
1288 struct tipc_link *bcl = l->bc_rcvlink;
1289 struct sk_buff *skb;
1290 struct tipc_msg *hdr;
1291 struct sk_buff_head *dfq = &l->deferdq;
1292 bool node_up = link_is_up(bcl);
1293 struct tipc_mon_state *mstate = &l->mon_state;
1297 /* Don't send protocol message during reset or link failover */
1298 if (tipc_link_is_blocked(l))
1301 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1304 if (!skb_queue_empty(dfq))
1305 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1307 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1308 tipc_max_domain_size, l->addr,
1309 tipc_own_addr(l->net), 0, 0, 0);
1314 data = msg_data(hdr);
1315 msg_set_session(hdr, l->session);
1316 msg_set_bearer_id(hdr, l->bearer_id);
1317 msg_set_net_plane(hdr, l->net_plane);
1318 msg_set_next_sent(hdr, l->snd_nxt);
1319 msg_set_ack(hdr, l->rcv_nxt - 1);
1320 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1321 msg_set_bc_ack_invalid(hdr, !node_up);
1322 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1323 msg_set_link_tolerance(hdr, tolerance);
1324 msg_set_linkprio(hdr, priority);
1325 msg_set_redundant_link(hdr, node_up);
1326 msg_set_seq_gap(hdr, 0);
1327 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1329 if (mtyp == STATE_MSG) {
1330 msg_set_seq_gap(hdr, rcvgap);
1331 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1332 msg_set_probe(hdr, probe);
1333 tipc_mon_prep(l->net, data, &dlen, mstate, l->bearer_id);
1334 msg_set_size(hdr, INT_H_SIZE + dlen);
1335 skb_trim(skb, INT_H_SIZE + dlen);
1336 l->stats.sent_states++;
1339 /* RESET_MSG or ACTIVATE_MSG */
1340 msg_set_max_pkt(hdr, l->advertised_mtu);
1341 strcpy(data, l->if_name);
1342 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1343 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1346 l->stats.sent_probes++;
1348 l->stats.sent_nacks++;
1349 skb->priority = TC_PRIO_CONTROL;
1350 __skb_queue_tail(xmitq, skb);
1353 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1354 * with contents of the link's transmit and backlog queues.
1356 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1357 int mtyp, struct sk_buff_head *xmitq)
1359 struct sk_buff *skb, *tnlskb;
1360 struct tipc_msg *hdr, tnlhdr;
1361 struct sk_buff_head *queue = &l->transmq;
1362 struct sk_buff_head tmpxq, tnlq;
1363 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1368 skb_queue_head_init(&tnlq);
1369 skb_queue_head_init(&tmpxq);
1371 /* At least one packet required for safe algorithm => add dummy */
1372 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1373 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1374 0, 0, TIPC_ERR_NO_PORT);
1376 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1379 skb_queue_tail(&tnlq, skb);
1380 tipc_link_xmit(l, &tnlq, &tmpxq);
1381 __skb_queue_purge(&tmpxq);
1383 /* Initialize reusable tunnel packet header */
1384 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1385 mtyp, INT_H_SIZE, l->addr);
1386 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1387 msg_set_msgcnt(&tnlhdr, pktcnt);
1388 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1390 /* Wrap each packet into a tunnel packet */
1391 skb_queue_walk(queue, skb) {
1393 if (queue == &l->backlogq)
1394 msg_set_seqno(hdr, seqno++);
1395 pktlen = msg_size(hdr);
1396 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1397 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1399 pr_warn("%sunable to send packet\n", link_co_err);
1402 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1403 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1404 __skb_queue_tail(&tnlq, tnlskb);
1406 if (queue != &l->backlogq) {
1407 queue = &l->backlogq;
1411 tipc_link_xmit(tnl, &tnlq, xmitq);
1413 if (mtyp == FAILOVER_MSG) {
1414 tnl->drop_point = l->rcv_nxt;
1415 tnl->failover_reasm_skb = l->reasm_buf;
1416 l->reasm_buf = NULL;
1420 /* tipc_link_proto_rcv(): receive link level protocol message :
1421 * Note that network plane id propagates through the network, and may
1422 * change at any time. The node with lowest numerical id determines
1425 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1426 struct sk_buff_head *xmitq)
1428 struct tipc_msg *hdr = buf_msg(skb);
1430 u16 ack = msg_ack(hdr);
1431 u16 gap = msg_seq_gap(hdr);
1432 u16 peers_snd_nxt = msg_next_sent(hdr);
1433 u16 peers_tol = msg_link_tolerance(hdr);
1434 u16 peers_prio = msg_linkprio(hdr);
1435 u16 rcv_nxt = l->rcv_nxt;
1436 u16 dlen = msg_data_sz(hdr);
1437 int mtyp = msg_type(hdr);
1442 if (tipc_link_is_blocked(l) || !xmitq)
1445 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1446 l->net_plane = msg_net_plane(hdr);
1450 data = msg_data(hdr);
1455 /* Ignore duplicate RESET with old session number */
1456 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1457 (l->peer_session != ANY_SESSION))
1463 /* Complete own link name with peer's interface name */
1464 if_name = strrchr(l->name, ':') + 1;
1465 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1467 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1469 strncpy(if_name, data, TIPC_MAX_IF_NAME);
1471 /* Update own tolerance if peer indicates a non-zero value */
1472 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1473 l->tolerance = peers_tol;
1475 /* Update own priority if peer's priority is higher */
1476 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1477 l->priority = peers_prio;
1479 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1480 if (msg_peer_stopping(hdr))
1481 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1482 else if ((mtyp == RESET_MSG) || !link_is_up(l))
1483 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1485 /* ACTIVATE_MSG takes up link if it was already locally reset */
1486 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1487 rc = TIPC_LINK_UP_EVT;
1489 l->peer_session = msg_session(hdr);
1490 l->peer_bearer_id = msg_bearer_id(hdr);
1491 if (l->mtu > msg_max_pkt(hdr))
1492 l->mtu = msg_max_pkt(hdr);
1497 /* Update own tolerance if peer indicates a non-zero value */
1498 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1499 l->tolerance = peers_tol;
1501 /* Update own prio if peer indicates a different value */
1502 if ((peers_prio != l->priority) &&
1503 in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
1504 l->priority = peers_prio;
1505 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1508 l->silent_intv_cnt = 0;
1509 l->stats.recv_states++;
1511 l->stats.recv_probes++;
1513 if (!link_is_up(l)) {
1514 if (l->state == LINK_ESTABLISHING)
1515 rc = TIPC_LINK_UP_EVT;
1518 tipc_mon_rcv(l->net, data, dlen, l->addr,
1519 &l->mon_state, l->bearer_id);
1521 /* Send NACK if peer has sent pkts we haven't received yet */
1522 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1523 rcvgap = peers_snd_nxt - l->rcv_nxt;
1524 if (rcvgap || (msg_probe(hdr)))
1525 tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1527 tipc_link_release_pkts(l, ack);
1529 /* If NACK, retransmit will now start at right position */
1531 rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
1532 l->stats.recv_nacks++;
1535 tipc_link_advance_backlog(l, xmitq);
1536 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1537 link_prepare_wakeup(l);
1544 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1546 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1548 struct sk_buff_head *xmitq)
1550 struct sk_buff *skb;
1551 struct tipc_msg *hdr;
1552 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1553 u16 ack = l->rcv_nxt - 1;
1554 u16 gap_to = peers_snd_nxt - 1;
1556 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1557 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1561 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1562 msg_set_bcast_ack(hdr, ack);
1563 msg_set_bcgap_after(hdr, ack);
1565 gap_to = buf_seqno(dfrd_skb) - 1;
1566 msg_set_bcgap_to(hdr, gap_to);
1567 msg_set_non_seq(hdr, bcast);
1568 __skb_queue_tail(xmitq, skb);
1572 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1574 * Give a newly added peer node the sequence number where it should
1575 * start receiving and acking broadcast packets.
1577 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1578 struct sk_buff_head *xmitq)
1580 struct sk_buff_head list;
1582 __skb_queue_head_init(&list);
1583 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1585 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
1586 tipc_link_xmit(l, &list, xmitq);
1589 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1591 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1593 int mtyp = msg_type(hdr);
1594 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1599 if (msg_user(hdr) == BCAST_PROTOCOL) {
1600 l->rcv_nxt = peers_snd_nxt;
1601 l->state = LINK_ESTABLISHED;
1605 if (l->peer_caps & TIPC_BCAST_SYNCH)
1608 if (msg_peer_node_is_up(hdr))
1611 /* Compatibility: accept older, less safe initial synch data */
1612 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1613 l->rcv_nxt = peers_snd_nxt;
1616 /* link_bc_retr eval()- check if the indicated range can be retransmitted now
1617 * - Adjust permitted range if there is overlap with previous retransmission
1619 static bool link_bc_retr_eval(struct tipc_link *l, u16 *from, u16 *to)
1621 unsigned long elapsed = jiffies_to_msecs(jiffies - l->prev_retr);
1623 if (less(*to, *from))
1626 /* New retransmission request */
1627 if ((elapsed > TIPC_BC_RETR_LIMIT) ||
1628 less(*to, l->prev_from) || more(*from, l->prev_to)) {
1629 l->prev_from = *from;
1631 l->prev_retr = jiffies;
1635 /* Inside range of previous retransmit */
1636 if (!less(*from, l->prev_from) && !more(*to, l->prev_to))
1639 /* Fully or partially outside previous range => exclude overlap */
1640 if (less(*from, l->prev_from)) {
1641 *to = l->prev_from - 1;
1642 l->prev_from = *from;
1644 if (more(*to, l->prev_to)) {
1645 *from = l->prev_to + 1;
1648 l->prev_retr = jiffies;
1652 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1654 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1655 struct sk_buff_head *xmitq)
1657 struct tipc_link *snd_l = l->bc_sndlink;
1658 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1659 u16 from = msg_bcast_ack(hdr) + 1;
1660 u16 to = from + msg_bc_gap(hdr) - 1;
1666 if (!msg_peer_node_is_up(hdr))
1669 /* Open when peer ackowledges our bcast init msg (pkt #1) */
1671 l->bc_peer_is_up = true;
1673 if (!l->bc_peer_is_up)
1676 l->stats.recv_nacks++;
1678 /* Ignore if peers_snd_nxt goes beyond receive window */
1679 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1682 if (link_bc_retr_eval(snd_l, &from, &to))
1683 rc = tipc_link_retrans(snd_l, from, to, xmitq);
1685 l->snd_nxt = peers_snd_nxt;
1686 if (link_bc_rcv_gap(l))
1687 rc |= TIPC_LINK_SND_STATE;
1689 /* Return now if sender supports nack via STATE messages */
1690 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
1693 /* Otherwise, be backwards compatible */
1695 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1696 l->nack_state = BC_NACK_SND_CONDITIONAL;
1700 /* Don't NACK if one was recently sent or peeked */
1701 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1702 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1706 /* Conditionally delay NACK sending until next synch rcv */
1707 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1708 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1709 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1713 /* Send NACK now but suppress next one */
1714 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1715 l->nack_state = BC_NACK_SND_SUPPRESS;
1719 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1720 struct sk_buff_head *xmitq)
1722 struct sk_buff *skb, *tmp;
1723 struct tipc_link *snd_l = l->bc_sndlink;
1725 if (!link_is_up(l) || !l->bc_peer_is_up)
1728 if (!more(acked, l->acked))
1731 /* Skip over packets peer has already acked */
1732 skb_queue_walk(&snd_l->transmq, skb) {
1733 if (more(buf_seqno(skb), l->acked))
1737 /* Update/release the packets peer is acking now */
1738 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1739 if (more(buf_seqno(skb), acked))
1741 if (!--TIPC_SKB_CB(skb)->ackers) {
1742 __skb_unlink(skb, &snd_l->transmq);
1747 tipc_link_advance_backlog(snd_l, xmitq);
1748 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1749 link_prepare_wakeup(snd_l);
1752 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1753 * This function is here for backwards compatibility, since
1754 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
1756 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1757 struct sk_buff_head *xmitq)
1759 struct tipc_msg *hdr = buf_msg(skb);
1760 u32 dnode = msg_destnode(hdr);
1761 int mtyp = msg_type(hdr);
1762 u16 acked = msg_bcast_ack(hdr);
1763 u16 from = acked + 1;
1764 u16 to = msg_bcgap_to(hdr);
1765 u16 peers_snd_nxt = to + 1;
1770 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1773 if (mtyp != STATE_MSG)
1776 if (dnode == tipc_own_addr(l->net)) {
1777 tipc_link_bc_ack_rcv(l, acked, xmitq);
1778 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1779 l->stats.recv_nacks++;
1783 /* Msg for other node => suppress own NACK at next sync if applicable */
1784 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1785 l->nack_state = BC_NACK_SND_SUPPRESS;
1790 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1792 int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1795 l->backlog[TIPC_LOW_IMPORTANCE].limit = max_t(u16, 50, win);
1796 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = max_t(u16, 100, win * 2);
1797 l->backlog[TIPC_HIGH_IMPORTANCE].limit = max_t(u16, 150, win * 3);
1798 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
1799 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1803 * link_reset_stats - reset link statistics
1804 * @l: pointer to link
1806 void tipc_link_reset_stats(struct tipc_link *l)
1808 memset(&l->stats, 0, sizeof(l->stats));
1811 static void link_print(struct tipc_link *l, const char *str)
1813 struct sk_buff *hskb = skb_peek(&l->transmq);
1814 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1815 u16 tail = l->snd_nxt - 1;
1817 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1818 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1819 skb_queue_len(&l->transmq), head, tail,
1820 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1823 /* Parse and validate nested (link) properties valid for media, bearer and link
1825 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1829 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1830 tipc_nl_prop_policy, NULL);
1834 if (props[TIPC_NLA_PROP_PRIO]) {
1837 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1838 if (prio > TIPC_MAX_LINK_PRI)
1842 if (props[TIPC_NLA_PROP_TOL]) {
1845 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1846 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1850 if (props[TIPC_NLA_PROP_WIN]) {
1853 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1854 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1861 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1864 struct nlattr *stats;
1871 struct nla_map map[] = {
1872 {TIPC_NLA_STATS_RX_INFO, 0},
1873 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1874 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1875 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1876 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1877 {TIPC_NLA_STATS_TX_INFO, 0},
1878 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1879 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1880 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1881 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1882 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1883 s->msg_length_counts : 1},
1884 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1885 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1886 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1887 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1888 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1889 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1890 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1891 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1892 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1893 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1894 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1895 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1896 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1897 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1898 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1899 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1900 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1901 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1902 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1903 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1904 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1905 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1906 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1909 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1913 for (i = 0; i < ARRAY_SIZE(map); i++)
1914 if (nla_put_u32(skb, map[i].key, map[i].val))
1917 nla_nest_end(skb, stats);
1921 nla_nest_cancel(skb, stats);
1926 /* Caller should hold appropriate locks to protect the link */
1927 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1928 struct tipc_link *link, int nlflags)
1932 struct nlattr *attrs;
1933 struct nlattr *prop;
1934 struct tipc_net *tn = net_generic(net, tipc_net_id);
1936 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1937 nlflags, TIPC_NL_LINK_GET);
1941 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1945 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1947 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1948 tipc_cluster_mask(tn->own_addr)))
1950 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1952 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
1954 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
1957 if (tipc_link_is_up(link))
1958 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1961 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
1964 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1967 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1969 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
1971 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
1974 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1976 nla_nest_end(msg->skb, prop);
1978 err = __tipc_nl_add_stats(msg->skb, &link->stats);
1982 nla_nest_end(msg->skb, attrs);
1983 genlmsg_end(msg->skb, hdr);
1988 nla_nest_cancel(msg->skb, prop);
1990 nla_nest_cancel(msg->skb, attrs);
1992 genlmsg_cancel(msg->skb, hdr);
1997 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
1998 struct tipc_stats *stats)
2001 struct nlattr *nest;
2008 struct nla_map map[] = {
2009 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2010 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2011 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2012 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2013 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2014 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2015 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2016 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2017 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2018 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2019 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2020 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2021 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2022 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2023 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2024 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2025 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2026 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2027 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2028 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2031 nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
2035 for (i = 0; i < ARRAY_SIZE(map); i++)
2036 if (nla_put_u32(skb, map[i].key, map[i].val))
2039 nla_nest_end(skb, nest);
2043 nla_nest_cancel(skb, nest);
2048 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2052 struct nlattr *attrs;
2053 struct nlattr *prop;
2054 struct tipc_net *tn = net_generic(net, tipc_net_id);
2055 struct tipc_link *bcl = tn->bcl;
2060 tipc_bcast_lock(net);
2062 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2063 NLM_F_MULTI, TIPC_NL_LINK_GET);
2065 tipc_bcast_unlock(net);
2069 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
2073 /* The broadcast link is always up */
2074 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2077 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2079 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2081 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2083 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2086 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2089 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2091 nla_nest_end(msg->skb, prop);
2093 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2097 tipc_bcast_unlock(net);
2098 nla_nest_end(msg->skb, attrs);
2099 genlmsg_end(msg->skb, hdr);
2104 nla_nest_cancel(msg->skb, prop);
2106 nla_nest_cancel(msg->skb, attrs);
2108 tipc_bcast_unlock(net);
2109 genlmsg_cancel(msg->skb, hdr);
2114 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2115 struct sk_buff_head *xmitq)
2118 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, tol, 0, xmitq);
2121 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2122 struct sk_buff_head *xmitq)
2125 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, prio, xmitq);
2128 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2130 l->abort_limit = limit;