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>
50 u32 sent_info; /* used in counting # sent packets */
51 u32 recv_info; /* used in counting # recv'd packets */
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;
188 struct tipc_stats stats;
192 * Error message prefixes
194 static const char *link_co_err = "Link tunneling error, ";
195 static const char *link_rst_msg = "Resetting link ";
197 /* Send states for broadcast NACKs
200 BC_NACK_SND_CONDITIONAL,
201 BC_NACK_SND_UNCONDITIONAL,
202 BC_NACK_SND_SUPPRESS,
206 * Interval between NACKs when packets arrive out of order
208 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
210 /* Wildcard value for link session numbers. When it is known that
211 * peer endpoint is down, any session number must be accepted.
213 #define ANY_SESSION 0x10000
218 LINK_ESTABLISHED = 0xe,
219 LINK_ESTABLISHING = 0xe << 4,
220 LINK_RESET = 0x1 << 8,
221 LINK_RESETTING = 0x2 << 12,
222 LINK_PEER_RESET = 0xd << 16,
223 LINK_FAILINGOVER = 0xf << 20,
224 LINK_SYNCHING = 0xc << 24
227 /* Link FSM state checking routines
229 static int link_is_up(struct tipc_link *l)
231 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
234 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
235 struct sk_buff_head *xmitq);
236 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
237 u16 rcvgap, int tolerance, int priority,
238 struct sk_buff_head *xmitq);
239 static void link_print(struct tipc_link *l, const char *str);
240 static void tipc_link_build_nack_msg(struct tipc_link *l,
241 struct sk_buff_head *xmitq);
242 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
243 struct sk_buff_head *xmitq);
244 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
247 * Simple non-static link routines (i.e. referenced outside this file)
249 bool tipc_link_is_up(struct tipc_link *l)
251 return link_is_up(l);
254 bool tipc_link_peer_is_down(struct tipc_link *l)
256 return l->state == LINK_PEER_RESET;
259 bool tipc_link_is_reset(struct tipc_link *l)
261 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
264 bool tipc_link_is_establishing(struct tipc_link *l)
266 return l->state == LINK_ESTABLISHING;
269 bool tipc_link_is_synching(struct tipc_link *l)
271 return l->state == LINK_SYNCHING;
274 bool tipc_link_is_failingover(struct tipc_link *l)
276 return l->state == LINK_FAILINGOVER;
279 bool tipc_link_is_blocked(struct tipc_link *l)
281 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
284 static bool link_is_bc_sndlink(struct tipc_link *l)
286 return !l->bc_sndlink;
289 static bool link_is_bc_rcvlink(struct tipc_link *l)
291 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
294 int tipc_link_is_active(struct tipc_link *l)
299 void tipc_link_set_active(struct tipc_link *l, bool active)
304 u32 tipc_link_id(struct tipc_link *l)
306 return l->peer_bearer_id << 16 | l->bearer_id;
309 int tipc_link_window(struct tipc_link *l)
314 int tipc_link_prio(struct tipc_link *l)
319 unsigned long tipc_link_tolerance(struct tipc_link *l)
324 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
329 char tipc_link_plane(struct tipc_link *l)
334 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
335 struct tipc_link *uc_l,
336 struct sk_buff_head *xmitq)
338 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
341 rcv_l->acked = snd_l->snd_nxt - 1;
342 snd_l->state = LINK_ESTABLISHED;
343 tipc_link_build_bc_init_msg(uc_l, xmitq);
346 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
347 struct tipc_link *rcv_l,
348 struct sk_buff_head *xmitq)
350 u16 ack = snd_l->snd_nxt - 1;
353 rcv_l->bc_peer_is_up = true;
354 rcv_l->state = LINK_ESTABLISHED;
355 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
356 tipc_link_reset(rcv_l);
357 rcv_l->state = LINK_RESET;
358 if (!snd_l->ackers) {
359 tipc_link_reset(snd_l);
360 snd_l->state = LINK_RESET;
361 __skb_queue_purge(xmitq);
365 int tipc_link_bc_peers(struct tipc_link *l)
370 u16 link_bc_rcv_gap(struct tipc_link *l)
372 struct sk_buff *skb = skb_peek(&l->deferdq);
375 if (more(l->snd_nxt, l->rcv_nxt))
376 gap = l->snd_nxt - l->rcv_nxt;
378 gap = buf_seqno(skb) - l->rcv_nxt;
382 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
387 int tipc_link_mtu(struct tipc_link *l)
392 u16 tipc_link_rcv_nxt(struct tipc_link *l)
397 u16 tipc_link_acked(struct tipc_link *l)
402 char *tipc_link_name(struct tipc_link *l)
408 * tipc_link_create - create a new link
409 * @n: pointer to associated node
410 * @if_name: associated interface name
411 * @bearer_id: id (index) of associated bearer
412 * @tolerance: link tolerance to be used by link
413 * @net_plane: network plane (A,B,c..) this link belongs to
414 * @mtu: mtu to be advertised by link
415 * @priority: priority to be used by link
416 * @window: send window to be used by link
417 * @session: session to be used by link
418 * @ownnode: identity of own node
419 * @peer: node id of peer node
420 * @peer_caps: bitmap describing peer node capabilities
421 * @bc_sndlink: the namespace global link used for broadcast sending
422 * @bc_rcvlink: the peer specific link used for broadcast reception
423 * @inputq: queue to put messages ready for delivery
424 * @namedq: queue to put binding table update messages ready for delivery
425 * @link: return value, pointer to put the created link
427 * Returns true if link was created, otherwise false
429 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
430 int tolerance, char net_plane, u32 mtu, int priority,
431 int window, u32 session, u32 ownnode, u32 peer,
433 struct tipc_link *bc_sndlink,
434 struct tipc_link *bc_rcvlink,
435 struct sk_buff_head *inputq,
436 struct sk_buff_head *namedq,
437 struct tipc_link **link)
441 l = kzalloc(sizeof(*l), GFP_ATOMIC);
445 l->session = session;
447 /* Note: peer i/f name is completed by reset/activate message */
448 sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
449 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
450 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
451 strcpy(l->if_name, if_name);
453 l->peer_caps = peer_caps;
455 l->peer_session = ANY_SESSION;
456 l->bearer_id = bearer_id;
457 l->tolerance = tolerance;
458 l->net_plane = net_plane;
459 l->advertised_mtu = mtu;
461 l->priority = priority;
462 tipc_link_set_queue_limits(l, window);
464 l->bc_sndlink = bc_sndlink;
465 l->bc_rcvlink = bc_rcvlink;
468 l->state = LINK_RESETTING;
469 __skb_queue_head_init(&l->transmq);
470 __skb_queue_head_init(&l->backlogq);
471 __skb_queue_head_init(&l->deferdq);
472 skb_queue_head_init(&l->wakeupq);
473 skb_queue_head_init(l->inputq);
478 * tipc_link_bc_create - create new link to be used for broadcast
479 * @n: pointer to associated node
480 * @mtu: mtu to be used
481 * @window: send window to be used
482 * @inputq: queue to put messages ready for delivery
483 * @namedq: queue to put binding table update messages ready for delivery
484 * @link: return value, pointer to put the created link
486 * Returns true if link was created, otherwise false
488 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
489 int mtu, int window, u16 peer_caps,
490 struct sk_buff_head *inputq,
491 struct sk_buff_head *namedq,
492 struct tipc_link *bc_sndlink,
493 struct tipc_link **link)
497 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
498 0, ownnode, peer, peer_caps, bc_sndlink,
499 NULL, inputq, namedq, link))
503 strcpy(l->name, tipc_bclink_name);
505 l->state = LINK_RESET;
509 /* Broadcast send link is always up */
510 if (link_is_bc_sndlink(l))
511 l->state = LINK_ESTABLISHED;
517 * tipc_link_fsm_evt - link finite state machine
518 * @l: pointer to link
519 * @evt: state machine event to be processed
521 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
528 case LINK_PEER_RESET_EVT:
529 l->state = LINK_PEER_RESET;
532 l->state = LINK_RESET;
534 case LINK_FAILURE_EVT:
535 case LINK_FAILOVER_BEGIN_EVT:
536 case LINK_ESTABLISH_EVT:
537 case LINK_FAILOVER_END_EVT:
538 case LINK_SYNCH_BEGIN_EVT:
539 case LINK_SYNCH_END_EVT:
546 case LINK_PEER_RESET_EVT:
547 l->state = LINK_ESTABLISHING;
549 case LINK_FAILOVER_BEGIN_EVT:
550 l->state = LINK_FAILINGOVER;
551 case LINK_FAILURE_EVT:
553 case LINK_ESTABLISH_EVT:
554 case LINK_FAILOVER_END_EVT:
556 case LINK_SYNCH_BEGIN_EVT:
557 case LINK_SYNCH_END_EVT:
562 case LINK_PEER_RESET:
565 l->state = LINK_ESTABLISHING;
567 case LINK_PEER_RESET_EVT:
568 case LINK_ESTABLISH_EVT:
569 case LINK_FAILURE_EVT:
571 case LINK_SYNCH_BEGIN_EVT:
572 case LINK_SYNCH_END_EVT:
573 case LINK_FAILOVER_BEGIN_EVT:
574 case LINK_FAILOVER_END_EVT:
579 case LINK_FAILINGOVER:
581 case LINK_FAILOVER_END_EVT:
582 l->state = LINK_RESET;
584 case LINK_PEER_RESET_EVT:
586 case LINK_ESTABLISH_EVT:
587 case LINK_FAILURE_EVT:
589 case LINK_FAILOVER_BEGIN_EVT:
590 case LINK_SYNCH_BEGIN_EVT:
591 case LINK_SYNCH_END_EVT:
596 case LINK_ESTABLISHING:
598 case LINK_ESTABLISH_EVT:
599 l->state = LINK_ESTABLISHED;
601 case LINK_FAILOVER_BEGIN_EVT:
602 l->state = LINK_FAILINGOVER;
605 l->state = LINK_RESET;
607 case LINK_FAILURE_EVT:
608 case LINK_PEER_RESET_EVT:
609 case LINK_SYNCH_BEGIN_EVT:
610 case LINK_FAILOVER_END_EVT:
612 case LINK_SYNCH_END_EVT:
617 case LINK_ESTABLISHED:
619 case LINK_PEER_RESET_EVT:
620 l->state = LINK_PEER_RESET;
621 rc |= TIPC_LINK_DOWN_EVT;
623 case LINK_FAILURE_EVT:
624 l->state = LINK_RESETTING;
625 rc |= TIPC_LINK_DOWN_EVT;
628 l->state = LINK_RESET;
630 case LINK_ESTABLISH_EVT:
631 case LINK_SYNCH_END_EVT:
633 case LINK_SYNCH_BEGIN_EVT:
634 l->state = LINK_SYNCHING;
636 case LINK_FAILOVER_BEGIN_EVT:
637 case LINK_FAILOVER_END_EVT:
644 case LINK_PEER_RESET_EVT:
645 l->state = LINK_PEER_RESET;
646 rc |= TIPC_LINK_DOWN_EVT;
648 case LINK_FAILURE_EVT:
649 l->state = LINK_RESETTING;
650 rc |= TIPC_LINK_DOWN_EVT;
653 l->state = LINK_RESET;
655 case LINK_ESTABLISH_EVT:
656 case LINK_SYNCH_BEGIN_EVT:
658 case LINK_SYNCH_END_EVT:
659 l->state = LINK_ESTABLISHED;
661 case LINK_FAILOVER_BEGIN_EVT:
662 case LINK_FAILOVER_END_EVT:
668 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
672 pr_err("Illegal FSM event %x in state %x on link %s\n",
673 evt, l->state, l->name);
677 /* link_profile_stats - update statistical profiling of traffic
679 static void link_profile_stats(struct tipc_link *l)
682 struct tipc_msg *msg;
685 /* Update counters used in statistical profiling of send traffic */
686 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
687 l->stats.queue_sz_counts++;
689 skb = skb_peek(&l->transmq);
693 length = msg_size(msg);
695 if (msg_user(msg) == MSG_FRAGMENTER) {
696 if (msg_type(msg) != FIRST_FRAGMENT)
698 length = msg_size(msg_get_wrapped(msg));
700 l->stats.msg_lengths_total += length;
701 l->stats.msg_length_counts++;
703 l->stats.msg_length_profile[0]++;
704 else if (length <= 256)
705 l->stats.msg_length_profile[1]++;
706 else if (length <= 1024)
707 l->stats.msg_length_profile[2]++;
708 else if (length <= 4096)
709 l->stats.msg_length_profile[3]++;
710 else if (length <= 16384)
711 l->stats.msg_length_profile[4]++;
712 else if (length <= 32768)
713 l->stats.msg_length_profile[5]++;
715 l->stats.msg_length_profile[6]++;
718 /* tipc_link_timeout - perform periodic task as instructed from node timeout
720 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
727 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
728 u16 bc_acked = l->bc_rcvlink->acked;
729 struct tipc_mon_state *mstate = &l->mon_state;
732 case LINK_ESTABLISHED:
735 link_profile_stats(l);
736 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
737 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
738 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
739 state = bc_acked != bc_snt;
740 state |= l->bc_rcvlink->rcv_unacked;
741 state |= l->rcv_unacked;
742 state |= !skb_queue_empty(&l->transmq);
743 state |= !skb_queue_empty(&l->deferdq);
744 probe = mstate->probing;
745 probe |= l->silent_intv_cnt;
746 if (probe || mstate->monitoring)
747 l->silent_intv_cnt++;
750 setup = l->rst_cnt++ <= 4;
751 setup |= !(l->rst_cnt % 16);
754 case LINK_ESTABLISHING:
758 case LINK_PEER_RESET:
760 case LINK_FAILINGOVER:
766 if (state || probe || setup)
767 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, xmitq);
773 * link_schedule_user - schedule a message sender for wakeup after congestion
774 * @link: congested link
775 * @list: message that was attempted sent
776 * Create pseudo msg to send back to user when congestion abates
777 * Does not consume buffer list
779 static int link_schedule_user(struct tipc_link *link, struct sk_buff_head *list)
781 struct tipc_msg *msg = buf_msg(skb_peek(list));
782 int imp = msg_importance(msg);
783 u32 oport = msg_origport(msg);
784 u32 addr = tipc_own_addr(link->net);
787 /* This really cannot happen... */
788 if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
789 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
792 /* Non-blocking sender: */
793 if (TIPC_SKB_CB(skb_peek(list))->wakeup_pending)
796 /* Create and schedule wakeup pseudo message */
797 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
798 addr, addr, oport, 0, 0);
801 TIPC_SKB_CB(skb)->chain_sz = skb_queue_len(list);
802 TIPC_SKB_CB(skb)->chain_imp = imp;
803 skb_queue_tail(&link->wakeupq, skb);
804 link->stats.link_congs++;
809 * link_prepare_wakeup - prepare users for wakeup after congestion
810 * @link: congested link
811 * Move a number of waiting users, as permitted by available space in
812 * the send queue, from link wait queue to node wait queue for wakeup
814 void link_prepare_wakeup(struct tipc_link *l)
816 int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
818 struct sk_buff *skb, *tmp;
820 skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
821 imp = TIPC_SKB_CB(skb)->chain_imp;
822 lim = l->backlog[imp].limit;
823 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
824 if ((pnd[imp] + l->backlog[imp].len) >= lim)
826 skb_unlink(skb, &l->wakeupq);
827 skb_queue_tail(l->inputq, skb);
831 void tipc_link_reset(struct tipc_link *l)
833 l->peer_session = ANY_SESSION;
835 l->mtu = l->advertised_mtu;
836 __skb_queue_purge(&l->transmq);
837 __skb_queue_purge(&l->deferdq);
838 skb_queue_splice_init(&l->wakeupq, l->inputq);
839 __skb_queue_purge(&l->backlogq);
840 l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
841 l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
842 l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
843 l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
844 l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
845 kfree_skb(l->reasm_buf);
846 kfree_skb(l->failover_reasm_skb);
848 l->failover_reasm_skb = NULL;
853 l->silent_intv_cnt = 0;
855 l->stats.recv_info = 0;
857 l->bc_peer_is_up = false;
858 memset(&l->mon_state, 0, sizeof(l->mon_state));
859 tipc_link_reset_stats(l);
863 * tipc_link_xmit(): enqueue buffer list according to queue situation
865 * @list: chain of buffers containing message
866 * @xmitq: returned list of packets to be sent by caller
868 * Consumes the buffer chain, except when returning -ELINKCONG,
869 * since the caller then may want to make more send attempts.
870 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
871 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
873 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
874 struct sk_buff_head *xmitq)
876 struct tipc_msg *hdr = buf_msg(skb_peek(list));
877 unsigned int maxwin = l->window;
878 unsigned int i, imp = msg_importance(hdr);
879 unsigned int mtu = l->mtu;
880 u16 ack = l->rcv_nxt - 1;
881 u16 seqno = l->snd_nxt;
882 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
883 struct sk_buff_head *transmq = &l->transmq;
884 struct sk_buff_head *backlogq = &l->backlogq;
885 struct sk_buff *skb, *_skb, *bskb;
887 /* Match msg importance against this and all higher backlog limits: */
888 if (!skb_queue_empty(backlogq)) {
889 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
890 if (unlikely(l->backlog[i].len >= l->backlog[i].limit))
891 return link_schedule_user(l, list);
894 if (unlikely(msg_size(hdr) > mtu)) {
895 skb_queue_purge(list);
899 /* Prepare each packet for sending, and add to relevant queue: */
900 while (skb_queue_len(list)) {
901 skb = skb_peek(list);
903 msg_set_seqno(hdr, seqno);
904 msg_set_ack(hdr, ack);
905 msg_set_bcast_ack(hdr, bc_ack);
907 if (likely(skb_queue_len(transmq) < maxwin)) {
908 _skb = skb_clone(skb, GFP_ATOMIC);
910 skb_queue_purge(list);
914 __skb_queue_tail(transmq, skb);
915 __skb_queue_tail(xmitq, _skb);
916 TIPC_SKB_CB(skb)->ackers = l->ackers;
921 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
922 kfree_skb(__skb_dequeue(list));
923 l->stats.sent_bundled++;
926 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
927 kfree_skb(__skb_dequeue(list));
928 __skb_queue_tail(backlogq, bskb);
929 l->backlog[msg_importance(buf_msg(bskb))].len++;
930 l->stats.sent_bundled++;
931 l->stats.sent_bundles++;
934 l->backlog[imp].len += skb_queue_len(list);
935 skb_queue_splice_tail_init(list, backlogq);
941 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
943 struct sk_buff *skb, *_skb;
944 struct tipc_msg *hdr;
945 u16 seqno = l->snd_nxt;
946 u16 ack = l->rcv_nxt - 1;
947 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
949 while (skb_queue_len(&l->transmq) < l->window) {
950 skb = skb_peek(&l->backlogq);
953 _skb = skb_clone(skb, GFP_ATOMIC);
956 __skb_dequeue(&l->backlogq);
958 l->backlog[msg_importance(hdr)].len--;
959 __skb_queue_tail(&l->transmq, skb);
960 __skb_queue_tail(xmitq, _skb);
961 TIPC_SKB_CB(skb)->ackers = l->ackers;
962 msg_set_seqno(hdr, seqno);
963 msg_set_ack(hdr, ack);
964 msg_set_bcast_ack(hdr, bc_ack);
971 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
973 struct tipc_msg *hdr = buf_msg(skb);
975 pr_warn("Retransmission failure on link <%s>\n", l->name);
976 link_print(l, "Resetting link ");
977 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
978 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
979 pr_info("sqno %u, prev: %x, src: %x\n",
980 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
983 int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
984 struct sk_buff_head *xmitq)
986 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
987 struct tipc_msg *hdr;
988 u16 ack = l->rcv_nxt - 1;
989 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
994 /* Detect repeated retransmit failures on same packet */
995 if (likely(l->last_retransm != buf_seqno(skb))) {
996 l->last_retransm = buf_seqno(skb);
998 } else if (++l->stale_count > 100) {
999 link_retransmit_failure(l, skb);
1000 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1003 /* Move forward to where retransmission should start */
1004 skb_queue_walk(&l->transmq, skb) {
1005 if (!less(buf_seqno(skb), from))
1009 skb_queue_walk_from(&l->transmq, skb) {
1010 if (more(buf_seqno(skb), to))
1013 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1016 hdr = buf_msg(_skb);
1017 msg_set_ack(hdr, ack);
1018 msg_set_bcast_ack(hdr, bc_ack);
1019 _skb->priority = TC_PRIO_CONTROL;
1020 __skb_queue_tail(xmitq, _skb);
1021 l->stats.retransmitted++;
1026 /* tipc_data_input - deliver data and name distr msgs to upper layer
1028 * Consumes buffer if message is of right type
1029 * Node lock must be held
1031 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1032 struct sk_buff_head *inputq)
1034 switch (msg_user(buf_msg(skb))) {
1035 case TIPC_LOW_IMPORTANCE:
1036 case TIPC_MEDIUM_IMPORTANCE:
1037 case TIPC_HIGH_IMPORTANCE:
1038 case TIPC_CRITICAL_IMPORTANCE:
1040 skb_queue_tail(inputq, skb);
1042 case NAME_DISTRIBUTOR:
1043 l->bc_rcvlink->state = LINK_ESTABLISHED;
1044 skb_queue_tail(l->namedq, skb);
1047 case TUNNEL_PROTOCOL:
1048 case MSG_FRAGMENTER:
1049 case BCAST_PROTOCOL:
1052 pr_warn("Dropping received illegal msg type\n");
1058 /* tipc_link_input - process packet that has passed link protocol check
1062 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1063 struct sk_buff_head *inputq)
1065 struct tipc_msg *hdr = buf_msg(skb);
1066 struct sk_buff **reasm_skb = &l->reasm_buf;
1067 struct sk_buff *iskb;
1068 struct sk_buff_head tmpq;
1069 int usr = msg_user(hdr);
1074 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1075 if (msg_type(hdr) == SYNCH_MSG) {
1076 __skb_queue_purge(&l->deferdq);
1079 if (!tipc_msg_extract(skb, &iskb, &ipos))
1084 if (less(msg_seqno(hdr), l->drop_point))
1086 if (tipc_data_input(l, skb, inputq))
1088 usr = msg_user(hdr);
1089 reasm_skb = &l->failover_reasm_skb;
1092 if (usr == MSG_BUNDLER) {
1093 skb_queue_head_init(&tmpq);
1094 l->stats.recv_bundles++;
1095 l->stats.recv_bundled += msg_msgcnt(hdr);
1096 while (tipc_msg_extract(skb, &iskb, &pos))
1097 tipc_data_input(l, iskb, &tmpq);
1098 tipc_skb_queue_splice_tail(&tmpq, inputq);
1100 } else if (usr == MSG_FRAGMENTER) {
1101 l->stats.recv_fragments++;
1102 if (tipc_buf_append(reasm_skb, &skb)) {
1103 l->stats.recv_fragmented++;
1104 tipc_data_input(l, skb, inputq);
1105 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1106 pr_warn_ratelimited("Unable to build fragment list\n");
1107 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1110 } else if (usr == BCAST_PROTOCOL) {
1111 tipc_bcast_lock(l->net);
1112 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1113 tipc_bcast_unlock(l->net);
1120 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1122 bool released = false;
1123 struct sk_buff *skb, *tmp;
1125 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1126 if (more(buf_seqno(skb), acked))
1128 __skb_unlink(skb, &l->transmq);
1135 /* tipc_link_build_state_msg: prepare link state message for transmission
1137 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1138 * risk of ack storms towards the sender
1140 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1145 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1146 if (link_is_bc_rcvlink(l)) {
1147 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1151 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1152 l->snd_nxt = l->rcv_nxt;
1153 return TIPC_LINK_SND_STATE;
1158 l->stats.sent_acks++;
1159 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1163 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1165 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1167 int mtyp = RESET_MSG;
1168 struct sk_buff *skb;
1170 if (l->state == LINK_ESTABLISHING)
1171 mtyp = ACTIVATE_MSG;
1173 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1175 /* Inform peer that this endpoint is going down if applicable */
1176 skb = skb_peek_tail(xmitq);
1177 if (skb && (l->state == LINK_RESET))
1178 msg_set_peer_stopping(buf_msg(skb), 1);
1181 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1183 static void tipc_link_build_nack_msg(struct tipc_link *l,
1184 struct sk_buff_head *xmitq)
1186 u32 def_cnt = ++l->stats.deferred_recv;
1188 if (link_is_bc_rcvlink(l))
1191 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1192 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1195 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1196 * @l: the link that should handle the message
1198 * @xmitq: queue to place packets to be sent after this call
1200 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1201 struct sk_buff_head *xmitq)
1203 struct sk_buff_head *defq = &l->deferdq;
1204 struct tipc_msg *hdr;
1205 u16 seqno, rcv_nxt, win_lim;
1210 seqno = msg_seqno(hdr);
1211 rcv_nxt = l->rcv_nxt;
1212 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1214 /* Verify and update link state */
1215 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1216 return tipc_link_proto_rcv(l, skb, xmitq);
1218 if (unlikely(!link_is_up(l))) {
1219 if (l->state == LINK_ESTABLISHING)
1220 rc = TIPC_LINK_UP_EVT;
1224 /* Don't send probe at next timeout expiration */
1225 l->silent_intv_cnt = 0;
1227 /* Drop if outside receive window */
1228 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1229 l->stats.duplicates++;
1233 /* Forward queues and wake up waiting users */
1234 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1235 tipc_link_advance_backlog(l, xmitq);
1236 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1237 link_prepare_wakeup(l);
1240 /* Defer delivery if sequence gap */
1241 if (unlikely(seqno != rcv_nxt)) {
1242 __tipc_skb_queue_sorted(defq, seqno, skb);
1243 tipc_link_build_nack_msg(l, xmitq);
1247 /* Deliver packet */
1249 l->stats.recv_info++;
1250 if (!tipc_data_input(l, skb, l->inputq))
1251 rc |= tipc_link_input(l, skb, l->inputq);
1252 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1253 rc |= tipc_link_build_state_msg(l, xmitq);
1254 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1256 } while ((skb = __skb_dequeue(defq)));
1264 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1265 u16 rcvgap, int tolerance, int priority,
1266 struct sk_buff_head *xmitq)
1268 struct tipc_link *bcl = l->bc_rcvlink;
1269 struct sk_buff *skb;
1270 struct tipc_msg *hdr;
1271 struct sk_buff_head *dfq = &l->deferdq;
1272 bool node_up = link_is_up(bcl);
1273 struct tipc_mon_state *mstate = &l->mon_state;
1277 /* Don't send protocol message during reset or link failover */
1278 if (tipc_link_is_blocked(l))
1281 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1284 if (!skb_queue_empty(dfq))
1285 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1287 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1288 tipc_max_domain_size, l->addr,
1289 tipc_own_addr(l->net), 0, 0, 0);
1294 data = msg_data(hdr);
1295 msg_set_session(hdr, l->session);
1296 msg_set_bearer_id(hdr, l->bearer_id);
1297 msg_set_net_plane(hdr, l->net_plane);
1298 msg_set_next_sent(hdr, l->snd_nxt);
1299 msg_set_ack(hdr, l->rcv_nxt - 1);
1300 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1301 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1302 msg_set_link_tolerance(hdr, tolerance);
1303 msg_set_linkprio(hdr, priority);
1304 msg_set_redundant_link(hdr, node_up);
1305 msg_set_seq_gap(hdr, 0);
1306 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1308 if (mtyp == STATE_MSG) {
1309 msg_set_seq_gap(hdr, rcvgap);
1310 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1311 msg_set_probe(hdr, probe);
1312 tipc_mon_prep(l->net, data, &dlen, mstate, l->bearer_id);
1313 msg_set_size(hdr, INT_H_SIZE + dlen);
1314 skb_trim(skb, INT_H_SIZE + dlen);
1315 l->stats.sent_states++;
1318 /* RESET_MSG or ACTIVATE_MSG */
1319 msg_set_max_pkt(hdr, l->advertised_mtu);
1320 strcpy(data, l->if_name);
1321 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1322 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1325 l->stats.sent_probes++;
1327 l->stats.sent_nacks++;
1328 skb->priority = TC_PRIO_CONTROL;
1329 __skb_queue_tail(xmitq, skb);
1332 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1333 * with contents of the link's transmit and backlog queues.
1335 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1336 int mtyp, struct sk_buff_head *xmitq)
1338 struct sk_buff *skb, *tnlskb;
1339 struct tipc_msg *hdr, tnlhdr;
1340 struct sk_buff_head *queue = &l->transmq;
1341 struct sk_buff_head tmpxq, tnlq;
1342 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1347 skb_queue_head_init(&tnlq);
1348 skb_queue_head_init(&tmpxq);
1350 /* At least one packet required for safe algorithm => add dummy */
1351 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1352 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1353 0, 0, TIPC_ERR_NO_PORT);
1355 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1358 skb_queue_tail(&tnlq, skb);
1359 tipc_link_xmit(l, &tnlq, &tmpxq);
1360 __skb_queue_purge(&tmpxq);
1362 /* Initialize reusable tunnel packet header */
1363 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1364 mtyp, INT_H_SIZE, l->addr);
1365 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1366 msg_set_msgcnt(&tnlhdr, pktcnt);
1367 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1369 /* Wrap each packet into a tunnel packet */
1370 skb_queue_walk(queue, skb) {
1372 if (queue == &l->backlogq)
1373 msg_set_seqno(hdr, seqno++);
1374 pktlen = msg_size(hdr);
1375 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1376 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE);
1378 pr_warn("%sunable to send packet\n", link_co_err);
1381 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1382 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1383 __skb_queue_tail(&tnlq, tnlskb);
1385 if (queue != &l->backlogq) {
1386 queue = &l->backlogq;
1390 tipc_link_xmit(tnl, &tnlq, xmitq);
1392 if (mtyp == FAILOVER_MSG) {
1393 tnl->drop_point = l->rcv_nxt;
1394 tnl->failover_reasm_skb = l->reasm_buf;
1395 l->reasm_buf = NULL;
1399 /* tipc_link_proto_rcv(): receive link level protocol message :
1400 * Note that network plane id propagates through the network, and may
1401 * change at any time. The node with lowest numerical id determines
1404 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1405 struct sk_buff_head *xmitq)
1407 struct tipc_msg *hdr = buf_msg(skb);
1409 u16 ack = msg_ack(hdr);
1410 u16 gap = msg_seq_gap(hdr);
1411 u16 peers_snd_nxt = msg_next_sent(hdr);
1412 u16 peers_tol = msg_link_tolerance(hdr);
1413 u16 peers_prio = msg_linkprio(hdr);
1414 u16 rcv_nxt = l->rcv_nxt;
1415 u16 dlen = msg_data_sz(hdr);
1416 int mtyp = msg_type(hdr);
1421 if (tipc_link_is_blocked(l) || !xmitq)
1424 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1425 l->net_plane = msg_net_plane(hdr);
1429 data = msg_data(hdr);
1434 /* Ignore duplicate RESET with old session number */
1435 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1436 (l->peer_session != ANY_SESSION))
1442 /* Complete own link name with peer's interface name */
1443 if_name = strrchr(l->name, ':') + 1;
1444 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1446 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1448 strncpy(if_name, data, TIPC_MAX_IF_NAME);
1450 /* Update own tolerance if peer indicates a non-zero value */
1451 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1452 l->tolerance = peers_tol;
1454 /* Update own priority if peer's priority is higher */
1455 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1456 l->priority = peers_prio;
1458 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1459 if (msg_peer_stopping(hdr))
1460 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1461 else if ((mtyp == RESET_MSG) || !link_is_up(l))
1462 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1464 /* ACTIVATE_MSG takes up link if it was already locally reset */
1465 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1466 rc = TIPC_LINK_UP_EVT;
1468 l->peer_session = msg_session(hdr);
1469 l->peer_bearer_id = msg_bearer_id(hdr);
1470 if (l->mtu > msg_max_pkt(hdr))
1471 l->mtu = msg_max_pkt(hdr);
1476 /* Update own tolerance if peer indicates a non-zero value */
1477 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1478 l->tolerance = peers_tol;
1480 if (peers_prio && in_range(peers_prio, TIPC_MIN_LINK_PRI,
1481 TIPC_MAX_LINK_PRI)) {
1482 l->priority = peers_prio;
1483 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1486 l->silent_intv_cnt = 0;
1487 l->stats.recv_states++;
1489 l->stats.recv_probes++;
1491 if (!link_is_up(l)) {
1492 if (l->state == LINK_ESTABLISHING)
1493 rc = TIPC_LINK_UP_EVT;
1496 tipc_mon_rcv(l->net, data, dlen, l->addr,
1497 &l->mon_state, l->bearer_id);
1499 /* Send NACK if peer has sent pkts we haven't received yet */
1500 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1501 rcvgap = peers_snd_nxt - l->rcv_nxt;
1502 if (rcvgap || (msg_probe(hdr)))
1503 tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1505 tipc_link_release_pkts(l, ack);
1507 /* If NACK, retransmit will now start at right position */
1509 rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
1510 l->stats.recv_nacks++;
1513 tipc_link_advance_backlog(l, xmitq);
1514 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1515 link_prepare_wakeup(l);
1522 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1524 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1526 struct sk_buff_head *xmitq)
1528 struct sk_buff *skb;
1529 struct tipc_msg *hdr;
1530 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1531 u16 ack = l->rcv_nxt - 1;
1532 u16 gap_to = peers_snd_nxt - 1;
1534 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1535 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1539 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1540 msg_set_bcast_ack(hdr, ack);
1541 msg_set_bcgap_after(hdr, ack);
1543 gap_to = buf_seqno(dfrd_skb) - 1;
1544 msg_set_bcgap_to(hdr, gap_to);
1545 msg_set_non_seq(hdr, bcast);
1546 __skb_queue_tail(xmitq, skb);
1550 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1552 * Give a newly added peer node the sequence number where it should
1553 * start receiving and acking broadcast packets.
1555 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1556 struct sk_buff_head *xmitq)
1558 struct sk_buff_head list;
1560 __skb_queue_head_init(&list);
1561 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1563 tipc_link_xmit(l, &list, xmitq);
1566 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1568 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1570 int mtyp = msg_type(hdr);
1571 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1576 if (msg_user(hdr) == BCAST_PROTOCOL) {
1577 l->rcv_nxt = peers_snd_nxt;
1578 l->state = LINK_ESTABLISHED;
1582 if (l->peer_caps & TIPC_BCAST_SYNCH)
1585 if (msg_peer_node_is_up(hdr))
1588 /* Compatibility: accept older, less safe initial synch data */
1589 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1590 l->rcv_nxt = peers_snd_nxt;
1593 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1595 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1596 struct sk_buff_head *xmitq)
1598 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1599 u16 from = msg_bcast_ack(hdr) + 1;
1600 u16 to = from + msg_bc_gap(hdr) - 1;
1606 if (!msg_peer_node_is_up(hdr))
1609 /* Open when peer ackowledges our bcast init msg (pkt #1) */
1611 l->bc_peer_is_up = true;
1613 if (!l->bc_peer_is_up)
1616 /* Ignore if peers_snd_nxt goes beyond receive window */
1617 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1620 if (!less(to, from)) {
1621 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1622 l->stats.recv_nacks++;
1625 l->snd_nxt = peers_snd_nxt;
1626 if (link_bc_rcv_gap(l))
1627 rc |= TIPC_LINK_SND_STATE;
1629 /* Return now if sender supports nack via STATE messages */
1630 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
1633 /* Otherwise, be backwards compatible */
1635 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1636 l->nack_state = BC_NACK_SND_CONDITIONAL;
1640 /* Don't NACK if one was recently sent or peeked */
1641 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1642 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1646 /* Conditionally delay NACK sending until next synch rcv */
1647 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1648 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1649 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1653 /* Send NACK now but suppress next one */
1654 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1655 l->nack_state = BC_NACK_SND_SUPPRESS;
1659 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1660 struct sk_buff_head *xmitq)
1662 struct sk_buff *skb, *tmp;
1663 struct tipc_link *snd_l = l->bc_sndlink;
1665 if (!link_is_up(l) || !l->bc_peer_is_up)
1668 if (!more(acked, l->acked))
1671 /* Skip over packets peer has already acked */
1672 skb_queue_walk(&snd_l->transmq, skb) {
1673 if (more(buf_seqno(skb), l->acked))
1677 /* Update/release the packets peer is acking now */
1678 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1679 if (more(buf_seqno(skb), acked))
1681 if (!--TIPC_SKB_CB(skb)->ackers) {
1682 __skb_unlink(skb, &snd_l->transmq);
1687 tipc_link_advance_backlog(snd_l, xmitq);
1688 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1689 link_prepare_wakeup(snd_l);
1692 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1693 * This function is here for backwards compatibility, since
1694 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
1696 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1697 struct sk_buff_head *xmitq)
1699 struct tipc_msg *hdr = buf_msg(skb);
1700 u32 dnode = msg_destnode(hdr);
1701 int mtyp = msg_type(hdr);
1702 u16 acked = msg_bcast_ack(hdr);
1703 u16 from = acked + 1;
1704 u16 to = msg_bcgap_to(hdr);
1705 u16 peers_snd_nxt = to + 1;
1710 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1713 if (mtyp != STATE_MSG)
1716 if (dnode == tipc_own_addr(l->net)) {
1717 tipc_link_bc_ack_rcv(l, acked, xmitq);
1718 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1719 l->stats.recv_nacks++;
1723 /* Msg for other node => suppress own NACK at next sync if applicable */
1724 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1725 l->nack_state = BC_NACK_SND_SUPPRESS;
1730 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1732 int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1735 l->backlog[TIPC_LOW_IMPORTANCE].limit = max_t(u16, 50, win);
1736 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = max_t(u16, 100, win * 2);
1737 l->backlog[TIPC_HIGH_IMPORTANCE].limit = max_t(u16, 150, win * 3);
1738 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
1739 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1743 * link_reset_stats - reset link statistics
1744 * @l: pointer to link
1746 void tipc_link_reset_stats(struct tipc_link *l)
1748 memset(&l->stats, 0, sizeof(l->stats));
1749 if (!link_is_bc_sndlink(l)) {
1750 l->stats.sent_info = l->snd_nxt;
1751 l->stats.recv_info = l->rcv_nxt;
1755 static void link_print(struct tipc_link *l, const char *str)
1757 struct sk_buff *hskb = skb_peek(&l->transmq);
1758 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1759 u16 tail = l->snd_nxt - 1;
1761 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1762 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1763 skb_queue_len(&l->transmq), head, tail,
1764 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1767 /* Parse and validate nested (link) properties valid for media, bearer and link
1769 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1773 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1774 tipc_nl_prop_policy);
1778 if (props[TIPC_NLA_PROP_PRIO]) {
1781 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1782 if (prio > TIPC_MAX_LINK_PRI)
1786 if (props[TIPC_NLA_PROP_TOL]) {
1789 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1790 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1794 if (props[TIPC_NLA_PROP_WIN]) {
1797 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1798 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1805 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1808 struct nlattr *stats;
1815 struct nla_map map[] = {
1816 {TIPC_NLA_STATS_RX_INFO, s->recv_info},
1817 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1818 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1819 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1820 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1821 {TIPC_NLA_STATS_TX_INFO, s->sent_info},
1822 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1823 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1824 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1825 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1826 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1827 s->msg_length_counts : 1},
1828 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1829 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1830 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1831 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1832 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1833 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1834 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1835 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1836 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1837 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1838 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1839 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1840 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1841 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1842 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1843 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1844 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1845 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1846 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1847 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1848 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1849 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1850 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1853 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1857 for (i = 0; i < ARRAY_SIZE(map); i++)
1858 if (nla_put_u32(skb, map[i].key, map[i].val))
1861 nla_nest_end(skb, stats);
1865 nla_nest_cancel(skb, stats);
1870 /* Caller should hold appropriate locks to protect the link */
1871 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1872 struct tipc_link *link, int nlflags)
1876 struct nlattr *attrs;
1877 struct nlattr *prop;
1878 struct tipc_net *tn = net_generic(net, tipc_net_id);
1880 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1881 nlflags, TIPC_NL_LINK_GET);
1885 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1889 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1891 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1892 tipc_cluster_mask(tn->own_addr)))
1894 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1896 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->rcv_nxt))
1898 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->snd_nxt))
1901 if (tipc_link_is_up(link))
1902 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1905 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
1908 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1911 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1913 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
1915 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
1918 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1920 nla_nest_end(msg->skb, prop);
1922 err = __tipc_nl_add_stats(msg->skb, &link->stats);
1926 nla_nest_end(msg->skb, attrs);
1927 genlmsg_end(msg->skb, hdr);
1932 nla_nest_cancel(msg->skb, prop);
1934 nla_nest_cancel(msg->skb, attrs);
1936 genlmsg_cancel(msg->skb, hdr);
1941 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
1942 struct tipc_stats *stats)
1945 struct nlattr *nest;
1952 struct nla_map map[] = {
1953 {TIPC_NLA_STATS_RX_INFO, stats->recv_info},
1954 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
1955 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
1956 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
1957 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
1958 {TIPC_NLA_STATS_TX_INFO, stats->sent_info},
1959 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
1960 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
1961 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
1962 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
1963 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
1964 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
1965 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
1966 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
1967 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
1968 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
1969 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
1970 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
1971 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
1972 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
1975 nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1979 for (i = 0; i < ARRAY_SIZE(map); i++)
1980 if (nla_put_u32(skb, map[i].key, map[i].val))
1983 nla_nest_end(skb, nest);
1987 nla_nest_cancel(skb, nest);
1992 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
1996 struct nlattr *attrs;
1997 struct nlattr *prop;
1998 struct tipc_net *tn = net_generic(net, tipc_net_id);
1999 struct tipc_link *bcl = tn->bcl;
2004 tipc_bcast_lock(net);
2006 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2007 NLM_F_MULTI, TIPC_NL_LINK_GET);
2009 tipc_bcast_unlock(net);
2013 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
2017 /* The broadcast link is always up */
2018 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2021 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2023 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2025 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, bcl->rcv_nxt))
2027 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, bcl->snd_nxt))
2030 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2033 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2035 nla_nest_end(msg->skb, prop);
2037 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2041 tipc_bcast_unlock(net);
2042 nla_nest_end(msg->skb, attrs);
2043 genlmsg_end(msg->skb, hdr);
2048 nla_nest_cancel(msg->skb, prop);
2050 nla_nest_cancel(msg->skb, attrs);
2052 tipc_bcast_unlock(net);
2053 genlmsg_cancel(msg->skb, hdr);
2058 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2059 struct sk_buff_head *xmitq)
2062 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, tol, 0, xmitq);
2065 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2066 struct sk_buff_head *xmitq)
2069 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, prio, xmitq);
2072 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2074 l->abort_limit = limit;