xref: /openbmc/linux/net/tipc/link.c (revision f220d3eb)
1 /*
2  * net/tipc/link.c: TIPC link code
3  *
4  * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5  * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
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.
19  *
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.
23  *
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.
35  */
36 
37 #include "core.h"
38 #include "subscr.h"
39 #include "link.h"
40 #include "bcast.h"
41 #include "socket.h"
42 #include "name_distr.h"
43 #include "discover.h"
44 #include "netlink.h"
45 #include "monitor.h"
46 
47 #include <linux/pkt_sched.h>
48 
49 struct tipc_stats {
50 	u32 sent_pkts;
51 	u32 recv_pkts;
52 	u32 sent_states;
53 	u32 recv_states;
54 	u32 sent_probes;
55 	u32 recv_probes;
56 	u32 sent_nacks;
57 	u32 recv_nacks;
58 	u32 sent_acks;
59 	u32 sent_bundled;
60 	u32 sent_bundles;
61 	u32 recv_bundled;
62 	u32 recv_bundles;
63 	u32 retransmitted;
64 	u32 sent_fragmented;
65 	u32 sent_fragments;
66 	u32 recv_fragmented;
67 	u32 recv_fragments;
68 	u32 link_congs;		/* # port sends blocked by congestion */
69 	u32 deferred_recv;
70 	u32 duplicates;
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 */
77 };
78 
79 /**
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
84  * @timer: link timer
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_cnt: counter for number of identical retransmit attempts
110  * @stale_limit: time when repeated identical retransmits must force link reset
111  * @ackers: # of peers that needs to ack each packet before it can be released
112  * @acked: # last packet acked by a certain peer. Used for broadcast.
113  * @rcv_nxt: next sequence number to expect for inbound messages
114  * @deferred_queue: deferred queue saved OOS b'cast message received from node
115  * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
116  * @inputq: buffer queue for messages to be delivered upwards
117  * @namedq: buffer queue for name table messages to be delivered upwards
118  * @next_out: ptr to first unsent outbound message in queue
119  * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
120  * @long_msg_seq_no: next identifier to use for outbound fragmented messages
121  * @reasm_buf: head of partially reassembled inbound message fragments
122  * @bc_rcvr: marks that this is a broadcast receiver link
123  * @stats: collects statistics regarding link activity
124  */
125 struct tipc_link {
126 	u32 addr;
127 	char name[TIPC_MAX_LINK_NAME];
128 	struct net *net;
129 
130 	/* Management and link supervision data */
131 	u16 peer_session;
132 	u16 session;
133 	u16 snd_nxt_state;
134 	u16 rcv_nxt_state;
135 	u32 peer_bearer_id;
136 	u32 bearer_id;
137 	u32 tolerance;
138 	u32 abort_limit;
139 	u32 state;
140 	u16 peer_caps;
141 	bool in_session;
142 	bool active;
143 	u32 silent_intv_cnt;
144 	char if_name[TIPC_MAX_IF_NAME];
145 	u32 priority;
146 	char net_plane;
147 	struct tipc_mon_state mon_state;
148 	u16 rst_cnt;
149 
150 	/* Failover/synch */
151 	u16 drop_point;
152 	struct sk_buff *failover_reasm_skb;
153 
154 	/* Max packet negotiation */
155 	u16 mtu;
156 	u16 advertised_mtu;
157 
158 	/* Sending */
159 	struct sk_buff_head transmq;
160 	struct sk_buff_head backlogq;
161 	struct {
162 		u16 len;
163 		u16 limit;
164 	} backlog[5];
165 	u16 snd_nxt;
166 	u16 last_retransm;
167 	u16 window;
168 	u16 stale_cnt;
169 	unsigned long stale_limit;
170 
171 	/* Reception */
172 	u16 rcv_nxt;
173 	u32 rcv_unacked;
174 	struct sk_buff_head deferdq;
175 	struct sk_buff_head *inputq;
176 	struct sk_buff_head *namedq;
177 
178 	/* Congestion handling */
179 	struct sk_buff_head wakeupq;
180 
181 	/* Fragmentation/reassembly */
182 	struct sk_buff *reasm_buf;
183 
184 	/* Broadcast */
185 	u16 ackers;
186 	u16 acked;
187 	struct tipc_link *bc_rcvlink;
188 	struct tipc_link *bc_sndlink;
189 	unsigned long prev_retr;
190 	u16 prev_from;
191 	u16 prev_to;
192 	u8 nack_state;
193 	bool bc_peer_is_up;
194 
195 	/* Statistics */
196 	struct tipc_stats stats;
197 };
198 
199 /*
200  * Error message prefixes
201  */
202 static const char *link_co_err = "Link tunneling error, ";
203 static const char *link_rst_msg = "Resetting link ";
204 
205 /* Send states for broadcast NACKs
206  */
207 enum {
208 	BC_NACK_SND_CONDITIONAL,
209 	BC_NACK_SND_UNCONDITIONAL,
210 	BC_NACK_SND_SUPPRESS,
211 };
212 
213 #define TIPC_BC_RETR_LIMIT 10   /* [ms] */
214 
215 /*
216  * Interval between NACKs when packets arrive out of order
217  */
218 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
219 
220 /* Link FSM states:
221  */
222 enum {
223 	LINK_ESTABLISHED     = 0xe,
224 	LINK_ESTABLISHING    = 0xe  << 4,
225 	LINK_RESET           = 0x1  << 8,
226 	LINK_RESETTING       = 0x2  << 12,
227 	LINK_PEER_RESET      = 0xd  << 16,
228 	LINK_FAILINGOVER     = 0xf  << 20,
229 	LINK_SYNCHING        = 0xc  << 24
230 };
231 
232 /* Link FSM state checking routines
233  */
234 static int link_is_up(struct tipc_link *l)
235 {
236 	return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
237 }
238 
239 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
240 			       struct sk_buff_head *xmitq);
241 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
242 				      bool probe_reply, u16 rcvgap,
243 				      int tolerance, int priority,
244 				      struct sk_buff_head *xmitq);
245 static void link_print(struct tipc_link *l, const char *str);
246 static int tipc_link_build_nack_msg(struct tipc_link *l,
247 				    struct sk_buff_head *xmitq);
248 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
249 					struct sk_buff_head *xmitq);
250 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
251 
252 /*
253  *  Simple non-static link routines (i.e. referenced outside this file)
254  */
255 bool tipc_link_is_up(struct tipc_link *l)
256 {
257 	return link_is_up(l);
258 }
259 
260 bool tipc_link_peer_is_down(struct tipc_link *l)
261 {
262 	return l->state == LINK_PEER_RESET;
263 }
264 
265 bool tipc_link_is_reset(struct tipc_link *l)
266 {
267 	return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
268 }
269 
270 bool tipc_link_is_establishing(struct tipc_link *l)
271 {
272 	return l->state == LINK_ESTABLISHING;
273 }
274 
275 bool tipc_link_is_synching(struct tipc_link *l)
276 {
277 	return l->state == LINK_SYNCHING;
278 }
279 
280 bool tipc_link_is_failingover(struct tipc_link *l)
281 {
282 	return l->state == LINK_FAILINGOVER;
283 }
284 
285 bool tipc_link_is_blocked(struct tipc_link *l)
286 {
287 	return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
288 }
289 
290 static bool link_is_bc_sndlink(struct tipc_link *l)
291 {
292 	return !l->bc_sndlink;
293 }
294 
295 static bool link_is_bc_rcvlink(struct tipc_link *l)
296 {
297 	return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
298 }
299 
300 void tipc_link_set_active(struct tipc_link *l, bool active)
301 {
302 	l->active = active;
303 }
304 
305 u32 tipc_link_id(struct tipc_link *l)
306 {
307 	return l->peer_bearer_id << 16 | l->bearer_id;
308 }
309 
310 int tipc_link_window(struct tipc_link *l)
311 {
312 	return l->window;
313 }
314 
315 int tipc_link_prio(struct tipc_link *l)
316 {
317 	return l->priority;
318 }
319 
320 unsigned long tipc_link_tolerance(struct tipc_link *l)
321 {
322 	return l->tolerance;
323 }
324 
325 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
326 {
327 	return l->inputq;
328 }
329 
330 char tipc_link_plane(struct tipc_link *l)
331 {
332 	return l->net_plane;
333 }
334 
335 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
336 {
337 	l->peer_caps = capabilities;
338 }
339 
340 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
341 			   struct tipc_link *uc_l,
342 			   struct sk_buff_head *xmitq)
343 {
344 	struct tipc_link *rcv_l = uc_l->bc_rcvlink;
345 
346 	snd_l->ackers++;
347 	rcv_l->acked = snd_l->snd_nxt - 1;
348 	snd_l->state = LINK_ESTABLISHED;
349 	tipc_link_build_bc_init_msg(uc_l, xmitq);
350 }
351 
352 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
353 			      struct tipc_link *rcv_l,
354 			      struct sk_buff_head *xmitq)
355 {
356 	u16 ack = snd_l->snd_nxt - 1;
357 
358 	snd_l->ackers--;
359 	rcv_l->bc_peer_is_up = true;
360 	rcv_l->state = LINK_ESTABLISHED;
361 	tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
362 	tipc_link_reset(rcv_l);
363 	rcv_l->state = LINK_RESET;
364 	if (!snd_l->ackers) {
365 		tipc_link_reset(snd_l);
366 		snd_l->state = LINK_RESET;
367 		__skb_queue_purge(xmitq);
368 	}
369 }
370 
371 int tipc_link_bc_peers(struct tipc_link *l)
372 {
373 	return l->ackers;
374 }
375 
376 static u16 link_bc_rcv_gap(struct tipc_link *l)
377 {
378 	struct sk_buff *skb = skb_peek(&l->deferdq);
379 	u16 gap = 0;
380 
381 	if (more(l->snd_nxt, l->rcv_nxt))
382 		gap = l->snd_nxt - l->rcv_nxt;
383 	if (skb)
384 		gap = buf_seqno(skb) - l->rcv_nxt;
385 	return gap;
386 }
387 
388 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
389 {
390 	l->mtu = mtu;
391 }
392 
393 int tipc_link_mtu(struct tipc_link *l)
394 {
395 	return l->mtu;
396 }
397 
398 u16 tipc_link_rcv_nxt(struct tipc_link *l)
399 {
400 	return l->rcv_nxt;
401 }
402 
403 u16 tipc_link_acked(struct tipc_link *l)
404 {
405 	return l->acked;
406 }
407 
408 char *tipc_link_name(struct tipc_link *l)
409 {
410 	return l->name;
411 }
412 
413 /**
414  * tipc_link_create - create a new link
415  * @n: pointer to associated node
416  * @if_name: associated interface name
417  * @bearer_id: id (index) of associated bearer
418  * @tolerance: link tolerance to be used by link
419  * @net_plane: network plane (A,B,c..) this link belongs to
420  * @mtu: mtu to be advertised by link
421  * @priority: priority to be used by link
422  * @window: send window to be used by link
423  * @session: session to be used by link
424  * @ownnode: identity of own node
425  * @peer: node id of peer node
426  * @peer_caps: bitmap describing peer node capabilities
427  * @bc_sndlink: the namespace global link used for broadcast sending
428  * @bc_rcvlink: the peer specific link used for broadcast reception
429  * @inputq: queue to put messages ready for delivery
430  * @namedq: queue to put binding table update messages ready for delivery
431  * @link: return value, pointer to put the created link
432  *
433  * Returns true if link was created, otherwise false
434  */
435 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
436 		      int tolerance, char net_plane, u32 mtu, int priority,
437 		      int window, u32 session, u32 self,
438 		      u32 peer, u8 *peer_id, u16 peer_caps,
439 		      struct tipc_link *bc_sndlink,
440 		      struct tipc_link *bc_rcvlink,
441 		      struct sk_buff_head *inputq,
442 		      struct sk_buff_head *namedq,
443 		      struct tipc_link **link)
444 {
445 	char peer_str[NODE_ID_STR_LEN] = {0,};
446 	char self_str[NODE_ID_STR_LEN] = {0,};
447 	struct tipc_link *l;
448 
449 	l = kzalloc(sizeof(*l), GFP_ATOMIC);
450 	if (!l)
451 		return false;
452 	*link = l;
453 	l->session = session;
454 
455 	/* Set link name for unicast links only */
456 	if (peer_id) {
457 		tipc_nodeid2string(self_str, tipc_own_id(net));
458 		if (strlen(self_str) > 16)
459 			sprintf(self_str, "%x", self);
460 		tipc_nodeid2string(peer_str, peer_id);
461 		if (strlen(peer_str) > 16)
462 			sprintf(peer_str, "%x", peer);
463 	}
464 	/* Peer i/f name will be completed by reset/activate message */
465 	snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
466 		 self_str, if_name, peer_str);
467 
468 	strcpy(l->if_name, if_name);
469 	l->addr = peer;
470 	l->peer_caps = peer_caps;
471 	l->net = net;
472 	l->in_session = false;
473 	l->bearer_id = bearer_id;
474 	l->tolerance = tolerance;
475 	l->net_plane = net_plane;
476 	l->advertised_mtu = mtu;
477 	l->mtu = mtu;
478 	l->priority = priority;
479 	tipc_link_set_queue_limits(l, window);
480 	l->ackers = 1;
481 	l->bc_sndlink = bc_sndlink;
482 	l->bc_rcvlink = bc_rcvlink;
483 	l->inputq = inputq;
484 	l->namedq = namedq;
485 	l->state = LINK_RESETTING;
486 	__skb_queue_head_init(&l->transmq);
487 	__skb_queue_head_init(&l->backlogq);
488 	__skb_queue_head_init(&l->deferdq);
489 	skb_queue_head_init(&l->wakeupq);
490 	skb_queue_head_init(l->inputq);
491 	return true;
492 }
493 
494 /**
495  * tipc_link_bc_create - create new link to be used for broadcast
496  * @n: pointer to associated node
497  * @mtu: mtu to be used initially if no peers
498  * @window: send window to be used
499  * @inputq: queue to put messages ready for delivery
500  * @namedq: queue to put binding table update messages ready for delivery
501  * @link: return value, pointer to put the created link
502  *
503  * Returns true if link was created, otherwise false
504  */
505 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
506 			 int mtu, int window, u16 peer_caps,
507 			 struct sk_buff_head *inputq,
508 			 struct sk_buff_head *namedq,
509 			 struct tipc_link *bc_sndlink,
510 			 struct tipc_link **link)
511 {
512 	struct tipc_link *l;
513 
514 	if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
515 			      0, ownnode, peer, NULL, peer_caps, bc_sndlink,
516 			      NULL, inputq, namedq, link))
517 		return false;
518 
519 	l = *link;
520 	strcpy(l->name, tipc_bclink_name);
521 	tipc_link_reset(l);
522 	l->state = LINK_RESET;
523 	l->ackers = 0;
524 	l->bc_rcvlink = l;
525 
526 	/* Broadcast send link is always up */
527 	if (link_is_bc_sndlink(l))
528 		l->state = LINK_ESTABLISHED;
529 
530 	/* Disable replicast if even a single peer doesn't support it */
531 	if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
532 		tipc_bcast_disable_rcast(net);
533 
534 	return true;
535 }
536 
537 /**
538  * tipc_link_fsm_evt - link finite state machine
539  * @l: pointer to link
540  * @evt: state machine event to be processed
541  */
542 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
543 {
544 	int rc = 0;
545 
546 	switch (l->state) {
547 	case LINK_RESETTING:
548 		switch (evt) {
549 		case LINK_PEER_RESET_EVT:
550 			l->state = LINK_PEER_RESET;
551 			break;
552 		case LINK_RESET_EVT:
553 			l->state = LINK_RESET;
554 			break;
555 		case LINK_FAILURE_EVT:
556 		case LINK_FAILOVER_BEGIN_EVT:
557 		case LINK_ESTABLISH_EVT:
558 		case LINK_FAILOVER_END_EVT:
559 		case LINK_SYNCH_BEGIN_EVT:
560 		case LINK_SYNCH_END_EVT:
561 		default:
562 			goto illegal_evt;
563 		}
564 		break;
565 	case LINK_RESET:
566 		switch (evt) {
567 		case LINK_PEER_RESET_EVT:
568 			l->state = LINK_ESTABLISHING;
569 			break;
570 		case LINK_FAILOVER_BEGIN_EVT:
571 			l->state = LINK_FAILINGOVER;
572 		case LINK_FAILURE_EVT:
573 		case LINK_RESET_EVT:
574 		case LINK_ESTABLISH_EVT:
575 		case LINK_FAILOVER_END_EVT:
576 			break;
577 		case LINK_SYNCH_BEGIN_EVT:
578 		case LINK_SYNCH_END_EVT:
579 		default:
580 			goto illegal_evt;
581 		}
582 		break;
583 	case LINK_PEER_RESET:
584 		switch (evt) {
585 		case LINK_RESET_EVT:
586 			l->state = LINK_ESTABLISHING;
587 			break;
588 		case LINK_PEER_RESET_EVT:
589 		case LINK_ESTABLISH_EVT:
590 		case LINK_FAILURE_EVT:
591 			break;
592 		case LINK_SYNCH_BEGIN_EVT:
593 		case LINK_SYNCH_END_EVT:
594 		case LINK_FAILOVER_BEGIN_EVT:
595 		case LINK_FAILOVER_END_EVT:
596 		default:
597 			goto illegal_evt;
598 		}
599 		break;
600 	case LINK_FAILINGOVER:
601 		switch (evt) {
602 		case LINK_FAILOVER_END_EVT:
603 			l->state = LINK_RESET;
604 			break;
605 		case LINK_PEER_RESET_EVT:
606 		case LINK_RESET_EVT:
607 		case LINK_ESTABLISH_EVT:
608 		case LINK_FAILURE_EVT:
609 			break;
610 		case LINK_FAILOVER_BEGIN_EVT:
611 		case LINK_SYNCH_BEGIN_EVT:
612 		case LINK_SYNCH_END_EVT:
613 		default:
614 			goto illegal_evt;
615 		}
616 		break;
617 	case LINK_ESTABLISHING:
618 		switch (evt) {
619 		case LINK_ESTABLISH_EVT:
620 			l->state = LINK_ESTABLISHED;
621 			break;
622 		case LINK_FAILOVER_BEGIN_EVT:
623 			l->state = LINK_FAILINGOVER;
624 			break;
625 		case LINK_RESET_EVT:
626 			l->state = LINK_RESET;
627 			break;
628 		case LINK_FAILURE_EVT:
629 		case LINK_PEER_RESET_EVT:
630 		case LINK_SYNCH_BEGIN_EVT:
631 		case LINK_FAILOVER_END_EVT:
632 			break;
633 		case LINK_SYNCH_END_EVT:
634 		default:
635 			goto illegal_evt;
636 		}
637 		break;
638 	case LINK_ESTABLISHED:
639 		switch (evt) {
640 		case LINK_PEER_RESET_EVT:
641 			l->state = LINK_PEER_RESET;
642 			rc |= TIPC_LINK_DOWN_EVT;
643 			break;
644 		case LINK_FAILURE_EVT:
645 			l->state = LINK_RESETTING;
646 			rc |= TIPC_LINK_DOWN_EVT;
647 			break;
648 		case LINK_RESET_EVT:
649 			l->state = LINK_RESET;
650 			break;
651 		case LINK_ESTABLISH_EVT:
652 		case LINK_SYNCH_END_EVT:
653 			break;
654 		case LINK_SYNCH_BEGIN_EVT:
655 			l->state = LINK_SYNCHING;
656 			break;
657 		case LINK_FAILOVER_BEGIN_EVT:
658 		case LINK_FAILOVER_END_EVT:
659 		default:
660 			goto illegal_evt;
661 		}
662 		break;
663 	case LINK_SYNCHING:
664 		switch (evt) {
665 		case LINK_PEER_RESET_EVT:
666 			l->state = LINK_PEER_RESET;
667 			rc |= TIPC_LINK_DOWN_EVT;
668 			break;
669 		case LINK_FAILURE_EVT:
670 			l->state = LINK_RESETTING;
671 			rc |= TIPC_LINK_DOWN_EVT;
672 			break;
673 		case LINK_RESET_EVT:
674 			l->state = LINK_RESET;
675 			break;
676 		case LINK_ESTABLISH_EVT:
677 		case LINK_SYNCH_BEGIN_EVT:
678 			break;
679 		case LINK_SYNCH_END_EVT:
680 			l->state = LINK_ESTABLISHED;
681 			break;
682 		case LINK_FAILOVER_BEGIN_EVT:
683 		case LINK_FAILOVER_END_EVT:
684 		default:
685 			goto illegal_evt;
686 		}
687 		break;
688 	default:
689 		pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
690 	}
691 	return rc;
692 illegal_evt:
693 	pr_err("Illegal FSM event %x in state %x on link %s\n",
694 	       evt, l->state, l->name);
695 	return rc;
696 }
697 
698 /* link_profile_stats - update statistical profiling of traffic
699  */
700 static void link_profile_stats(struct tipc_link *l)
701 {
702 	struct sk_buff *skb;
703 	struct tipc_msg *msg;
704 	int length;
705 
706 	/* Update counters used in statistical profiling of send traffic */
707 	l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
708 	l->stats.queue_sz_counts++;
709 
710 	skb = skb_peek(&l->transmq);
711 	if (!skb)
712 		return;
713 	msg = buf_msg(skb);
714 	length = msg_size(msg);
715 
716 	if (msg_user(msg) == MSG_FRAGMENTER) {
717 		if (msg_type(msg) != FIRST_FRAGMENT)
718 			return;
719 		length = msg_size(msg_get_wrapped(msg));
720 	}
721 	l->stats.msg_lengths_total += length;
722 	l->stats.msg_length_counts++;
723 	if (length <= 64)
724 		l->stats.msg_length_profile[0]++;
725 	else if (length <= 256)
726 		l->stats.msg_length_profile[1]++;
727 	else if (length <= 1024)
728 		l->stats.msg_length_profile[2]++;
729 	else if (length <= 4096)
730 		l->stats.msg_length_profile[3]++;
731 	else if (length <= 16384)
732 		l->stats.msg_length_profile[4]++;
733 	else if (length <= 32768)
734 		l->stats.msg_length_profile[5]++;
735 	else
736 		l->stats.msg_length_profile[6]++;
737 }
738 
739 /* tipc_link_timeout - perform periodic task as instructed from node timeout
740  */
741 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
742 {
743 	int mtyp = 0;
744 	int rc = 0;
745 	bool state = false;
746 	bool probe = false;
747 	bool setup = false;
748 	u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
749 	u16 bc_acked = l->bc_rcvlink->acked;
750 	struct tipc_mon_state *mstate = &l->mon_state;
751 
752 	switch (l->state) {
753 	case LINK_ESTABLISHED:
754 	case LINK_SYNCHING:
755 		mtyp = STATE_MSG;
756 		link_profile_stats(l);
757 		tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
758 		if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
759 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
760 		state = bc_acked != bc_snt;
761 		state |= l->bc_rcvlink->rcv_unacked;
762 		state |= l->rcv_unacked;
763 		state |= !skb_queue_empty(&l->transmq);
764 		state |= !skb_queue_empty(&l->deferdq);
765 		probe = mstate->probing;
766 		probe |= l->silent_intv_cnt;
767 		if (probe || mstate->monitoring)
768 			l->silent_intv_cnt++;
769 		break;
770 	case LINK_RESET:
771 		setup = l->rst_cnt++ <= 4;
772 		setup |= !(l->rst_cnt % 16);
773 		mtyp = RESET_MSG;
774 		break;
775 	case LINK_ESTABLISHING:
776 		setup = true;
777 		mtyp = ACTIVATE_MSG;
778 		break;
779 	case LINK_PEER_RESET:
780 	case LINK_RESETTING:
781 	case LINK_FAILINGOVER:
782 		break;
783 	default:
784 		break;
785 	}
786 
787 	if (state || probe || setup)
788 		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
789 
790 	return rc;
791 }
792 
793 /**
794  * link_schedule_user - schedule a message sender for wakeup after congestion
795  * @l: congested link
796  * @hdr: header of message that is being sent
797  * Create pseudo msg to send back to user when congestion abates
798  */
799 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
800 {
801 	u32 dnode = tipc_own_addr(l->net);
802 	u32 dport = msg_origport(hdr);
803 	struct sk_buff *skb;
804 
805 	/* Create and schedule wakeup pseudo message */
806 	skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
807 			      dnode, l->addr, dport, 0, 0);
808 	if (!skb)
809 		return -ENOBUFS;
810 	msg_set_dest_droppable(buf_msg(skb), true);
811 	TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
812 	skb_queue_tail(&l->wakeupq, skb);
813 	l->stats.link_congs++;
814 	return -ELINKCONG;
815 }
816 
817 /**
818  * link_prepare_wakeup - prepare users for wakeup after congestion
819  * @l: congested link
820  * Wake up a number of waiting users, as permitted by available space
821  * in the send queue
822  */
823 static void link_prepare_wakeup(struct tipc_link *l)
824 {
825 	struct sk_buff *skb, *tmp;
826 	int imp, i = 0;
827 
828 	skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
829 		imp = TIPC_SKB_CB(skb)->chain_imp;
830 		if (l->backlog[imp].len < l->backlog[imp].limit) {
831 			skb_unlink(skb, &l->wakeupq);
832 			skb_queue_tail(l->inputq, skb);
833 		} else if (i++ > 10) {
834 			break;
835 		}
836 	}
837 }
838 
839 void tipc_link_reset(struct tipc_link *l)
840 {
841 	l->in_session = false;
842 	l->session++;
843 	l->mtu = l->advertised_mtu;
844 	__skb_queue_purge(&l->transmq);
845 	__skb_queue_purge(&l->deferdq);
846 	skb_queue_splice_init(&l->wakeupq, l->inputq);
847 	__skb_queue_purge(&l->backlogq);
848 	l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
849 	l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
850 	l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
851 	l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
852 	l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
853 	kfree_skb(l->reasm_buf);
854 	kfree_skb(l->failover_reasm_skb);
855 	l->reasm_buf = NULL;
856 	l->failover_reasm_skb = NULL;
857 	l->rcv_unacked = 0;
858 	l->snd_nxt = 1;
859 	l->rcv_nxt = 1;
860 	l->snd_nxt_state = 1;
861 	l->rcv_nxt_state = 1;
862 	l->acked = 0;
863 	l->silent_intv_cnt = 0;
864 	l->rst_cnt = 0;
865 	l->stale_cnt = 0;
866 	l->bc_peer_is_up = false;
867 	memset(&l->mon_state, 0, sizeof(l->mon_state));
868 	tipc_link_reset_stats(l);
869 }
870 
871 /**
872  * tipc_link_xmit(): enqueue buffer list according to queue situation
873  * @link: link to use
874  * @list: chain of buffers containing message
875  * @xmitq: returned list of packets to be sent by caller
876  *
877  * Consumes the buffer chain.
878  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
879  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
880  */
881 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
882 		   struct sk_buff_head *xmitq)
883 {
884 	struct tipc_msg *hdr = buf_msg(skb_peek(list));
885 	unsigned int maxwin = l->window;
886 	int imp = msg_importance(hdr);
887 	unsigned int mtu = l->mtu;
888 	u16 ack = l->rcv_nxt - 1;
889 	u16 seqno = l->snd_nxt;
890 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
891 	struct sk_buff_head *transmq = &l->transmq;
892 	struct sk_buff_head *backlogq = &l->backlogq;
893 	struct sk_buff *skb, *_skb, *bskb;
894 	int pkt_cnt = skb_queue_len(list);
895 	int rc = 0;
896 
897 	if (unlikely(msg_size(hdr) > mtu)) {
898 		skb_queue_purge(list);
899 		return -EMSGSIZE;
900 	}
901 
902 	/* Allow oversubscription of one data msg per source at congestion */
903 	if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
904 		if (imp == TIPC_SYSTEM_IMPORTANCE) {
905 			pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
906 			return -ENOBUFS;
907 		}
908 		rc = link_schedule_user(l, hdr);
909 	}
910 
911 	if (pkt_cnt > 1) {
912 		l->stats.sent_fragmented++;
913 		l->stats.sent_fragments += pkt_cnt;
914 	}
915 
916 	/* Prepare each packet for sending, and add to relevant queue: */
917 	while (skb_queue_len(list)) {
918 		skb = skb_peek(list);
919 		hdr = buf_msg(skb);
920 		msg_set_seqno(hdr, seqno);
921 		msg_set_ack(hdr, ack);
922 		msg_set_bcast_ack(hdr, bc_ack);
923 
924 		if (likely(skb_queue_len(transmq) < maxwin)) {
925 			_skb = skb_clone(skb, GFP_ATOMIC);
926 			if (!_skb) {
927 				skb_queue_purge(list);
928 				return -ENOBUFS;
929 			}
930 			__skb_dequeue(list);
931 			__skb_queue_tail(transmq, skb);
932 			__skb_queue_tail(xmitq, _skb);
933 			TIPC_SKB_CB(skb)->ackers = l->ackers;
934 			l->rcv_unacked = 0;
935 			l->stats.sent_pkts++;
936 			seqno++;
937 			continue;
938 		}
939 		if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
940 			kfree_skb(__skb_dequeue(list));
941 			l->stats.sent_bundled++;
942 			continue;
943 		}
944 		if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
945 			kfree_skb(__skb_dequeue(list));
946 			__skb_queue_tail(backlogq, bskb);
947 			l->backlog[msg_importance(buf_msg(bskb))].len++;
948 			l->stats.sent_bundled++;
949 			l->stats.sent_bundles++;
950 			continue;
951 		}
952 		l->backlog[imp].len += skb_queue_len(list);
953 		skb_queue_splice_tail_init(list, backlogq);
954 	}
955 	l->snd_nxt = seqno;
956 	return rc;
957 }
958 
959 static void tipc_link_advance_backlog(struct tipc_link *l,
960 				      struct sk_buff_head *xmitq)
961 {
962 	struct sk_buff *skb, *_skb;
963 	struct tipc_msg *hdr;
964 	u16 seqno = l->snd_nxt;
965 	u16 ack = l->rcv_nxt - 1;
966 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
967 
968 	while (skb_queue_len(&l->transmq) < l->window) {
969 		skb = skb_peek(&l->backlogq);
970 		if (!skb)
971 			break;
972 		_skb = skb_clone(skb, GFP_ATOMIC);
973 		if (!_skb)
974 			break;
975 		__skb_dequeue(&l->backlogq);
976 		hdr = buf_msg(skb);
977 		l->backlog[msg_importance(hdr)].len--;
978 		__skb_queue_tail(&l->transmq, skb);
979 		__skb_queue_tail(xmitq, _skb);
980 		TIPC_SKB_CB(skb)->ackers = l->ackers;
981 		msg_set_seqno(hdr, seqno);
982 		msg_set_ack(hdr, ack);
983 		msg_set_bcast_ack(hdr, bc_ack);
984 		l->rcv_unacked = 0;
985 		l->stats.sent_pkts++;
986 		seqno++;
987 	}
988 	l->snd_nxt = seqno;
989 }
990 
991 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
992 {
993 	struct tipc_msg *hdr = buf_msg(skb);
994 
995 	pr_warn("Retransmission failure on link <%s>\n", l->name);
996 	link_print(l, "State of link ");
997 	pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
998 		msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
999 	pr_info("sqno %u, prev: %x, src: %x\n",
1000 		msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
1001 }
1002 
1003 /* tipc_link_retrans() - retransmit one or more packets
1004  * @l: the link to transmit on
1005  * @r: the receiving link ordering the retransmit. Same as l if unicast
1006  * @from: retransmit from (inclusive) this sequence number
1007  * @to: retransmit to (inclusive) this sequence number
1008  * xmitq: queue for accumulating the retransmitted packets
1009  */
1010 static int tipc_link_retrans(struct tipc_link *l, struct tipc_link *r,
1011 			     u16 from, u16 to, struct sk_buff_head *xmitq)
1012 {
1013 	struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1014 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1015 	u16 ack = l->rcv_nxt - 1;
1016 	struct tipc_msg *hdr;
1017 
1018 	if (!skb)
1019 		return 0;
1020 
1021 	/* Detect repeated retransmit failures on same packet */
1022 	if (r->last_retransm != buf_seqno(skb)) {
1023 		r->last_retransm = buf_seqno(skb);
1024 		r->stale_limit = jiffies + msecs_to_jiffies(l->tolerance);
1025 	} else if (++r->stale_cnt > 99 && time_after(jiffies, r->stale_limit)) {
1026 		link_retransmit_failure(l, skb);
1027 		if (link_is_bc_sndlink(l))
1028 			return TIPC_LINK_DOWN_EVT;
1029 		return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1030 	}
1031 
1032 	skb_queue_walk(&l->transmq, skb) {
1033 		hdr = buf_msg(skb);
1034 		if (less(msg_seqno(hdr), from))
1035 			continue;
1036 		if (more(msg_seqno(hdr), to))
1037 			break;
1038 		_skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1039 		if (!_skb)
1040 			return 0;
1041 		hdr = buf_msg(_skb);
1042 		msg_set_ack(hdr, ack);
1043 		msg_set_bcast_ack(hdr, bc_ack);
1044 		_skb->priority = TC_PRIO_CONTROL;
1045 		__skb_queue_tail(xmitq, _skb);
1046 		l->stats.retransmitted++;
1047 	}
1048 	return 0;
1049 }
1050 
1051 /* tipc_data_input - deliver data and name distr msgs to upper layer
1052  *
1053  * Consumes buffer if message is of right type
1054  * Node lock must be held
1055  */
1056 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1057 			    struct sk_buff_head *inputq)
1058 {
1059 	struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1060 	struct tipc_msg *hdr = buf_msg(skb);
1061 
1062 	switch (msg_user(hdr)) {
1063 	case TIPC_LOW_IMPORTANCE:
1064 	case TIPC_MEDIUM_IMPORTANCE:
1065 	case TIPC_HIGH_IMPORTANCE:
1066 	case TIPC_CRITICAL_IMPORTANCE:
1067 		if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1068 			skb_queue_tail(mc_inputq, skb);
1069 			return true;
1070 		}
1071 		/* else: fall through */
1072 	case CONN_MANAGER:
1073 		skb_queue_tail(inputq, skb);
1074 		return true;
1075 	case GROUP_PROTOCOL:
1076 		skb_queue_tail(mc_inputq, skb);
1077 		return true;
1078 	case NAME_DISTRIBUTOR:
1079 		l->bc_rcvlink->state = LINK_ESTABLISHED;
1080 		skb_queue_tail(l->namedq, skb);
1081 		return true;
1082 	case MSG_BUNDLER:
1083 	case TUNNEL_PROTOCOL:
1084 	case MSG_FRAGMENTER:
1085 	case BCAST_PROTOCOL:
1086 		return false;
1087 	default:
1088 		pr_warn("Dropping received illegal msg type\n");
1089 		kfree_skb(skb);
1090 		return false;
1091 	};
1092 }
1093 
1094 /* tipc_link_input - process packet that has passed link protocol check
1095  *
1096  * Consumes buffer
1097  */
1098 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1099 			   struct sk_buff_head *inputq)
1100 {
1101 	struct tipc_msg *hdr = buf_msg(skb);
1102 	struct sk_buff **reasm_skb = &l->reasm_buf;
1103 	struct sk_buff *iskb;
1104 	struct sk_buff_head tmpq;
1105 	int usr = msg_user(hdr);
1106 	int rc = 0;
1107 	int pos = 0;
1108 	int ipos = 0;
1109 
1110 	if (unlikely(usr == TUNNEL_PROTOCOL)) {
1111 		if (msg_type(hdr) == SYNCH_MSG) {
1112 			__skb_queue_purge(&l->deferdq);
1113 			goto drop;
1114 		}
1115 		if (!tipc_msg_extract(skb, &iskb, &ipos))
1116 			return rc;
1117 		kfree_skb(skb);
1118 		skb = iskb;
1119 		hdr = buf_msg(skb);
1120 		if (less(msg_seqno(hdr), l->drop_point))
1121 			goto drop;
1122 		if (tipc_data_input(l, skb, inputq))
1123 			return rc;
1124 		usr = msg_user(hdr);
1125 		reasm_skb = &l->failover_reasm_skb;
1126 	}
1127 
1128 	if (usr == MSG_BUNDLER) {
1129 		skb_queue_head_init(&tmpq);
1130 		l->stats.recv_bundles++;
1131 		l->stats.recv_bundled += msg_msgcnt(hdr);
1132 		while (tipc_msg_extract(skb, &iskb, &pos))
1133 			tipc_data_input(l, iskb, &tmpq);
1134 		tipc_skb_queue_splice_tail(&tmpq, inputq);
1135 		return 0;
1136 	} else if (usr == MSG_FRAGMENTER) {
1137 		l->stats.recv_fragments++;
1138 		if (tipc_buf_append(reasm_skb, &skb)) {
1139 			l->stats.recv_fragmented++;
1140 			tipc_data_input(l, skb, inputq);
1141 		} else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1142 			pr_warn_ratelimited("Unable to build fragment list\n");
1143 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1144 		}
1145 		return 0;
1146 	} else if (usr == BCAST_PROTOCOL) {
1147 		tipc_bcast_lock(l->net);
1148 		tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1149 		tipc_bcast_unlock(l->net);
1150 	}
1151 drop:
1152 	kfree_skb(skb);
1153 	return 0;
1154 }
1155 
1156 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1157 {
1158 	bool released = false;
1159 	struct sk_buff *skb, *tmp;
1160 
1161 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1162 		if (more(buf_seqno(skb), acked))
1163 			break;
1164 		__skb_unlink(skb, &l->transmq);
1165 		kfree_skb(skb);
1166 		released = true;
1167 	}
1168 	return released;
1169 }
1170 
1171 /* tipc_link_build_state_msg: prepare link state message for transmission
1172  *
1173  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1174  * risk of ack storms towards the sender
1175  */
1176 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1177 {
1178 	if (!l)
1179 		return 0;
1180 
1181 	/* Broadcast ACK must be sent via a unicast link => defer to caller */
1182 	if (link_is_bc_rcvlink(l)) {
1183 		if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1184 			return 0;
1185 		l->rcv_unacked = 0;
1186 
1187 		/* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1188 		l->snd_nxt = l->rcv_nxt;
1189 		return TIPC_LINK_SND_STATE;
1190 	}
1191 
1192 	/* Unicast ACK */
1193 	l->rcv_unacked = 0;
1194 	l->stats.sent_acks++;
1195 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1196 	return 0;
1197 }
1198 
1199 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1200  */
1201 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1202 {
1203 	int mtyp = RESET_MSG;
1204 	struct sk_buff *skb;
1205 
1206 	if (l->state == LINK_ESTABLISHING)
1207 		mtyp = ACTIVATE_MSG;
1208 
1209 	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1210 
1211 	/* Inform peer that this endpoint is going down if applicable */
1212 	skb = skb_peek_tail(xmitq);
1213 	if (skb && (l->state == LINK_RESET))
1214 		msg_set_peer_stopping(buf_msg(skb), 1);
1215 }
1216 
1217 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1218  * Note that sending of broadcast NACK is coordinated among nodes, to
1219  * reduce the risk of NACK storms towards the sender
1220  */
1221 static int tipc_link_build_nack_msg(struct tipc_link *l,
1222 				    struct sk_buff_head *xmitq)
1223 {
1224 	u32 def_cnt = ++l->stats.deferred_recv;
1225 	int match1, match2;
1226 
1227 	if (link_is_bc_rcvlink(l)) {
1228 		match1 = def_cnt & 0xf;
1229 		match2 = tipc_own_addr(l->net) & 0xf;
1230 		if (match1 == match2)
1231 			return TIPC_LINK_SND_STATE;
1232 		return 0;
1233 	}
1234 
1235 	if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1236 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1237 	return 0;
1238 }
1239 
1240 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1241  * @l: the link that should handle the message
1242  * @skb: TIPC packet
1243  * @xmitq: queue to place packets to be sent after this call
1244  */
1245 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1246 		  struct sk_buff_head *xmitq)
1247 {
1248 	struct sk_buff_head *defq = &l->deferdq;
1249 	struct tipc_msg *hdr;
1250 	u16 seqno, rcv_nxt, win_lim;
1251 	int rc = 0;
1252 
1253 	do {
1254 		hdr = buf_msg(skb);
1255 		seqno = msg_seqno(hdr);
1256 		rcv_nxt = l->rcv_nxt;
1257 		win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1258 
1259 		/* Verify and update link state */
1260 		if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1261 			return tipc_link_proto_rcv(l, skb, xmitq);
1262 
1263 		if (unlikely(!link_is_up(l))) {
1264 			if (l->state == LINK_ESTABLISHING)
1265 				rc = TIPC_LINK_UP_EVT;
1266 			goto drop;
1267 		}
1268 
1269 		/* Don't send probe at next timeout expiration */
1270 		l->silent_intv_cnt = 0;
1271 
1272 		/* Drop if outside receive window */
1273 		if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1274 			l->stats.duplicates++;
1275 			goto drop;
1276 		}
1277 
1278 		/* Forward queues and wake up waiting users */
1279 		if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1280 			l->stale_cnt = 0;
1281 			tipc_link_advance_backlog(l, xmitq);
1282 			if (unlikely(!skb_queue_empty(&l->wakeupq)))
1283 				link_prepare_wakeup(l);
1284 		}
1285 
1286 		/* Defer delivery if sequence gap */
1287 		if (unlikely(seqno != rcv_nxt)) {
1288 			__tipc_skb_queue_sorted(defq, seqno, skb);
1289 			rc |= tipc_link_build_nack_msg(l, xmitq);
1290 			break;
1291 		}
1292 
1293 		/* Deliver packet */
1294 		l->rcv_nxt++;
1295 		l->stats.recv_pkts++;
1296 		if (!tipc_data_input(l, skb, l->inputq))
1297 			rc |= tipc_link_input(l, skb, l->inputq);
1298 		if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1299 			rc |= tipc_link_build_state_msg(l, xmitq);
1300 		if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1301 			break;
1302 	} while ((skb = __skb_dequeue(defq)));
1303 
1304 	return rc;
1305 drop:
1306 	kfree_skb(skb);
1307 	return rc;
1308 }
1309 
1310 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1311 				      bool probe_reply, u16 rcvgap,
1312 				      int tolerance, int priority,
1313 				      struct sk_buff_head *xmitq)
1314 {
1315 	struct tipc_link *bcl = l->bc_rcvlink;
1316 	struct sk_buff *skb;
1317 	struct tipc_msg *hdr;
1318 	struct sk_buff_head *dfq = &l->deferdq;
1319 	bool node_up = link_is_up(bcl);
1320 	struct tipc_mon_state *mstate = &l->mon_state;
1321 	int dlen = 0;
1322 	void *data;
1323 
1324 	/* Don't send protocol message during reset or link failover */
1325 	if (tipc_link_is_blocked(l))
1326 		return;
1327 
1328 	if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1329 		return;
1330 
1331 	if (!skb_queue_empty(dfq))
1332 		rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1333 
1334 	skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1335 			      tipc_max_domain_size, l->addr,
1336 			      tipc_own_addr(l->net), 0, 0, 0);
1337 	if (!skb)
1338 		return;
1339 
1340 	hdr = buf_msg(skb);
1341 	data = msg_data(hdr);
1342 	msg_set_session(hdr, l->session);
1343 	msg_set_bearer_id(hdr, l->bearer_id);
1344 	msg_set_net_plane(hdr, l->net_plane);
1345 	msg_set_next_sent(hdr, l->snd_nxt);
1346 	msg_set_ack(hdr, l->rcv_nxt - 1);
1347 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1348 	msg_set_bc_ack_invalid(hdr, !node_up);
1349 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1350 	msg_set_link_tolerance(hdr, tolerance);
1351 	msg_set_linkprio(hdr, priority);
1352 	msg_set_redundant_link(hdr, node_up);
1353 	msg_set_seq_gap(hdr, 0);
1354 	msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1355 
1356 	if (mtyp == STATE_MSG) {
1357 		if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1358 			msg_set_seqno(hdr, l->snd_nxt_state++);
1359 		msg_set_seq_gap(hdr, rcvgap);
1360 		msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1361 		msg_set_probe(hdr, probe);
1362 		msg_set_is_keepalive(hdr, probe || probe_reply);
1363 		tipc_mon_prep(l->net, data, &dlen, mstate, l->bearer_id);
1364 		msg_set_size(hdr, INT_H_SIZE + dlen);
1365 		skb_trim(skb, INT_H_SIZE + dlen);
1366 		l->stats.sent_states++;
1367 		l->rcv_unacked = 0;
1368 	} else {
1369 		/* RESET_MSG or ACTIVATE_MSG */
1370 		msg_set_max_pkt(hdr, l->advertised_mtu);
1371 		strcpy(data, l->if_name);
1372 		msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1373 		skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1374 	}
1375 	if (probe)
1376 		l->stats.sent_probes++;
1377 	if (rcvgap)
1378 		l->stats.sent_nacks++;
1379 	skb->priority = TC_PRIO_CONTROL;
1380 	__skb_queue_tail(xmitq, skb);
1381 }
1382 
1383 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1384  * with contents of the link's transmit and backlog queues.
1385  */
1386 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1387 			   int mtyp, struct sk_buff_head *xmitq)
1388 {
1389 	struct sk_buff *skb, *tnlskb;
1390 	struct tipc_msg *hdr, tnlhdr;
1391 	struct sk_buff_head *queue = &l->transmq;
1392 	struct sk_buff_head tmpxq, tnlq;
1393 	u16 pktlen, pktcnt, seqno = l->snd_nxt;
1394 
1395 	if (!tnl)
1396 		return;
1397 
1398 	skb_queue_head_init(&tnlq);
1399 	skb_queue_head_init(&tmpxq);
1400 
1401 	/* At least one packet required for safe algorithm => add dummy */
1402 	skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1403 			      BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1404 			      0, 0, TIPC_ERR_NO_PORT);
1405 	if (!skb) {
1406 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1407 		return;
1408 	}
1409 	skb_queue_tail(&tnlq, skb);
1410 	tipc_link_xmit(l, &tnlq, &tmpxq);
1411 	__skb_queue_purge(&tmpxq);
1412 
1413 	/* Initialize reusable tunnel packet header */
1414 	tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1415 		      mtyp, INT_H_SIZE, l->addr);
1416 	pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1417 	msg_set_msgcnt(&tnlhdr, pktcnt);
1418 	msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1419 tnl:
1420 	/* Wrap each packet into a tunnel packet */
1421 	skb_queue_walk(queue, skb) {
1422 		hdr = buf_msg(skb);
1423 		if (queue == &l->backlogq)
1424 			msg_set_seqno(hdr, seqno++);
1425 		pktlen = msg_size(hdr);
1426 		msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1427 		tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1428 		if (!tnlskb) {
1429 			pr_warn("%sunable to send packet\n", link_co_err);
1430 			return;
1431 		}
1432 		skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1433 		skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1434 		__skb_queue_tail(&tnlq, tnlskb);
1435 	}
1436 	if (queue != &l->backlogq) {
1437 		queue = &l->backlogq;
1438 		goto tnl;
1439 	}
1440 
1441 	tipc_link_xmit(tnl, &tnlq, xmitq);
1442 
1443 	if (mtyp == FAILOVER_MSG) {
1444 		tnl->drop_point = l->rcv_nxt;
1445 		tnl->failover_reasm_skb = l->reasm_buf;
1446 		l->reasm_buf = NULL;
1447 	}
1448 }
1449 
1450 /* tipc_link_validate_msg(): validate message against current link state
1451  * Returns true if message should be accepted, otherwise false
1452  */
1453 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
1454 {
1455 	u16 curr_session = l->peer_session;
1456 	u16 session = msg_session(hdr);
1457 	int mtyp = msg_type(hdr);
1458 
1459 	if (msg_user(hdr) != LINK_PROTOCOL)
1460 		return true;
1461 
1462 	switch (mtyp) {
1463 	case RESET_MSG:
1464 		if (!l->in_session)
1465 			return true;
1466 		/* Accept only RESET with new session number */
1467 		return more(session, curr_session);
1468 	case ACTIVATE_MSG:
1469 		if (!l->in_session)
1470 			return true;
1471 		/* Accept only ACTIVATE with new or current session number */
1472 		return !less(session, curr_session);
1473 	case STATE_MSG:
1474 		/* Accept only STATE with current session number */
1475 		if (!l->in_session)
1476 			return false;
1477 		if (session != curr_session)
1478 			return false;
1479 		if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
1480 			return true;
1481 		/* Accept only STATE with new sequence number */
1482 		return !less(msg_seqno(hdr), l->rcv_nxt_state);
1483 	default:
1484 		return false;
1485 	}
1486 }
1487 
1488 /* tipc_link_proto_rcv(): receive link level protocol message :
1489  * Note that network plane id propagates through the network, and may
1490  * change at any time. The node with lowest numerical id determines
1491  * network plane
1492  */
1493 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1494 			       struct sk_buff_head *xmitq)
1495 {
1496 	struct tipc_msg *hdr = buf_msg(skb);
1497 	u16 rcvgap = 0;
1498 	u16 ack = msg_ack(hdr);
1499 	u16 gap = msg_seq_gap(hdr);
1500 	u16 peers_snd_nxt =  msg_next_sent(hdr);
1501 	u16 peers_tol = msg_link_tolerance(hdr);
1502 	u16 peers_prio = msg_linkprio(hdr);
1503 	u16 rcv_nxt = l->rcv_nxt;
1504 	u16 dlen = msg_data_sz(hdr);
1505 	int mtyp = msg_type(hdr);
1506 	bool reply = msg_probe(hdr);
1507 	void *data;
1508 	char *if_name;
1509 	int rc = 0;
1510 
1511 	if (tipc_link_is_blocked(l) || !xmitq)
1512 		goto exit;
1513 
1514 	if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1515 		l->net_plane = msg_net_plane(hdr);
1516 
1517 	skb_linearize(skb);
1518 	hdr = buf_msg(skb);
1519 	data = msg_data(hdr);
1520 
1521 	if (!tipc_link_validate_msg(l, hdr))
1522 		goto exit;
1523 
1524 	switch (mtyp) {
1525 	case RESET_MSG:
1526 	case ACTIVATE_MSG:
1527 		/* Complete own link name with peer's interface name */
1528 		if_name =  strrchr(l->name, ':') + 1;
1529 		if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1530 			break;
1531 		if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1532 			break;
1533 		strncpy(if_name, data, TIPC_MAX_IF_NAME);
1534 
1535 		/* Update own tolerance if peer indicates a non-zero value */
1536 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1537 			l->tolerance = peers_tol;
1538 
1539 		/* Update own priority if peer's priority is higher */
1540 		if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1541 			l->priority = peers_prio;
1542 
1543 		/* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1544 		if (msg_peer_stopping(hdr))
1545 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1546 		else if ((mtyp == RESET_MSG) || !link_is_up(l))
1547 			rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1548 
1549 		/* ACTIVATE_MSG takes up link if it was already locally reset */
1550 		if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1551 			rc = TIPC_LINK_UP_EVT;
1552 
1553 		l->peer_session = msg_session(hdr);
1554 		l->in_session = true;
1555 		l->peer_bearer_id = msg_bearer_id(hdr);
1556 		if (l->mtu > msg_max_pkt(hdr))
1557 			l->mtu = msg_max_pkt(hdr);
1558 		break;
1559 
1560 	case STATE_MSG:
1561 		l->rcv_nxt_state = msg_seqno(hdr) + 1;
1562 
1563 		/* Update own tolerance if peer indicates a non-zero value */
1564 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1565 			l->tolerance = peers_tol;
1566 
1567 		/* Update own prio if peer indicates a different value */
1568 		if ((peers_prio != l->priority) &&
1569 		    in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
1570 			l->priority = peers_prio;
1571 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1572 		}
1573 
1574 		l->silent_intv_cnt = 0;
1575 		l->stats.recv_states++;
1576 		if (msg_probe(hdr))
1577 			l->stats.recv_probes++;
1578 
1579 		if (!link_is_up(l)) {
1580 			if (l->state == LINK_ESTABLISHING)
1581 				rc = TIPC_LINK_UP_EVT;
1582 			break;
1583 		}
1584 		tipc_mon_rcv(l->net, data, dlen, l->addr,
1585 			     &l->mon_state, l->bearer_id);
1586 
1587 		/* Send NACK if peer has sent pkts we haven't received yet */
1588 		if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1589 			rcvgap = peers_snd_nxt - l->rcv_nxt;
1590 		if (rcvgap || reply)
1591 			tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
1592 						  rcvgap, 0, 0, xmitq);
1593 		tipc_link_release_pkts(l, ack);
1594 
1595 		/* If NACK, retransmit will now start at right position */
1596 		if (gap) {
1597 			rc = tipc_link_retrans(l, l, ack + 1, ack + gap, xmitq);
1598 			l->stats.recv_nacks++;
1599 		}
1600 
1601 		tipc_link_advance_backlog(l, xmitq);
1602 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
1603 			link_prepare_wakeup(l);
1604 	}
1605 exit:
1606 	kfree_skb(skb);
1607 	return rc;
1608 }
1609 
1610 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1611  */
1612 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1613 					 u16 peers_snd_nxt,
1614 					 struct sk_buff_head *xmitq)
1615 {
1616 	struct sk_buff *skb;
1617 	struct tipc_msg *hdr;
1618 	struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1619 	u16 ack = l->rcv_nxt - 1;
1620 	u16 gap_to = peers_snd_nxt - 1;
1621 
1622 	skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1623 			      0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1624 	if (!skb)
1625 		return false;
1626 	hdr = buf_msg(skb);
1627 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1628 	msg_set_bcast_ack(hdr, ack);
1629 	msg_set_bcgap_after(hdr, ack);
1630 	if (dfrd_skb)
1631 		gap_to = buf_seqno(dfrd_skb) - 1;
1632 	msg_set_bcgap_to(hdr, gap_to);
1633 	msg_set_non_seq(hdr, bcast);
1634 	__skb_queue_tail(xmitq, skb);
1635 	return true;
1636 }
1637 
1638 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1639  *
1640  * Give a newly added peer node the sequence number where it should
1641  * start receiving and acking broadcast packets.
1642  */
1643 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1644 					struct sk_buff_head *xmitq)
1645 {
1646 	struct sk_buff_head list;
1647 
1648 	__skb_queue_head_init(&list);
1649 	if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1650 		return;
1651 	msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
1652 	tipc_link_xmit(l, &list, xmitq);
1653 }
1654 
1655 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1656  */
1657 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1658 {
1659 	int mtyp = msg_type(hdr);
1660 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1661 
1662 	if (link_is_up(l))
1663 		return;
1664 
1665 	if (msg_user(hdr) == BCAST_PROTOCOL) {
1666 		l->rcv_nxt = peers_snd_nxt;
1667 		l->state = LINK_ESTABLISHED;
1668 		return;
1669 	}
1670 
1671 	if (l->peer_caps & TIPC_BCAST_SYNCH)
1672 		return;
1673 
1674 	if (msg_peer_node_is_up(hdr))
1675 		return;
1676 
1677 	/* Compatibility: accept older, less safe initial synch data */
1678 	if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1679 		l->rcv_nxt = peers_snd_nxt;
1680 }
1681 
1682 /* link_bc_retr eval()- check if the indicated range can be retransmitted now
1683  * - Adjust permitted range if there is overlap with previous retransmission
1684  */
1685 static bool link_bc_retr_eval(struct tipc_link *l, u16 *from, u16 *to)
1686 {
1687 	unsigned long elapsed = jiffies_to_msecs(jiffies - l->prev_retr);
1688 
1689 	if (less(*to, *from))
1690 		return false;
1691 
1692 	/* New retransmission request */
1693 	if ((elapsed > TIPC_BC_RETR_LIMIT) ||
1694 	    less(*to, l->prev_from) || more(*from, l->prev_to)) {
1695 		l->prev_from = *from;
1696 		l->prev_to = *to;
1697 		l->prev_retr = jiffies;
1698 		return true;
1699 	}
1700 
1701 	/* Inside range of previous retransmit */
1702 	if (!less(*from, l->prev_from) && !more(*to, l->prev_to))
1703 		return false;
1704 
1705 	/* Fully or partially outside previous range => exclude overlap */
1706 	if (less(*from, l->prev_from)) {
1707 		*to = l->prev_from - 1;
1708 		l->prev_from = *from;
1709 	}
1710 	if (more(*to, l->prev_to)) {
1711 		*from = l->prev_to + 1;
1712 		l->prev_to = *to;
1713 	}
1714 	l->prev_retr = jiffies;
1715 	return true;
1716 }
1717 
1718 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1719  */
1720 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1721 			  struct sk_buff_head *xmitq)
1722 {
1723 	struct tipc_link *snd_l = l->bc_sndlink;
1724 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1725 	u16 from = msg_bcast_ack(hdr) + 1;
1726 	u16 to = from + msg_bc_gap(hdr) - 1;
1727 	int rc = 0;
1728 
1729 	if (!link_is_up(l))
1730 		return rc;
1731 
1732 	if (!msg_peer_node_is_up(hdr))
1733 		return rc;
1734 
1735 	/* Open when peer ackowledges our bcast init msg (pkt #1) */
1736 	if (msg_ack(hdr))
1737 		l->bc_peer_is_up = true;
1738 
1739 	if (!l->bc_peer_is_up)
1740 		return rc;
1741 
1742 	l->stats.recv_nacks++;
1743 
1744 	/* Ignore if peers_snd_nxt goes beyond receive window */
1745 	if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1746 		return rc;
1747 
1748 	if (link_bc_retr_eval(snd_l, &from, &to))
1749 		rc = tipc_link_retrans(snd_l, l, from, to, xmitq);
1750 
1751 	l->snd_nxt = peers_snd_nxt;
1752 	if (link_bc_rcv_gap(l))
1753 		rc |= TIPC_LINK_SND_STATE;
1754 
1755 	/* Return now if sender supports nack via STATE messages */
1756 	if (l->peer_caps & TIPC_BCAST_STATE_NACK)
1757 		return rc;
1758 
1759 	/* Otherwise, be backwards compatible */
1760 
1761 	if (!more(peers_snd_nxt, l->rcv_nxt)) {
1762 		l->nack_state = BC_NACK_SND_CONDITIONAL;
1763 		return 0;
1764 	}
1765 
1766 	/* Don't NACK if one was recently sent or peeked */
1767 	if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1768 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1769 		return 0;
1770 	}
1771 
1772 	/* Conditionally delay NACK sending until next synch rcv */
1773 	if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1774 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1775 		if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1776 			return 0;
1777 	}
1778 
1779 	/* Send NACK now but suppress next one */
1780 	tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1781 	l->nack_state = BC_NACK_SND_SUPPRESS;
1782 	return 0;
1783 }
1784 
1785 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1786 			  struct sk_buff_head *xmitq)
1787 {
1788 	struct sk_buff *skb, *tmp;
1789 	struct tipc_link *snd_l = l->bc_sndlink;
1790 
1791 	if (!link_is_up(l) || !l->bc_peer_is_up)
1792 		return;
1793 
1794 	if (!more(acked, l->acked))
1795 		return;
1796 
1797 	/* Skip over packets peer has already acked */
1798 	skb_queue_walk(&snd_l->transmq, skb) {
1799 		if (more(buf_seqno(skb), l->acked))
1800 			break;
1801 	}
1802 
1803 	/* Update/release the packets peer is acking now */
1804 	skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1805 		if (more(buf_seqno(skb), acked))
1806 			break;
1807 		if (!--TIPC_SKB_CB(skb)->ackers) {
1808 			__skb_unlink(skb, &snd_l->transmq);
1809 			kfree_skb(skb);
1810 		}
1811 	}
1812 	l->acked = acked;
1813 	tipc_link_advance_backlog(snd_l, xmitq);
1814 	if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1815 		link_prepare_wakeup(snd_l);
1816 }
1817 
1818 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1819  * This function is here for backwards compatibility, since
1820  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
1821  */
1822 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1823 			  struct sk_buff_head *xmitq)
1824 {
1825 	struct tipc_msg *hdr = buf_msg(skb);
1826 	u32 dnode = msg_destnode(hdr);
1827 	int mtyp = msg_type(hdr);
1828 	u16 acked = msg_bcast_ack(hdr);
1829 	u16 from = acked + 1;
1830 	u16 to = msg_bcgap_to(hdr);
1831 	u16 peers_snd_nxt = to + 1;
1832 	int rc = 0;
1833 
1834 	kfree_skb(skb);
1835 
1836 	if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1837 		return 0;
1838 
1839 	if (mtyp != STATE_MSG)
1840 		return 0;
1841 
1842 	if (dnode == tipc_own_addr(l->net)) {
1843 		tipc_link_bc_ack_rcv(l, acked, xmitq);
1844 		rc = tipc_link_retrans(l->bc_sndlink, l, from, to, xmitq);
1845 		l->stats.recv_nacks++;
1846 		return rc;
1847 	}
1848 
1849 	/* Msg for other node => suppress own NACK at next sync if applicable */
1850 	if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1851 		l->nack_state = BC_NACK_SND_SUPPRESS;
1852 
1853 	return 0;
1854 }
1855 
1856 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1857 {
1858 	int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
1859 
1860 	l->window = win;
1861 	l->backlog[TIPC_LOW_IMPORTANCE].limit      = max_t(u16, 50, win);
1862 	l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = max_t(u16, 100, win * 2);
1863 	l->backlog[TIPC_HIGH_IMPORTANCE].limit     = max_t(u16, 150, win * 3);
1864 	l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
1865 	l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
1866 }
1867 
1868 /**
1869  * link_reset_stats - reset link statistics
1870  * @l: pointer to link
1871  */
1872 void tipc_link_reset_stats(struct tipc_link *l)
1873 {
1874 	memset(&l->stats, 0, sizeof(l->stats));
1875 }
1876 
1877 static void link_print(struct tipc_link *l, const char *str)
1878 {
1879 	struct sk_buff *hskb = skb_peek(&l->transmq);
1880 	u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1881 	u16 tail = l->snd_nxt - 1;
1882 
1883 	pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1884 	pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1885 		skb_queue_len(&l->transmq), head, tail,
1886 		skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1887 }
1888 
1889 /* Parse and validate nested (link) properties valid for media, bearer and link
1890  */
1891 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1892 {
1893 	int err;
1894 
1895 	err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1896 			       tipc_nl_prop_policy, NULL);
1897 	if (err)
1898 		return err;
1899 
1900 	if (props[TIPC_NLA_PROP_PRIO]) {
1901 		u32 prio;
1902 
1903 		prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1904 		if (prio > TIPC_MAX_LINK_PRI)
1905 			return -EINVAL;
1906 	}
1907 
1908 	if (props[TIPC_NLA_PROP_TOL]) {
1909 		u32 tol;
1910 
1911 		tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1912 		if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1913 			return -EINVAL;
1914 	}
1915 
1916 	if (props[TIPC_NLA_PROP_WIN]) {
1917 		u32 win;
1918 
1919 		win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1920 		if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1921 			return -EINVAL;
1922 	}
1923 
1924 	return 0;
1925 }
1926 
1927 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1928 {
1929 	int i;
1930 	struct nlattr *stats;
1931 
1932 	struct nla_map {
1933 		u32 key;
1934 		u32 val;
1935 	};
1936 
1937 	struct nla_map map[] = {
1938 		{TIPC_NLA_STATS_RX_INFO, 0},
1939 		{TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1940 		{TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1941 		{TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1942 		{TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1943 		{TIPC_NLA_STATS_TX_INFO, 0},
1944 		{TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1945 		{TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1946 		{TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1947 		{TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1948 		{TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1949 			s->msg_length_counts : 1},
1950 		{TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1951 		{TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1952 		{TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1953 		{TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1954 		{TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1955 		{TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1956 		{TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1957 		{TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1958 		{TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1959 		{TIPC_NLA_STATS_RX_STATES, s->recv_states},
1960 		{TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1961 		{TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1962 		{TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1963 		{TIPC_NLA_STATS_TX_STATES, s->sent_states},
1964 		{TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1965 		{TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1966 		{TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1967 		{TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1968 		{TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1969 		{TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1970 		{TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1971 		{TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1972 			(s->accu_queue_sz / s->queue_sz_counts) : 0}
1973 	};
1974 
1975 	stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1976 	if (!stats)
1977 		return -EMSGSIZE;
1978 
1979 	for (i = 0; i <  ARRAY_SIZE(map); i++)
1980 		if (nla_put_u32(skb, map[i].key, map[i].val))
1981 			goto msg_full;
1982 
1983 	nla_nest_end(skb, stats);
1984 
1985 	return 0;
1986 msg_full:
1987 	nla_nest_cancel(skb, stats);
1988 
1989 	return -EMSGSIZE;
1990 }
1991 
1992 /* Caller should hold appropriate locks to protect the link */
1993 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1994 		       struct tipc_link *link, int nlflags)
1995 {
1996 	u32 self = tipc_own_addr(net);
1997 	struct nlattr *attrs;
1998 	struct nlattr *prop;
1999 	void *hdr;
2000 	int err;
2001 
2002 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2003 			  nlflags, TIPC_NL_LINK_GET);
2004 	if (!hdr)
2005 		return -EMSGSIZE;
2006 
2007 	attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
2008 	if (!attrs)
2009 		goto msg_full;
2010 
2011 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2012 		goto attr_msg_full;
2013 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2014 		goto attr_msg_full;
2015 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2016 		goto attr_msg_full;
2017 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2018 		goto attr_msg_full;
2019 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2020 		goto attr_msg_full;
2021 
2022 	if (tipc_link_is_up(link))
2023 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2024 			goto attr_msg_full;
2025 	if (link->active)
2026 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2027 			goto attr_msg_full;
2028 
2029 	prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2030 	if (!prop)
2031 		goto attr_msg_full;
2032 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2033 		goto prop_msg_full;
2034 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2035 		goto prop_msg_full;
2036 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2037 			link->window))
2038 		goto prop_msg_full;
2039 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2040 		goto prop_msg_full;
2041 	nla_nest_end(msg->skb, prop);
2042 
2043 	err = __tipc_nl_add_stats(msg->skb, &link->stats);
2044 	if (err)
2045 		goto attr_msg_full;
2046 
2047 	nla_nest_end(msg->skb, attrs);
2048 	genlmsg_end(msg->skb, hdr);
2049 
2050 	return 0;
2051 
2052 prop_msg_full:
2053 	nla_nest_cancel(msg->skb, prop);
2054 attr_msg_full:
2055 	nla_nest_cancel(msg->skb, attrs);
2056 msg_full:
2057 	genlmsg_cancel(msg->skb, hdr);
2058 
2059 	return -EMSGSIZE;
2060 }
2061 
2062 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2063 				      struct tipc_stats *stats)
2064 {
2065 	int i;
2066 	struct nlattr *nest;
2067 
2068 	struct nla_map {
2069 		__u32 key;
2070 		__u32 val;
2071 	};
2072 
2073 	struct nla_map map[] = {
2074 		{TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2075 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2076 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2077 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2078 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2079 		{TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2080 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2081 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2082 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2083 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2084 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2085 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2086 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2087 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2088 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2089 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2090 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2091 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2092 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2093 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2094 	};
2095 
2096 	nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
2097 	if (!nest)
2098 		return -EMSGSIZE;
2099 
2100 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2101 		if (nla_put_u32(skb, map[i].key, map[i].val))
2102 			goto msg_full;
2103 
2104 	nla_nest_end(skb, nest);
2105 
2106 	return 0;
2107 msg_full:
2108 	nla_nest_cancel(skb, nest);
2109 
2110 	return -EMSGSIZE;
2111 }
2112 
2113 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2114 {
2115 	int err;
2116 	void *hdr;
2117 	struct nlattr *attrs;
2118 	struct nlattr *prop;
2119 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2120 	struct tipc_link *bcl = tn->bcl;
2121 
2122 	if (!bcl)
2123 		return 0;
2124 
2125 	tipc_bcast_lock(net);
2126 
2127 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2128 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
2129 	if (!hdr) {
2130 		tipc_bcast_unlock(net);
2131 		return -EMSGSIZE;
2132 	}
2133 
2134 	attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
2135 	if (!attrs)
2136 		goto msg_full;
2137 
2138 	/* The broadcast link is always up */
2139 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2140 		goto attr_msg_full;
2141 
2142 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2143 		goto attr_msg_full;
2144 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2145 		goto attr_msg_full;
2146 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2147 		goto attr_msg_full;
2148 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2149 		goto attr_msg_full;
2150 
2151 	prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2152 	if (!prop)
2153 		goto attr_msg_full;
2154 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2155 		goto prop_msg_full;
2156 	nla_nest_end(msg->skb, prop);
2157 
2158 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2159 	if (err)
2160 		goto attr_msg_full;
2161 
2162 	tipc_bcast_unlock(net);
2163 	nla_nest_end(msg->skb, attrs);
2164 	genlmsg_end(msg->skb, hdr);
2165 
2166 	return 0;
2167 
2168 prop_msg_full:
2169 	nla_nest_cancel(msg->skb, prop);
2170 attr_msg_full:
2171 	nla_nest_cancel(msg->skb, attrs);
2172 msg_full:
2173 	tipc_bcast_unlock(net);
2174 	genlmsg_cancel(msg->skb, hdr);
2175 
2176 	return -EMSGSIZE;
2177 }
2178 
2179 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2180 			     struct sk_buff_head *xmitq)
2181 {
2182 	l->tolerance = tol;
2183 	if (link_is_up(l))
2184 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2185 }
2186 
2187 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2188 			struct sk_buff_head *xmitq)
2189 {
2190 	l->priority = prio;
2191 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2192 }
2193 
2194 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2195 {
2196 	l->abort_limit = limit;
2197 }
2198