2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, 2012-2015, 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"
46 #include <linux/pkt_sched.h>
49 u32 sent_info; /* used in counting # sent packets */
50 u32 recv_info; /* used in counting # recv'd packets */
67 u32 link_congs; /* # port sends blocked by congestion */
70 u32 max_queue_sz; /* send queue size high water mark */
71 u32 accu_queue_sz; /* used for send queue size profiling */
72 u32 queue_sz_counts; /* used for send queue size profiling */
73 u32 msg_length_counts; /* used for message length profiling */
74 u32 msg_lengths_total; /* used for message length profiling */
75 u32 msg_length_profile[7]; /* used for msg. length profiling */
79 * struct tipc_link - TIPC link data structure
80 * @addr: network address of link's peer node
81 * @name: link name character string
82 * @media_addr: media address to use when sending messages over link
84 * @net: pointer to namespace struct
85 * @refcnt: reference counter for permanent references (owner node & timer)
86 * @peer_session: link session # being used by peer end of link
87 * @peer_bearer_id: bearer id used by link's peer endpoint
88 * @bearer_id: local bearer id used by link
89 * @tolerance: minimum link continuity loss needed to reset link [in ms]
90 * @keepalive_intv: link keepalive timer interval
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 * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
100 * @exp_msg_count: # of tunnelled messages expected during link changeover
101 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
102 * @mtu: current maximum packet size for this link
103 * @advertised_mtu: advertised own mtu when link is being established
104 * @transmitq: queue for sent, non-acked messages
105 * @backlogq: queue for messages waiting to be sent
106 * @snt_nxt: next sequence number to use for outbound messages
107 * @last_retransmitted: sequence number of most recently retransmitted message
108 * @stale_count: # of identical retransmit requests made by peer
109 * @ackers: # of peers that needs to ack each packet before it can be released
110 * @acked: # last packet acked by a certain peer. Used for broadcast.
111 * @rcv_nxt: next sequence number to expect for inbound messages
112 * @deferred_queue: deferred queue saved OOS b'cast message received from node
113 * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
114 * @inputq: buffer queue for messages to be delivered upwards
115 * @namedq: buffer queue for name table messages to be delivered upwards
116 * @next_out: ptr to first unsent outbound message in queue
117 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
118 * @long_msg_seq_no: next identifier to use for outbound fragmented messages
119 * @reasm_buf: head of partially reassembled inbound message fragments
120 * @bc_rcvr: marks that this is a broadcast receiver link
121 * @stats: collects statistics regarding link activity
125 char name[TIPC_MAX_LINK_NAME];
126 struct tipc_media_addr *media_addr;
129 /* Management and link supervision data */
134 unsigned long keepalive_intv;
141 unchar hdr[INT_H_SIZE];
142 unchar body[TIPC_MAX_IF_NAME];
144 struct tipc_msg *pmsg;
150 struct sk_buff *failover_reasm_skb;
152 /* Max packet negotiation */
157 struct sk_buff_head transmq;
158 struct sk_buff_head backlogq;
171 struct sk_buff_head deferdq;
172 struct sk_buff_head *inputq;
173 struct sk_buff_head *namedq;
175 /* Congestion handling */
176 struct sk_buff_head wakeupq;
178 /* Fragmentation/reassembly */
179 struct sk_buff *reasm_buf;
184 struct tipc_link *bc_rcvlink;
185 struct tipc_link *bc_sndlink;
190 struct tipc_stats stats;
194 * Error message prefixes
196 static const char *link_co_err = "Link tunneling error, ";
197 static const char *link_rst_msg = "Resetting link ";
199 /* Properties valid for media, bearar and link */
200 static const struct nla_policy tipc_nl_prop_policy[TIPC_NLA_PROP_MAX + 1] = {
201 [TIPC_NLA_PROP_UNSPEC] = { .type = NLA_UNSPEC },
202 [TIPC_NLA_PROP_PRIO] = { .type = NLA_U32 },
203 [TIPC_NLA_PROP_TOL] = { .type = NLA_U32 },
204 [TIPC_NLA_PROP_WIN] = { .type = NLA_U32 }
207 /* Send states for broadcast NACKs
210 BC_NACK_SND_CONDITIONAL,
211 BC_NACK_SND_UNCONDITIONAL,
212 BC_NACK_SND_SUPPRESS,
216 * Interval between NACKs when packets arrive out of order
218 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
220 * Out-of-range value for link session numbers
222 #define WILDCARD_SESSION 0x10000
227 LINK_ESTABLISHED = 0xe,
228 LINK_ESTABLISHING = 0xe << 4,
229 LINK_RESET = 0x1 << 8,
230 LINK_RESETTING = 0x2 << 12,
231 LINK_PEER_RESET = 0xd << 16,
232 LINK_FAILINGOVER = 0xf << 20,
233 LINK_SYNCHING = 0xc << 24
236 /* Link FSM state checking routines
238 static int link_is_up(struct tipc_link *l)
240 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
243 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
244 struct sk_buff_head *xmitq);
245 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
246 u16 rcvgap, int tolerance, int priority,
247 struct sk_buff_head *xmitq);
248 static void link_print(struct tipc_link *l_ptr, const char *str);
249 static void tipc_link_build_nack_msg(struct tipc_link *l,
250 struct sk_buff_head *xmitq);
251 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
252 struct sk_buff_head *xmitq);
253 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
256 * Simple non-static link routines (i.e. referenced outside this file)
258 bool tipc_link_is_up(struct tipc_link *l)
260 return link_is_up(l);
263 bool tipc_link_peer_is_down(struct tipc_link *l)
265 return l->state == LINK_PEER_RESET;
268 bool tipc_link_is_reset(struct tipc_link *l)
270 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
273 bool tipc_link_is_establishing(struct tipc_link *l)
275 return l->state == LINK_ESTABLISHING;
278 bool tipc_link_is_synching(struct tipc_link *l)
280 return l->state == LINK_SYNCHING;
283 bool tipc_link_is_failingover(struct tipc_link *l)
285 return l->state == LINK_FAILINGOVER;
288 bool tipc_link_is_blocked(struct tipc_link *l)
290 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
293 static bool link_is_bc_sndlink(struct tipc_link *l)
295 return !l->bc_sndlink;
298 static bool link_is_bc_rcvlink(struct tipc_link *l)
300 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
303 int tipc_link_is_active(struct tipc_link *l)
308 void tipc_link_set_active(struct tipc_link *l, bool active)
313 u32 tipc_link_id(struct tipc_link *l)
315 return l->peer_bearer_id << 16 | l->bearer_id;
318 int tipc_link_window(struct tipc_link *l)
323 int tipc_link_prio(struct tipc_link *l)
328 unsigned long tipc_link_tolerance(struct tipc_link *l)
333 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
338 char tipc_link_plane(struct tipc_link *l)
343 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
344 struct tipc_link *uc_l,
345 struct sk_buff_head *xmitq)
347 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
350 rcv_l->acked = snd_l->snd_nxt - 1;
351 tipc_link_build_bc_init_msg(uc_l, xmitq);
354 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
355 struct tipc_link *rcv_l,
356 struct sk_buff_head *xmitq)
358 u16 ack = snd_l->snd_nxt - 1;
361 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
362 tipc_link_reset(rcv_l);
363 rcv_l->state = LINK_RESET;
364 if (!snd_l->ackers) {
365 tipc_link_reset(snd_l);
366 __skb_queue_purge(xmitq);
370 int tipc_link_bc_peers(struct tipc_link *l)
375 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
380 int tipc_link_mtu(struct tipc_link *l)
385 u16 tipc_link_rcv_nxt(struct tipc_link *l)
390 u16 tipc_link_acked(struct tipc_link *l)
395 char *tipc_link_name(struct tipc_link *l)
400 static u32 link_own_addr(struct tipc_link *l)
402 return msg_prevnode(l->pmsg);
405 void tipc_link_reinit(struct tipc_link *l, u32 addr)
407 msg_set_prevnode(l->pmsg, addr);
411 * tipc_link_create - create a new link
412 * @n: pointer to associated node
413 * @if_name: associated interface name
414 * @bearer_id: id (index) of associated bearer
415 * @tolerance: link tolerance to be used by link
416 * @net_plane: network plane (A,B,c..) this link belongs to
417 * @mtu: mtu to be advertised by link
418 * @priority: priority to be used by link
419 * @window: send window to be used by link
420 * @session: session to be used by link
421 * @ownnode: identity of own node
422 * @peer: node id of peer node
423 * @peer_caps: bitmap describing peer node capabilities
424 * @bc_sndlink: the namespace global link used for broadcast sending
425 * @bc_rcvlink: the peer specific link used for broadcast reception
426 * @inputq: queue to put messages ready for delivery
427 * @namedq: queue to put binding table update messages ready for delivery
428 * @link: return value, pointer to put the created link
430 * Returns true if link was created, otherwise false
432 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
433 int tolerance, char net_plane, u32 mtu, int priority,
434 int window, u32 session, u32 ownnode, u32 peer,
436 struct tipc_link *bc_sndlink,
437 struct tipc_link *bc_rcvlink,
438 struct sk_buff_head *inputq,
439 struct sk_buff_head *namedq,
440 struct tipc_link **link)
443 struct tipc_msg *hdr;
445 l = kzalloc(sizeof(*l), GFP_ATOMIC);
449 l->pmsg = (struct tipc_msg *)&l->proto_msg;
451 tipc_msg_init(ownnode, hdr, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, peer);
452 msg_set_size(hdr, sizeof(l->proto_msg));
453 msg_set_session(hdr, session);
454 msg_set_bearer_id(hdr, l->bearer_id);
456 /* Note: peer i/f name is completed by reset/activate message */
457 sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
458 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
459 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
460 strcpy((char *)msg_data(hdr), if_name);
463 l->peer_caps = peer_caps;
465 l->peer_session = WILDCARD_SESSION;
466 l->bearer_id = bearer_id;
467 l->tolerance = tolerance;
468 l->net_plane = net_plane;
469 l->advertised_mtu = mtu;
471 l->priority = priority;
472 tipc_link_set_queue_limits(l, window);
474 l->bc_sndlink = bc_sndlink;
475 l->bc_rcvlink = bc_rcvlink;
478 l->state = LINK_RESETTING;
479 __skb_queue_head_init(&l->transmq);
480 __skb_queue_head_init(&l->backlogq);
481 __skb_queue_head_init(&l->deferdq);
482 skb_queue_head_init(&l->wakeupq);
483 skb_queue_head_init(l->inputq);
488 * tipc_link_bc_create - create new link to be used for broadcast
489 * @n: pointer to associated node
490 * @mtu: mtu to be used
491 * @window: send window to be used
492 * @inputq: queue to put messages ready for delivery
493 * @namedq: queue to put binding table update messages ready for delivery
494 * @link: return value, pointer to put the created link
496 * Returns true if link was created, otherwise false
498 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
499 int mtu, int window, u16 peer_caps,
500 struct sk_buff_head *inputq,
501 struct sk_buff_head *namedq,
502 struct tipc_link *bc_sndlink,
503 struct tipc_link **link)
507 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
508 0, ownnode, peer, peer_caps, bc_sndlink,
509 NULL, inputq, namedq, link))
513 strcpy(l->name, tipc_bclink_name);
515 l->state = LINK_RESET;
519 /* Broadcast send link is always up */
520 if (link_is_bc_sndlink(l))
521 l->state = LINK_ESTABLISHED;
527 * tipc_link_fsm_evt - link finite state machine
528 * @l: pointer to link
529 * @evt: state machine event to be processed
531 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
538 case LINK_PEER_RESET_EVT:
539 l->state = LINK_PEER_RESET;
542 l->state = LINK_RESET;
544 case LINK_FAILURE_EVT:
545 case LINK_FAILOVER_BEGIN_EVT:
546 case LINK_ESTABLISH_EVT:
547 case LINK_FAILOVER_END_EVT:
548 case LINK_SYNCH_BEGIN_EVT:
549 case LINK_SYNCH_END_EVT:
556 case LINK_PEER_RESET_EVT:
557 l->state = LINK_ESTABLISHING;
559 case LINK_FAILOVER_BEGIN_EVT:
560 l->state = LINK_FAILINGOVER;
561 case LINK_FAILURE_EVT:
563 case LINK_ESTABLISH_EVT:
564 case LINK_FAILOVER_END_EVT:
566 case LINK_SYNCH_BEGIN_EVT:
567 case LINK_SYNCH_END_EVT:
572 case LINK_PEER_RESET:
575 l->state = LINK_ESTABLISHING;
577 case LINK_PEER_RESET_EVT:
578 case LINK_ESTABLISH_EVT:
579 case LINK_FAILURE_EVT:
581 case LINK_SYNCH_BEGIN_EVT:
582 case LINK_SYNCH_END_EVT:
583 case LINK_FAILOVER_BEGIN_EVT:
584 case LINK_FAILOVER_END_EVT:
589 case LINK_FAILINGOVER:
591 case LINK_FAILOVER_END_EVT:
592 l->state = LINK_RESET;
594 case LINK_PEER_RESET_EVT:
596 case LINK_ESTABLISH_EVT:
597 case LINK_FAILURE_EVT:
599 case LINK_FAILOVER_BEGIN_EVT:
600 case LINK_SYNCH_BEGIN_EVT:
601 case LINK_SYNCH_END_EVT:
606 case LINK_ESTABLISHING:
608 case LINK_ESTABLISH_EVT:
609 l->state = LINK_ESTABLISHED;
611 case LINK_FAILOVER_BEGIN_EVT:
612 l->state = LINK_FAILINGOVER;
615 l->state = LINK_RESET;
617 case LINK_FAILURE_EVT:
618 case LINK_PEER_RESET_EVT:
619 case LINK_SYNCH_BEGIN_EVT:
620 case LINK_FAILOVER_END_EVT:
622 case LINK_SYNCH_END_EVT:
627 case LINK_ESTABLISHED:
629 case LINK_PEER_RESET_EVT:
630 l->state = LINK_PEER_RESET;
631 rc |= TIPC_LINK_DOWN_EVT;
633 case LINK_FAILURE_EVT:
634 l->state = LINK_RESETTING;
635 rc |= TIPC_LINK_DOWN_EVT;
638 l->state = LINK_RESET;
640 case LINK_ESTABLISH_EVT:
641 case LINK_SYNCH_END_EVT:
643 case LINK_SYNCH_BEGIN_EVT:
644 l->state = LINK_SYNCHING;
646 case LINK_FAILOVER_BEGIN_EVT:
647 case LINK_FAILOVER_END_EVT:
654 case LINK_PEER_RESET_EVT:
655 l->state = LINK_PEER_RESET;
656 rc |= TIPC_LINK_DOWN_EVT;
658 case LINK_FAILURE_EVT:
659 l->state = LINK_RESETTING;
660 rc |= TIPC_LINK_DOWN_EVT;
663 l->state = LINK_RESET;
665 case LINK_ESTABLISH_EVT:
666 case LINK_SYNCH_BEGIN_EVT:
668 case LINK_SYNCH_END_EVT:
669 l->state = LINK_ESTABLISHED;
671 case LINK_FAILOVER_BEGIN_EVT:
672 case LINK_FAILOVER_END_EVT:
678 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
682 pr_err("Illegal FSM event %x in state %x on link %s\n",
683 evt, l->state, l->name);
687 /* link_profile_stats - update statistical profiling of traffic
689 static void link_profile_stats(struct tipc_link *l)
692 struct tipc_msg *msg;
695 /* Update counters used in statistical profiling of send traffic */
696 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
697 l->stats.queue_sz_counts++;
699 skb = skb_peek(&l->transmq);
703 length = msg_size(msg);
705 if (msg_user(msg) == MSG_FRAGMENTER) {
706 if (msg_type(msg) != FIRST_FRAGMENT)
708 length = msg_size(msg_get_wrapped(msg));
710 l->stats.msg_lengths_total += length;
711 l->stats.msg_length_counts++;
713 l->stats.msg_length_profile[0]++;
714 else if (length <= 256)
715 l->stats.msg_length_profile[1]++;
716 else if (length <= 1024)
717 l->stats.msg_length_profile[2]++;
718 else if (length <= 4096)
719 l->stats.msg_length_profile[3]++;
720 else if (length <= 16384)
721 l->stats.msg_length_profile[4]++;
722 else if (length <= 32768)
723 l->stats.msg_length_profile[5]++;
725 l->stats.msg_length_profile[6]++;
728 /* tipc_link_timeout - perform periodic task as instructed from node timeout
730 /* tipc_link_timeout - perform periodic task as instructed from node timeout
732 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
735 int mtyp = STATE_MSG;
738 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
739 u16 bc_acked = l->bc_rcvlink->acked;
740 bool bc_up = link_is_up(l->bc_rcvlink);
742 link_profile_stats(l);
745 case LINK_ESTABLISHED:
747 if (!l->silent_intv_cnt) {
748 if (bc_up && (bc_acked != bc_snt))
750 } else if (l->silent_intv_cnt <= l->abort_limit) {
754 rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
756 l->silent_intv_cnt++;
762 case LINK_ESTABLISHING:
766 case LINK_PEER_RESET:
768 case LINK_FAILINGOVER:
775 tipc_link_build_proto_msg(l, mtyp, prb, 0, 0, 0, xmitq);
781 * link_schedule_user - schedule a message sender for wakeup after congestion
782 * @link: congested link
783 * @list: message that was attempted sent
784 * Create pseudo msg to send back to user when congestion abates
785 * Does not consume buffer list
787 static int link_schedule_user(struct tipc_link *link, struct sk_buff_head *list)
789 struct tipc_msg *msg = buf_msg(skb_peek(list));
790 int imp = msg_importance(msg);
791 u32 oport = msg_origport(msg);
792 u32 addr = link_own_addr(link);
795 /* This really cannot happen... */
796 if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
797 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
800 /* Non-blocking sender: */
801 if (TIPC_SKB_CB(skb_peek(list))->wakeup_pending)
804 /* Create and schedule wakeup pseudo message */
805 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
806 addr, addr, oport, 0, 0);
809 TIPC_SKB_CB(skb)->chain_sz = skb_queue_len(list);
810 TIPC_SKB_CB(skb)->chain_imp = imp;
811 skb_queue_tail(&link->wakeupq, skb);
812 link->stats.link_congs++;
817 * link_prepare_wakeup - prepare users for wakeup after congestion
818 * @link: congested link
819 * Move a number of waiting users, as permitted by available space in
820 * the send queue, from link wait queue to node wait queue for wakeup
822 void link_prepare_wakeup(struct tipc_link *l)
824 int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
826 struct sk_buff *skb, *tmp;
828 skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
829 imp = TIPC_SKB_CB(skb)->chain_imp;
830 lim = l->window + l->backlog[imp].limit;
831 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
832 if ((pnd[imp] + l->backlog[imp].len) >= lim)
834 skb_unlink(skb, &l->wakeupq);
835 skb_queue_tail(l->inputq, skb);
839 void tipc_link_reset(struct tipc_link *l)
841 /* Link is down, accept any session */
842 l->peer_session = WILDCARD_SESSION;
844 /* If peer is up, it only accepts an incremented session number */
845 msg_set_session(l->pmsg, msg_session(l->pmsg) + 1);
847 /* Prepare for renewed mtu size negotiation */
848 l->mtu = l->advertised_mtu;
850 /* Clean up all queues and counters: */
851 __skb_queue_purge(&l->transmq);
852 __skb_queue_purge(&l->deferdq);
853 skb_queue_splice_init(&l->wakeupq, l->inputq);
854 __skb_queue_purge(&l->backlogq);
855 l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
856 l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
857 l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
858 l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
859 l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
860 kfree_skb(l->reasm_buf);
861 kfree_skb(l->failover_reasm_skb);
863 l->failover_reasm_skb = NULL;
868 l->silent_intv_cnt = 0;
869 l->stats.recv_info = 0;
871 l->bc_peer_is_up = false;
872 tipc_link_reset_stats(l);
876 * tipc_link_xmit(): enqueue buffer list according to queue situation
878 * @list: chain of buffers containing message
879 * @xmitq: returned list of packets to be sent by caller
881 * Consumes the buffer chain, except when returning -ELINKCONG,
882 * since the caller then may want to make more send attempts.
883 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
884 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
886 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
887 struct sk_buff_head *xmitq)
889 struct tipc_msg *hdr = buf_msg(skb_peek(list));
890 unsigned int maxwin = l->window;
891 unsigned int i, imp = msg_importance(hdr);
892 unsigned int mtu = l->mtu;
893 u16 ack = l->rcv_nxt - 1;
894 u16 seqno = l->snd_nxt;
895 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
896 struct sk_buff_head *transmq = &l->transmq;
897 struct sk_buff_head *backlogq = &l->backlogq;
898 struct sk_buff *skb, *_skb, *bskb;
900 /* Match msg importance against this and all higher backlog limits: */
901 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
902 if (unlikely(l->backlog[i].len >= l->backlog[i].limit))
903 return link_schedule_user(l, list);
905 if (unlikely(msg_size(hdr) > mtu))
908 /* Prepare each packet for sending, and add to relevant queue: */
909 while (skb_queue_len(list)) {
910 skb = skb_peek(list);
912 msg_set_seqno(hdr, seqno);
913 msg_set_ack(hdr, ack);
914 msg_set_bcast_ack(hdr, bc_ack);
916 if (likely(skb_queue_len(transmq) < maxwin)) {
917 _skb = skb_clone(skb, GFP_ATOMIC);
921 __skb_queue_tail(transmq, skb);
922 __skb_queue_tail(xmitq, _skb);
923 TIPC_SKB_CB(skb)->ackers = l->ackers;
928 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
929 kfree_skb(__skb_dequeue(list));
930 l->stats.sent_bundled++;
933 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
934 kfree_skb(__skb_dequeue(list));
935 __skb_queue_tail(backlogq, bskb);
936 l->backlog[msg_importance(buf_msg(bskb))].len++;
937 l->stats.sent_bundled++;
938 l->stats.sent_bundles++;
941 l->backlog[imp].len += skb_queue_len(list);
942 skb_queue_splice_tail_init(list, backlogq);
948 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
950 struct sk_buff *skb, *_skb;
951 struct tipc_msg *hdr;
952 u16 seqno = l->snd_nxt;
953 u16 ack = l->rcv_nxt - 1;
954 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
956 while (skb_queue_len(&l->transmq) < l->window) {
957 skb = skb_peek(&l->backlogq);
960 _skb = skb_clone(skb, GFP_ATOMIC);
963 __skb_dequeue(&l->backlogq);
965 l->backlog[msg_importance(hdr)].len--;
966 __skb_queue_tail(&l->transmq, skb);
967 __skb_queue_tail(xmitq, _skb);
968 TIPC_SKB_CB(skb)->ackers = l->ackers;
969 msg_set_seqno(hdr, seqno);
970 msg_set_ack(hdr, ack);
971 msg_set_bcast_ack(hdr, bc_ack);
978 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
980 struct tipc_msg *hdr = buf_msg(skb);
982 pr_warn("Retransmission failure on link <%s>\n", l->name);
983 link_print(l, "Resetting link ");
984 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
985 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
986 pr_info("sqno %u, prev: %x, src: %x\n",
987 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
990 int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
991 struct sk_buff_head *xmitq)
993 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
994 struct tipc_msg *hdr;
995 u16 ack = l->rcv_nxt - 1;
996 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1001 /* Detect repeated retransmit failures on same packet */
1002 if (likely(l->last_retransm != buf_seqno(skb))) {
1003 l->last_retransm = buf_seqno(skb);
1005 } else if (++l->stale_count > 100) {
1006 link_retransmit_failure(l, skb);
1007 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1010 /* Move forward to where retransmission should start */
1011 skb_queue_walk(&l->transmq, skb) {
1012 if (!less(buf_seqno(skb), from))
1016 skb_queue_walk_from(&l->transmq, skb) {
1017 if (more(buf_seqno(skb), to))
1020 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1023 hdr = buf_msg(_skb);
1024 msg_set_ack(hdr, ack);
1025 msg_set_bcast_ack(hdr, bc_ack);
1026 _skb->priority = TC_PRIO_CONTROL;
1027 __skb_queue_tail(xmitq, _skb);
1028 l->stats.retransmitted++;
1033 /* tipc_data_input - deliver data and name distr msgs to upper layer
1035 * Consumes buffer if message is of right type
1036 * Node lock must be held
1038 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1039 struct sk_buff_head *inputq)
1041 switch (msg_user(buf_msg(skb))) {
1042 case TIPC_LOW_IMPORTANCE:
1043 case TIPC_MEDIUM_IMPORTANCE:
1044 case TIPC_HIGH_IMPORTANCE:
1045 case TIPC_CRITICAL_IMPORTANCE:
1047 skb_queue_tail(inputq, skb);
1049 case NAME_DISTRIBUTOR:
1050 l->bc_rcvlink->state = LINK_ESTABLISHED;
1051 skb_queue_tail(l->namedq, skb);
1054 case TUNNEL_PROTOCOL:
1055 case MSG_FRAGMENTER:
1056 case BCAST_PROTOCOL:
1059 pr_warn("Dropping received illegal msg type\n");
1065 /* tipc_link_input - process packet that has passed link protocol check
1069 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1070 struct sk_buff_head *inputq)
1072 struct tipc_msg *hdr = buf_msg(skb);
1073 struct sk_buff **reasm_skb = &l->reasm_buf;
1074 struct sk_buff *iskb;
1075 struct sk_buff_head tmpq;
1076 int usr = msg_user(hdr);
1081 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1082 if (msg_type(hdr) == SYNCH_MSG) {
1083 __skb_queue_purge(&l->deferdq);
1086 if (!tipc_msg_extract(skb, &iskb, &ipos))
1091 if (less(msg_seqno(hdr), l->drop_point))
1093 if (tipc_data_input(l, skb, inputq))
1095 usr = msg_user(hdr);
1096 reasm_skb = &l->failover_reasm_skb;
1099 if (usr == MSG_BUNDLER) {
1100 skb_queue_head_init(&tmpq);
1101 l->stats.recv_bundles++;
1102 l->stats.recv_bundled += msg_msgcnt(hdr);
1103 while (tipc_msg_extract(skb, &iskb, &pos))
1104 tipc_data_input(l, iskb, &tmpq);
1105 tipc_skb_queue_splice_tail(&tmpq, inputq);
1107 } else if (usr == MSG_FRAGMENTER) {
1108 l->stats.recv_fragments++;
1109 if (tipc_buf_append(reasm_skb, &skb)) {
1110 l->stats.recv_fragmented++;
1111 tipc_data_input(l, skb, inputq);
1112 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1113 pr_warn_ratelimited("Unable to build fragment list\n");
1114 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1117 } else if (usr == BCAST_PROTOCOL) {
1118 tipc_bcast_lock(l->net);
1119 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1120 tipc_bcast_unlock(l->net);
1127 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1129 bool released = false;
1130 struct sk_buff *skb, *tmp;
1132 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1133 if (more(buf_seqno(skb), acked))
1135 __skb_unlink(skb, &l->transmq);
1142 /* tipc_link_build_ack_msg: prepare link acknowledge message for transmission
1144 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1145 * risk of ack storms towards the sender
1147 int tipc_link_build_ack_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1152 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1153 if (link_is_bc_rcvlink(l)) {
1154 if (((l->rcv_nxt ^ link_own_addr(l)) & 0xf) != 0xf)
1157 return TIPC_LINK_SND_BC_ACK;
1162 l->stats.sent_acks++;
1163 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1167 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1169 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1171 int mtyp = RESET_MSG;
1173 if (l->state == LINK_ESTABLISHING)
1174 mtyp = ACTIVATE_MSG;
1176 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1179 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1181 static void tipc_link_build_nack_msg(struct tipc_link *l,
1182 struct sk_buff_head *xmitq)
1184 u32 def_cnt = ++l->stats.deferred_recv;
1186 if (link_is_bc_rcvlink(l))
1189 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1190 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1193 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1194 * @l: the link that should handle the message
1196 * @xmitq: queue to place packets to be sent after this call
1198 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1199 struct sk_buff_head *xmitq)
1201 struct sk_buff_head *defq = &l->deferdq;
1202 struct tipc_msg *hdr;
1203 u16 seqno, rcv_nxt, win_lim;
1208 seqno = msg_seqno(hdr);
1209 rcv_nxt = l->rcv_nxt;
1210 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1212 /* Verify and update link state */
1213 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1214 return tipc_link_proto_rcv(l, skb, xmitq);
1216 if (unlikely(!link_is_up(l))) {
1217 if (l->state == LINK_ESTABLISHING)
1218 rc = TIPC_LINK_UP_EVT;
1222 /* Don't send probe at next timeout expiration */
1223 l->silent_intv_cnt = 0;
1225 /* Drop if outside receive window */
1226 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1227 l->stats.duplicates++;
1231 /* Forward queues and wake up waiting users */
1232 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1233 tipc_link_advance_backlog(l, xmitq);
1234 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1235 link_prepare_wakeup(l);
1238 /* Defer delivery if sequence gap */
1239 if (unlikely(seqno != rcv_nxt)) {
1240 __tipc_skb_queue_sorted(defq, seqno, skb);
1241 tipc_link_build_nack_msg(l, xmitq);
1245 /* Deliver packet */
1247 l->stats.recv_info++;
1248 if (!tipc_data_input(l, skb, l->inputq))
1249 rc |= tipc_link_input(l, skb, l->inputq);
1250 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1251 rc |= tipc_link_build_ack_msg(l, xmitq);
1252 if (unlikely(rc & ~TIPC_LINK_SND_BC_ACK))
1254 } while ((skb = __skb_dequeue(defq)));
1263 * Send protocol message to the other endpoint.
1265 static void tipc_link_proto_xmit(struct tipc_link *l, u32 msg_typ,
1266 int probe_msg, u32 gap, u32 tolerance,
1269 struct sk_buff *skb = NULL;
1270 struct sk_buff_head xmitq;
1272 __skb_queue_head_init(&xmitq);
1273 tipc_link_build_proto_msg(l, msg_typ, probe_msg, gap,
1274 tolerance, priority, &xmitq);
1275 skb = __skb_dequeue(&xmitq);
1278 tipc_bearer_xmit_skb(l->net, l->bearer_id, skb, l->media_addr);
1282 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1283 u16 rcvgap, int tolerance, int priority,
1284 struct sk_buff_head *xmitq)
1286 struct sk_buff *skb = NULL;
1287 struct tipc_msg *hdr = l->pmsg;
1288 bool node_up = link_is_up(l->bc_rcvlink);
1290 /* Don't send protocol message during reset or link failover */
1291 if (tipc_link_is_blocked(l))
1294 msg_set_type(hdr, mtyp);
1295 msg_set_net_plane(hdr, l->net_plane);
1296 msg_set_next_sent(hdr, l->snd_nxt);
1297 msg_set_ack(hdr, l->rcv_nxt - 1);
1298 msg_set_bcast_ack(hdr, l->bc_rcvlink->rcv_nxt - 1);
1299 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1300 msg_set_link_tolerance(hdr, tolerance);
1301 msg_set_linkprio(hdr, priority);
1302 msg_set_redundant_link(hdr, node_up);
1303 msg_set_seq_gap(hdr, 0);
1305 /* Compatibility: created msg must not be in sequence with pkt flow */
1306 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1308 if (mtyp == STATE_MSG) {
1309 if (!tipc_link_is_up(l))
1312 /* Override rcvgap if there are packets in deferred queue */
1313 if (!skb_queue_empty(&l->deferdq))
1314 rcvgap = buf_seqno(skb_peek(&l->deferdq)) - l->rcv_nxt;
1316 msg_set_seq_gap(hdr, rcvgap);
1317 l->stats.sent_nacks++;
1319 msg_set_probe(hdr, probe);
1321 l->stats.sent_probes++;
1322 l->stats.sent_states++;
1325 /* RESET_MSG or ACTIVATE_MSG */
1326 msg_set_max_pkt(hdr, l->advertised_mtu);
1327 msg_set_ack(hdr, l->rcv_nxt - 1);
1328 msg_set_next_sent(hdr, 1);
1330 skb = tipc_buf_acquire(msg_size(hdr));
1333 skb_copy_to_linear_data(skb, hdr, msg_size(hdr));
1334 skb->priority = TC_PRIO_CONTROL;
1335 __skb_queue_tail(xmitq, skb);
1338 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1339 * with contents of the link's transmit and backlog queues.
1341 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1342 int mtyp, struct sk_buff_head *xmitq)
1344 struct sk_buff *skb, *tnlskb;
1345 struct tipc_msg *hdr, tnlhdr;
1346 struct sk_buff_head *queue = &l->transmq;
1347 struct sk_buff_head tmpxq, tnlq;
1348 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1353 skb_queue_head_init(&tnlq);
1354 skb_queue_head_init(&tmpxq);
1356 /* At least one packet required for safe algorithm => add dummy */
1357 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1358 BASIC_H_SIZE, 0, l->addr, link_own_addr(l),
1359 0, 0, TIPC_ERR_NO_PORT);
1361 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1364 skb_queue_tail(&tnlq, skb);
1365 tipc_link_xmit(l, &tnlq, &tmpxq);
1366 __skb_queue_purge(&tmpxq);
1368 /* Initialize reusable tunnel packet header */
1369 tipc_msg_init(link_own_addr(l), &tnlhdr, TUNNEL_PROTOCOL,
1370 mtyp, INT_H_SIZE, l->addr);
1371 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1372 msg_set_msgcnt(&tnlhdr, pktcnt);
1373 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1375 /* Wrap each packet into a tunnel packet */
1376 skb_queue_walk(queue, skb) {
1378 if (queue == &l->backlogq)
1379 msg_set_seqno(hdr, seqno++);
1380 pktlen = msg_size(hdr);
1381 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1382 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE);
1384 pr_warn("%sunable to send packet\n", link_co_err);
1387 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1388 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1389 __skb_queue_tail(&tnlq, tnlskb);
1391 if (queue != &l->backlogq) {
1392 queue = &l->backlogq;
1396 tipc_link_xmit(tnl, &tnlq, xmitq);
1398 if (mtyp == FAILOVER_MSG) {
1399 tnl->drop_point = l->rcv_nxt;
1400 tnl->failover_reasm_skb = l->reasm_buf;
1401 l->reasm_buf = NULL;
1405 /* tipc_link_proto_rcv(): receive link level protocol message :
1406 * Note that network plane id propagates through the network, and may
1407 * change at any time. The node with lowest numerical id determines
1410 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1411 struct sk_buff_head *xmitq)
1413 struct tipc_msg *hdr = buf_msg(skb);
1415 u16 ack = msg_ack(hdr);
1416 u16 gap = msg_seq_gap(hdr);
1417 u16 peers_snd_nxt = msg_next_sent(hdr);
1418 u16 peers_tol = msg_link_tolerance(hdr);
1419 u16 peers_prio = msg_linkprio(hdr);
1420 u16 rcv_nxt = l->rcv_nxt;
1421 int mtyp = msg_type(hdr);
1425 if (tipc_link_is_blocked(l) || !xmitq)
1428 if (link_own_addr(l) > msg_prevnode(hdr))
1429 l->net_plane = msg_net_plane(hdr);
1434 /* Ignore duplicate RESET with old session number */
1435 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1436 (l->peer_session != WILDCARD_SESSION))
1444 /* Complete own link name with peer's interface name */
1445 if_name = strrchr(l->name, ':') + 1;
1446 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1448 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1450 strncpy(if_name, msg_data(hdr), TIPC_MAX_IF_NAME);
1452 /* Update own tolerance if peer indicates a non-zero value */
1453 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1454 l->tolerance = peers_tol;
1456 /* Update own priority if peer's priority is higher */
1457 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1458 l->priority = peers_prio;
1460 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1461 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 l->silent_intv_cnt = 0;
1481 l->stats.recv_states++;
1483 l->stats.recv_probes++;
1485 if (!link_is_up(l)) {
1486 if (l->state == LINK_ESTABLISHING)
1487 rc = TIPC_LINK_UP_EVT;
1491 /* Send NACK if peer has sent pkts we haven't received yet */
1492 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1493 rcvgap = peers_snd_nxt - l->rcv_nxt;
1494 if (rcvgap || (msg_probe(hdr)))
1495 tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1497 tipc_link_release_pkts(l, ack);
1499 /* If NACK, retransmit will now start at right position */
1501 rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
1502 l->stats.recv_nacks++;
1505 tipc_link_advance_backlog(l, xmitq);
1506 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1507 link_prepare_wakeup(l);
1514 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1516 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1518 struct sk_buff_head *xmitq)
1520 struct sk_buff *skb;
1521 struct tipc_msg *hdr;
1522 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1523 u16 ack = l->rcv_nxt - 1;
1524 u16 gap_to = peers_snd_nxt - 1;
1526 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1527 0, l->addr, link_own_addr(l), 0, 0, 0);
1531 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1532 msg_set_bcast_ack(hdr, ack);
1533 msg_set_bcgap_after(hdr, ack);
1535 gap_to = buf_seqno(dfrd_skb) - 1;
1536 msg_set_bcgap_to(hdr, gap_to);
1537 msg_set_non_seq(hdr, bcast);
1538 __skb_queue_tail(xmitq, skb);
1542 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1544 * Give a newly added peer node the sequence number where it should
1545 * start receiving and acking broadcast packets.
1547 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1548 struct sk_buff_head *xmitq)
1550 struct sk_buff_head list;
1552 __skb_queue_head_init(&list);
1553 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1555 tipc_link_xmit(l, &list, xmitq);
1558 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1560 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1562 int mtyp = msg_type(hdr);
1563 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1568 if (msg_user(hdr) == BCAST_PROTOCOL) {
1569 l->rcv_nxt = peers_snd_nxt;
1570 l->state = LINK_ESTABLISHED;
1574 if (l->peer_caps & TIPC_BCAST_SYNCH)
1577 if (msg_peer_node_is_up(hdr))
1580 /* Compatibility: accept older, less safe initial synch data */
1581 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1582 l->rcv_nxt = peers_snd_nxt;
1585 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1587 void tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1588 struct sk_buff_head *xmitq)
1590 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1595 if (!msg_peer_node_is_up(hdr))
1598 l->bc_peer_is_up = true;
1600 /* Ignore if peers_snd_nxt goes beyond receive window */
1601 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1604 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1605 l->nack_state = BC_NACK_SND_CONDITIONAL;
1609 /* Don't NACK if one was recently sent or peeked */
1610 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1611 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1615 /* Conditionally delay NACK sending until next synch rcv */
1616 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1617 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1618 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1622 /* Send NACK now but suppress next one */
1623 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1624 l->nack_state = BC_NACK_SND_SUPPRESS;
1627 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1628 struct sk_buff_head *xmitq)
1630 struct sk_buff *skb, *tmp;
1631 struct tipc_link *snd_l = l->bc_sndlink;
1633 if (!link_is_up(l) || !l->bc_peer_is_up)
1636 if (!more(acked, l->acked))
1639 /* Skip over packets peer has already acked */
1640 skb_queue_walk(&snd_l->transmq, skb) {
1641 if (more(buf_seqno(skb), l->acked))
1645 /* Update/release the packets peer is acking now */
1646 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1647 if (more(buf_seqno(skb), acked))
1649 if (!--TIPC_SKB_CB(skb)->ackers) {
1650 __skb_unlink(skb, &snd_l->transmq);
1655 tipc_link_advance_backlog(snd_l, xmitq);
1656 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1657 link_prepare_wakeup(snd_l);
1660 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1662 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1663 struct sk_buff_head *xmitq)
1665 struct tipc_msg *hdr = buf_msg(skb);
1666 u32 dnode = msg_destnode(hdr);
1667 int mtyp = msg_type(hdr);
1668 u16 acked = msg_bcast_ack(hdr);
1669 u16 from = acked + 1;
1670 u16 to = msg_bcgap_to(hdr);
1671 u16 peers_snd_nxt = to + 1;
1676 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1679 if (mtyp != STATE_MSG)
1682 if (dnode == link_own_addr(l)) {
1683 tipc_link_bc_ack_rcv(l, acked, xmitq);
1684 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1685 l->stats.recv_nacks++;
1689 /* Msg for other node => suppress own NACK at next sync if applicable */
1690 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1691 l->nack_state = BC_NACK_SND_SUPPRESS;
1696 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1698 int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1701 l->backlog[TIPC_LOW_IMPORTANCE].limit = win / 2;
1702 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = win;
1703 l->backlog[TIPC_HIGH_IMPORTANCE].limit = win / 2 * 3;
1704 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = win * 2;
1705 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1709 * link_reset_stats - reset link statistics
1710 * @l_ptr: pointer to link
1712 void tipc_link_reset_stats(struct tipc_link *l)
1714 memset(&l->stats, 0, sizeof(l->stats));
1715 if (!link_is_bc_sndlink(l)) {
1716 l->stats.sent_info = l->snd_nxt;
1717 l->stats.recv_info = l->rcv_nxt;
1721 static void link_print(struct tipc_link *l, const char *str)
1723 struct sk_buff *hskb = skb_peek(&l->transmq);
1724 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1725 u16 tail = l->snd_nxt - 1;
1727 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1728 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1729 skb_queue_len(&l->transmq), head, tail,
1730 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1733 /* Parse and validate nested (link) properties valid for media, bearer and link
1735 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1739 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1740 tipc_nl_prop_policy);
1744 if (props[TIPC_NLA_PROP_PRIO]) {
1747 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1748 if (prio > TIPC_MAX_LINK_PRI)
1752 if (props[TIPC_NLA_PROP_TOL]) {
1755 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1756 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1760 if (props[TIPC_NLA_PROP_WIN]) {
1763 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1764 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1771 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1774 struct nlattr *stats;
1781 struct nla_map map[] = {
1782 {TIPC_NLA_STATS_RX_INFO, s->recv_info},
1783 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1784 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1785 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1786 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1787 {TIPC_NLA_STATS_TX_INFO, s->sent_info},
1788 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1789 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1790 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1791 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1792 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1793 s->msg_length_counts : 1},
1794 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1795 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1796 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1797 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1798 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1799 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1800 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1801 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1802 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1803 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1804 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1805 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1806 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1807 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1808 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1809 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1810 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1811 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1812 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1813 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1814 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1815 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1816 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1819 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1823 for (i = 0; i < ARRAY_SIZE(map); i++)
1824 if (nla_put_u32(skb, map[i].key, map[i].val))
1827 nla_nest_end(skb, stats);
1831 nla_nest_cancel(skb, stats);
1836 /* Caller should hold appropriate locks to protect the link */
1837 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1838 struct tipc_link *link, int nlflags)
1842 struct nlattr *attrs;
1843 struct nlattr *prop;
1844 struct tipc_net *tn = net_generic(net, tipc_net_id);
1846 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1847 nlflags, TIPC_NL_LINK_GET);
1851 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1855 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1857 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1858 tipc_cluster_mask(tn->own_addr)))
1860 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1862 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->rcv_nxt))
1864 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->snd_nxt))
1867 if (tipc_link_is_up(link))
1868 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1871 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
1874 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1877 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1879 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
1881 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
1884 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1886 nla_nest_end(msg->skb, prop);
1888 err = __tipc_nl_add_stats(msg->skb, &link->stats);
1892 nla_nest_end(msg->skb, attrs);
1893 genlmsg_end(msg->skb, hdr);
1898 nla_nest_cancel(msg->skb, prop);
1900 nla_nest_cancel(msg->skb, attrs);
1902 genlmsg_cancel(msg->skb, hdr);
1907 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
1908 struct tipc_stats *stats)
1911 struct nlattr *nest;
1918 struct nla_map map[] = {
1919 {TIPC_NLA_STATS_RX_INFO, stats->recv_info},
1920 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
1921 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
1922 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
1923 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
1924 {TIPC_NLA_STATS_TX_INFO, stats->sent_info},
1925 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
1926 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
1927 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
1928 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
1929 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
1930 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
1931 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
1932 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
1933 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
1934 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
1935 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
1936 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
1937 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
1938 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
1941 nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1945 for (i = 0; i < ARRAY_SIZE(map); i++)
1946 if (nla_put_u32(skb, map[i].key, map[i].val))
1949 nla_nest_end(skb, nest);
1953 nla_nest_cancel(skb, nest);
1958 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
1962 struct nlattr *attrs;
1963 struct nlattr *prop;
1964 struct tipc_net *tn = net_generic(net, tipc_net_id);
1965 struct tipc_link *bcl = tn->bcl;
1970 tipc_bcast_lock(net);
1972 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1973 NLM_F_MULTI, TIPC_NL_LINK_GET);
1977 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1981 /* The broadcast link is always up */
1982 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1985 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
1987 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
1989 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, bcl->rcv_nxt))
1991 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, bcl->snd_nxt))
1994 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1997 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
1999 nla_nest_end(msg->skb, prop);
2001 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2005 tipc_bcast_unlock(net);
2006 nla_nest_end(msg->skb, attrs);
2007 genlmsg_end(msg->skb, hdr);
2012 nla_nest_cancel(msg->skb, prop);
2014 nla_nest_cancel(msg->skb, attrs);
2016 tipc_bcast_unlock(net);
2017 genlmsg_cancel(msg->skb, hdr);
2022 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol)
2025 tipc_link_proto_xmit(l, STATE_MSG, 0, 0, tol, 0);
2028 void tipc_link_set_prio(struct tipc_link *l, u32 prio)
2031 tipc_link_proto_xmit(l, STATE_MSG, 0, 0, 0, prio);
2034 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2036 l->abort_limit = limit;