xref: /openbmc/linux/net/tipc/link.c (revision e8bd417a)
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 #include "trace.h"
47 
48 #include <linux/pkt_sched.h>
49 
50 struct tipc_stats {
51 	u32 sent_pkts;
52 	u32 recv_pkts;
53 	u32 sent_states;
54 	u32 recv_states;
55 	u32 sent_probes;
56 	u32 recv_probes;
57 	u32 sent_nacks;
58 	u32 recv_nacks;
59 	u32 sent_acks;
60 	u32 sent_bundled;
61 	u32 sent_bundles;
62 	u32 recv_bundled;
63 	u32 recv_bundles;
64 	u32 retransmitted;
65 	u32 sent_fragmented;
66 	u32 sent_fragments;
67 	u32 recv_fragmented;
68 	u32 recv_fragments;
69 	u32 link_congs;		/* # port sends blocked by congestion */
70 	u32 deferred_recv;
71 	u32 duplicates;
72 	u32 max_queue_sz;	/* send queue size high water mark */
73 	u32 accu_queue_sz;	/* used for send queue size profiling */
74 	u32 queue_sz_counts;	/* used for send queue size profiling */
75 	u32 msg_length_counts;	/* used for message length profiling */
76 	u32 msg_lengths_total;	/* used for message length profiling */
77 	u32 msg_length_profile[7]; /* used for msg. length profiling */
78 };
79 
80 /**
81  * struct tipc_link - TIPC link data structure
82  * @addr: network address of link's peer node
83  * @name: link name character string
84  * @media_addr: media address to use when sending messages over link
85  * @timer: link timer
86  * @net: pointer to namespace struct
87  * @refcnt: reference counter for permanent references (owner node & timer)
88  * @peer_session: link session # being used by peer end of link
89  * @peer_bearer_id: bearer id used by link's peer endpoint
90  * @bearer_id: local bearer id used by link
91  * @tolerance: minimum link continuity loss needed to reset link [in ms]
92  * @abort_limit: # of unacknowledged continuity probes needed to reset link
93  * @state: current state of link FSM
94  * @peer_caps: bitmap describing capabilities of peer node
95  * @silent_intv_cnt: # of timer intervals without any reception from peer
96  * @proto_msg: template for control messages generated by link
97  * @pmsg: convenience pointer to "proto_msg" field
98  * @priority: current link priority
99  * @net_plane: current link network plane ('A' through 'H')
100  * @mon_state: cookie with information needed by link monitor
101  * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
102  * @exp_msg_count: # of tunnelled messages expected during link changeover
103  * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
104  * @mtu: current maximum packet size for this link
105  * @advertised_mtu: advertised own mtu when link is being established
106  * @transmitq: queue for sent, non-acked messages
107  * @backlogq: queue for messages waiting to be sent
108  * @snt_nxt: next sequence number to use for outbound messages
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
122  */
123 struct tipc_link {
124 	u32 addr;
125 	char name[TIPC_MAX_LINK_NAME];
126 	struct net *net;
127 
128 	/* Management and link supervision data */
129 	u16 peer_session;
130 	u16 session;
131 	u16 snd_nxt_state;
132 	u16 rcv_nxt_state;
133 	u32 peer_bearer_id;
134 	u32 bearer_id;
135 	u32 tolerance;
136 	u32 abort_limit;
137 	u32 state;
138 	u16 peer_caps;
139 	bool in_session;
140 	bool active;
141 	u32 silent_intv_cnt;
142 	char if_name[TIPC_MAX_IF_NAME];
143 	u32 priority;
144 	char net_plane;
145 	struct tipc_mon_state mon_state;
146 	u16 rst_cnt;
147 
148 	/* Failover/synch */
149 	u16 drop_point;
150 	struct sk_buff *failover_reasm_skb;
151 	struct sk_buff_head failover_deferdq;
152 
153 	/* Max packet negotiation */
154 	u16 mtu;
155 	u16 advertised_mtu;
156 
157 	/* Sending */
158 	struct sk_buff_head transmq;
159 	struct sk_buff_head backlogq;
160 	struct {
161 		u16 len;
162 		u16 limit;
163 	} backlog[5];
164 	u16 snd_nxt;
165 	u16 window;
166 
167 	/* Reception */
168 	u16 rcv_nxt;
169 	u32 rcv_unacked;
170 	struct sk_buff_head deferdq;
171 	struct sk_buff_head *inputq;
172 	struct sk_buff_head *namedq;
173 
174 	/* Congestion handling */
175 	struct sk_buff_head wakeupq;
176 
177 	/* Fragmentation/reassembly */
178 	struct sk_buff *reasm_buf;
179 
180 	/* Broadcast */
181 	u16 ackers;
182 	u16 acked;
183 	struct tipc_link *bc_rcvlink;
184 	struct tipc_link *bc_sndlink;
185 	u8 nack_state;
186 	bool bc_peer_is_up;
187 
188 	/* Statistics */
189 	struct tipc_stats stats;
190 };
191 
192 /*
193  * Error message prefixes
194  */
195 static const char *link_co_err = "Link tunneling error, ";
196 static const char *link_rst_msg = "Resetting link ";
197 
198 /* Send states for broadcast NACKs
199  */
200 enum {
201 	BC_NACK_SND_CONDITIONAL,
202 	BC_NACK_SND_UNCONDITIONAL,
203 	BC_NACK_SND_SUPPRESS,
204 };
205 
206 #define TIPC_BC_RETR_LIM  (jiffies + msecs_to_jiffies(10))
207 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
208 
209 /*
210  * Interval between NACKs when packets arrive out of order
211  */
212 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
213 
214 /* Link FSM states:
215  */
216 enum {
217 	LINK_ESTABLISHED     = 0xe,
218 	LINK_ESTABLISHING    = 0xe  << 4,
219 	LINK_RESET           = 0x1  << 8,
220 	LINK_RESETTING       = 0x2  << 12,
221 	LINK_PEER_RESET      = 0xd  << 16,
222 	LINK_FAILINGOVER     = 0xf  << 20,
223 	LINK_SYNCHING        = 0xc  << 24
224 };
225 
226 /* Link FSM state checking routines
227  */
228 static int link_is_up(struct tipc_link *l)
229 {
230 	return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
231 }
232 
233 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
234 			       struct sk_buff_head *xmitq);
235 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
236 				      bool probe_reply, u16 rcvgap,
237 				      int tolerance, int priority,
238 				      struct sk_buff_head *xmitq);
239 static void link_print(struct tipc_link *l, const char *str);
240 static int tipc_link_build_nack_msg(struct tipc_link *l,
241 				    struct sk_buff_head *xmitq);
242 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
243 					struct sk_buff_head *xmitq);
244 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
245 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data);
246 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
247 				     struct tipc_gap_ack_blks *ga,
248 				     struct sk_buff_head *xmitq);
249 
250 /*
251  *  Simple non-static link routines (i.e. referenced outside this file)
252  */
253 bool tipc_link_is_up(struct tipc_link *l)
254 {
255 	return link_is_up(l);
256 }
257 
258 bool tipc_link_peer_is_down(struct tipc_link *l)
259 {
260 	return l->state == LINK_PEER_RESET;
261 }
262 
263 bool tipc_link_is_reset(struct tipc_link *l)
264 {
265 	return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
266 }
267 
268 bool tipc_link_is_establishing(struct tipc_link *l)
269 {
270 	return l->state == LINK_ESTABLISHING;
271 }
272 
273 bool tipc_link_is_synching(struct tipc_link *l)
274 {
275 	return l->state == LINK_SYNCHING;
276 }
277 
278 bool tipc_link_is_failingover(struct tipc_link *l)
279 {
280 	return l->state == LINK_FAILINGOVER;
281 }
282 
283 bool tipc_link_is_blocked(struct tipc_link *l)
284 {
285 	return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
286 }
287 
288 static bool link_is_bc_sndlink(struct tipc_link *l)
289 {
290 	return !l->bc_sndlink;
291 }
292 
293 static bool link_is_bc_rcvlink(struct tipc_link *l)
294 {
295 	return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
296 }
297 
298 void tipc_link_set_active(struct tipc_link *l, bool active)
299 {
300 	l->active = active;
301 }
302 
303 u32 tipc_link_id(struct tipc_link *l)
304 {
305 	return l->peer_bearer_id << 16 | l->bearer_id;
306 }
307 
308 int tipc_link_window(struct tipc_link *l)
309 {
310 	return l->window;
311 }
312 
313 int tipc_link_prio(struct tipc_link *l)
314 {
315 	return l->priority;
316 }
317 
318 unsigned long tipc_link_tolerance(struct tipc_link *l)
319 {
320 	return l->tolerance;
321 }
322 
323 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
324 {
325 	return l->inputq;
326 }
327 
328 char tipc_link_plane(struct tipc_link *l)
329 {
330 	return l->net_plane;
331 }
332 
333 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
334 {
335 	l->peer_caps = capabilities;
336 }
337 
338 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
339 			   struct tipc_link *uc_l,
340 			   struct sk_buff_head *xmitq)
341 {
342 	struct tipc_link *rcv_l = uc_l->bc_rcvlink;
343 
344 	snd_l->ackers++;
345 	rcv_l->acked = snd_l->snd_nxt - 1;
346 	snd_l->state = LINK_ESTABLISHED;
347 	tipc_link_build_bc_init_msg(uc_l, xmitq);
348 }
349 
350 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
351 			      struct tipc_link *rcv_l,
352 			      struct sk_buff_head *xmitq)
353 {
354 	u16 ack = snd_l->snd_nxt - 1;
355 
356 	snd_l->ackers--;
357 	rcv_l->bc_peer_is_up = true;
358 	rcv_l->state = LINK_ESTABLISHED;
359 	tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
360 	trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
361 	tipc_link_reset(rcv_l);
362 	rcv_l->state = LINK_RESET;
363 	if (!snd_l->ackers) {
364 		trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero 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 u32 tipc_link_state(struct tipc_link *l)
414 {
415 	return l->state;
416 }
417 
418 /**
419  * tipc_link_create - create a new link
420  * @n: pointer to associated node
421  * @if_name: associated interface name
422  * @bearer_id: id (index) of associated bearer
423  * @tolerance: link tolerance to be used by link
424  * @net_plane: network plane (A,B,c..) this link belongs to
425  * @mtu: mtu to be advertised by link
426  * @priority: priority to be used by link
427  * @window: send window to be used by link
428  * @session: session to be used by link
429  * @ownnode: identity of own node
430  * @peer: node id of peer node
431  * @peer_caps: bitmap describing peer node capabilities
432  * @bc_sndlink: the namespace global link used for broadcast sending
433  * @bc_rcvlink: the peer specific link used for broadcast reception
434  * @inputq: queue to put messages ready for delivery
435  * @namedq: queue to put binding table update messages ready for delivery
436  * @link: return value, pointer to put the created link
437  *
438  * Returns true if link was created, otherwise false
439  */
440 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
441 		      int tolerance, char net_plane, u32 mtu, int priority,
442 		      int window, u32 session, u32 self,
443 		      u32 peer, u8 *peer_id, u16 peer_caps,
444 		      struct tipc_link *bc_sndlink,
445 		      struct tipc_link *bc_rcvlink,
446 		      struct sk_buff_head *inputq,
447 		      struct sk_buff_head *namedq,
448 		      struct tipc_link **link)
449 {
450 	char peer_str[NODE_ID_STR_LEN] = {0,};
451 	char self_str[NODE_ID_STR_LEN] = {0,};
452 	struct tipc_link *l;
453 
454 	l = kzalloc(sizeof(*l), GFP_ATOMIC);
455 	if (!l)
456 		return false;
457 	*link = l;
458 	l->session = session;
459 
460 	/* Set link name for unicast links only */
461 	if (peer_id) {
462 		tipc_nodeid2string(self_str, tipc_own_id(net));
463 		if (strlen(self_str) > 16)
464 			sprintf(self_str, "%x", self);
465 		tipc_nodeid2string(peer_str, peer_id);
466 		if (strlen(peer_str) > 16)
467 			sprintf(peer_str, "%x", peer);
468 	}
469 	/* Peer i/f name will be completed by reset/activate message */
470 	snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
471 		 self_str, if_name, peer_str);
472 
473 	strcpy(l->if_name, if_name);
474 	l->addr = peer;
475 	l->peer_caps = peer_caps;
476 	l->net = net;
477 	l->in_session = false;
478 	l->bearer_id = bearer_id;
479 	l->tolerance = tolerance;
480 	if (bc_rcvlink)
481 		bc_rcvlink->tolerance = tolerance;
482 	l->net_plane = net_plane;
483 	l->advertised_mtu = mtu;
484 	l->mtu = mtu;
485 	l->priority = priority;
486 	tipc_link_set_queue_limits(l, window);
487 	l->ackers = 1;
488 	l->bc_sndlink = bc_sndlink;
489 	l->bc_rcvlink = bc_rcvlink;
490 	l->inputq = inputq;
491 	l->namedq = namedq;
492 	l->state = LINK_RESETTING;
493 	__skb_queue_head_init(&l->transmq);
494 	__skb_queue_head_init(&l->backlogq);
495 	__skb_queue_head_init(&l->deferdq);
496 	__skb_queue_head_init(&l->failover_deferdq);
497 	skb_queue_head_init(&l->wakeupq);
498 	skb_queue_head_init(l->inputq);
499 	return true;
500 }
501 
502 /**
503  * tipc_link_bc_create - create new link to be used for broadcast
504  * @n: pointer to associated node
505  * @mtu: mtu to be used initially if no peers
506  * @window: send window to be used
507  * @inputq: queue to put messages ready for delivery
508  * @namedq: queue to put binding table update messages ready for delivery
509  * @link: return value, pointer to put the created link
510  *
511  * Returns true if link was created, otherwise false
512  */
513 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
514 			 int mtu, int window, u16 peer_caps,
515 			 struct sk_buff_head *inputq,
516 			 struct sk_buff_head *namedq,
517 			 struct tipc_link *bc_sndlink,
518 			 struct tipc_link **link)
519 {
520 	struct tipc_link *l;
521 
522 	if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
523 			      0, ownnode, peer, NULL, peer_caps, bc_sndlink,
524 			      NULL, inputq, namedq, link))
525 		return false;
526 
527 	l = *link;
528 	strcpy(l->name, tipc_bclink_name);
529 	trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
530 	tipc_link_reset(l);
531 	l->state = LINK_RESET;
532 	l->ackers = 0;
533 	l->bc_rcvlink = l;
534 
535 	/* Broadcast send link is always up */
536 	if (link_is_bc_sndlink(l))
537 		l->state = LINK_ESTABLISHED;
538 
539 	/* Disable replicast if even a single peer doesn't support it */
540 	if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
541 		tipc_bcast_disable_rcast(net);
542 
543 	return true;
544 }
545 
546 /**
547  * tipc_link_fsm_evt - link finite state machine
548  * @l: pointer to link
549  * @evt: state machine event to be processed
550  */
551 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
552 {
553 	int rc = 0;
554 	int old_state = l->state;
555 
556 	switch (l->state) {
557 	case LINK_RESETTING:
558 		switch (evt) {
559 		case LINK_PEER_RESET_EVT:
560 			l->state = LINK_PEER_RESET;
561 			break;
562 		case LINK_RESET_EVT:
563 			l->state = LINK_RESET;
564 			break;
565 		case LINK_FAILURE_EVT:
566 		case LINK_FAILOVER_BEGIN_EVT:
567 		case LINK_ESTABLISH_EVT:
568 		case LINK_FAILOVER_END_EVT:
569 		case LINK_SYNCH_BEGIN_EVT:
570 		case LINK_SYNCH_END_EVT:
571 		default:
572 			goto illegal_evt;
573 		}
574 		break;
575 	case LINK_RESET:
576 		switch (evt) {
577 		case LINK_PEER_RESET_EVT:
578 			l->state = LINK_ESTABLISHING;
579 			break;
580 		case LINK_FAILOVER_BEGIN_EVT:
581 			l->state = LINK_FAILINGOVER;
582 		case LINK_FAILURE_EVT:
583 		case LINK_RESET_EVT:
584 		case LINK_ESTABLISH_EVT:
585 		case LINK_FAILOVER_END_EVT:
586 			break;
587 		case LINK_SYNCH_BEGIN_EVT:
588 		case LINK_SYNCH_END_EVT:
589 		default:
590 			goto illegal_evt;
591 		}
592 		break;
593 	case LINK_PEER_RESET:
594 		switch (evt) {
595 		case LINK_RESET_EVT:
596 			l->state = LINK_ESTABLISHING;
597 			break;
598 		case LINK_PEER_RESET_EVT:
599 		case LINK_ESTABLISH_EVT:
600 		case LINK_FAILURE_EVT:
601 			break;
602 		case LINK_SYNCH_BEGIN_EVT:
603 		case LINK_SYNCH_END_EVT:
604 		case LINK_FAILOVER_BEGIN_EVT:
605 		case LINK_FAILOVER_END_EVT:
606 		default:
607 			goto illegal_evt;
608 		}
609 		break;
610 	case LINK_FAILINGOVER:
611 		switch (evt) {
612 		case LINK_FAILOVER_END_EVT:
613 			l->state = LINK_RESET;
614 			break;
615 		case LINK_PEER_RESET_EVT:
616 		case LINK_RESET_EVT:
617 		case LINK_ESTABLISH_EVT:
618 		case LINK_FAILURE_EVT:
619 			break;
620 		case LINK_FAILOVER_BEGIN_EVT:
621 		case LINK_SYNCH_BEGIN_EVT:
622 		case LINK_SYNCH_END_EVT:
623 		default:
624 			goto illegal_evt;
625 		}
626 		break;
627 	case LINK_ESTABLISHING:
628 		switch (evt) {
629 		case LINK_ESTABLISH_EVT:
630 			l->state = LINK_ESTABLISHED;
631 			break;
632 		case LINK_FAILOVER_BEGIN_EVT:
633 			l->state = LINK_FAILINGOVER;
634 			break;
635 		case LINK_RESET_EVT:
636 			l->state = LINK_RESET;
637 			break;
638 		case LINK_FAILURE_EVT:
639 		case LINK_PEER_RESET_EVT:
640 		case LINK_SYNCH_BEGIN_EVT:
641 		case LINK_FAILOVER_END_EVT:
642 			break;
643 		case LINK_SYNCH_END_EVT:
644 		default:
645 			goto illegal_evt;
646 		}
647 		break;
648 	case LINK_ESTABLISHED:
649 		switch (evt) {
650 		case LINK_PEER_RESET_EVT:
651 			l->state = LINK_PEER_RESET;
652 			rc |= TIPC_LINK_DOWN_EVT;
653 			break;
654 		case LINK_FAILURE_EVT:
655 			l->state = LINK_RESETTING;
656 			rc |= TIPC_LINK_DOWN_EVT;
657 			break;
658 		case LINK_RESET_EVT:
659 			l->state = LINK_RESET;
660 			break;
661 		case LINK_ESTABLISH_EVT:
662 		case LINK_SYNCH_END_EVT:
663 			break;
664 		case LINK_SYNCH_BEGIN_EVT:
665 			l->state = LINK_SYNCHING;
666 			break;
667 		case LINK_FAILOVER_BEGIN_EVT:
668 		case LINK_FAILOVER_END_EVT:
669 		default:
670 			goto illegal_evt;
671 		}
672 		break;
673 	case LINK_SYNCHING:
674 		switch (evt) {
675 		case LINK_PEER_RESET_EVT:
676 			l->state = LINK_PEER_RESET;
677 			rc |= TIPC_LINK_DOWN_EVT;
678 			break;
679 		case LINK_FAILURE_EVT:
680 			l->state = LINK_RESETTING;
681 			rc |= TIPC_LINK_DOWN_EVT;
682 			break;
683 		case LINK_RESET_EVT:
684 			l->state = LINK_RESET;
685 			break;
686 		case LINK_ESTABLISH_EVT:
687 		case LINK_SYNCH_BEGIN_EVT:
688 			break;
689 		case LINK_SYNCH_END_EVT:
690 			l->state = LINK_ESTABLISHED;
691 			break;
692 		case LINK_FAILOVER_BEGIN_EVT:
693 		case LINK_FAILOVER_END_EVT:
694 		default:
695 			goto illegal_evt;
696 		}
697 		break;
698 	default:
699 		pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
700 	}
701 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
702 	return rc;
703 illegal_evt:
704 	pr_err("Illegal FSM event %x in state %x on link %s\n",
705 	       evt, l->state, l->name);
706 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
707 	return rc;
708 }
709 
710 /* link_profile_stats - update statistical profiling of traffic
711  */
712 static void link_profile_stats(struct tipc_link *l)
713 {
714 	struct sk_buff *skb;
715 	struct tipc_msg *msg;
716 	int length;
717 
718 	/* Update counters used in statistical profiling of send traffic */
719 	l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
720 	l->stats.queue_sz_counts++;
721 
722 	skb = skb_peek(&l->transmq);
723 	if (!skb)
724 		return;
725 	msg = buf_msg(skb);
726 	length = msg_size(msg);
727 
728 	if (msg_user(msg) == MSG_FRAGMENTER) {
729 		if (msg_type(msg) != FIRST_FRAGMENT)
730 			return;
731 		length = msg_size(msg_inner_hdr(msg));
732 	}
733 	l->stats.msg_lengths_total += length;
734 	l->stats.msg_length_counts++;
735 	if (length <= 64)
736 		l->stats.msg_length_profile[0]++;
737 	else if (length <= 256)
738 		l->stats.msg_length_profile[1]++;
739 	else if (length <= 1024)
740 		l->stats.msg_length_profile[2]++;
741 	else if (length <= 4096)
742 		l->stats.msg_length_profile[3]++;
743 	else if (length <= 16384)
744 		l->stats.msg_length_profile[4]++;
745 	else if (length <= 32768)
746 		l->stats.msg_length_profile[5]++;
747 	else
748 		l->stats.msg_length_profile[6]++;
749 }
750 
751 /**
752  * tipc_link_too_silent - check if link is "too silent"
753  * @l: tipc link to be checked
754  *
755  * Returns true if the link 'silent_intv_cnt' is about to reach the
756  * 'abort_limit' value, otherwise false
757  */
758 bool tipc_link_too_silent(struct tipc_link *l)
759 {
760 	return (l->silent_intv_cnt + 2 > l->abort_limit);
761 }
762 
763 /* tipc_link_timeout - perform periodic task as instructed from node timeout
764  */
765 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
766 {
767 	int mtyp = 0;
768 	int rc = 0;
769 	bool state = false;
770 	bool probe = false;
771 	bool setup = false;
772 	u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
773 	u16 bc_acked = l->bc_rcvlink->acked;
774 	struct tipc_mon_state *mstate = &l->mon_state;
775 
776 	trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
777 	trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
778 	switch (l->state) {
779 	case LINK_ESTABLISHED:
780 	case LINK_SYNCHING:
781 		mtyp = STATE_MSG;
782 		link_profile_stats(l);
783 		tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
784 		if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
785 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
786 		state = bc_acked != bc_snt;
787 		state |= l->bc_rcvlink->rcv_unacked;
788 		state |= l->rcv_unacked;
789 		state |= !skb_queue_empty(&l->transmq);
790 		state |= !skb_queue_empty(&l->deferdq);
791 		probe = mstate->probing;
792 		probe |= l->silent_intv_cnt;
793 		if (probe || mstate->monitoring)
794 			l->silent_intv_cnt++;
795 		break;
796 	case LINK_RESET:
797 		setup = l->rst_cnt++ <= 4;
798 		setup |= !(l->rst_cnt % 16);
799 		mtyp = RESET_MSG;
800 		break;
801 	case LINK_ESTABLISHING:
802 		setup = true;
803 		mtyp = ACTIVATE_MSG;
804 		break;
805 	case LINK_PEER_RESET:
806 	case LINK_RESETTING:
807 	case LINK_FAILINGOVER:
808 		break;
809 	default:
810 		break;
811 	}
812 
813 	if (state || probe || setup)
814 		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
815 
816 	return rc;
817 }
818 
819 /**
820  * link_schedule_user - schedule a message sender for wakeup after congestion
821  * @l: congested link
822  * @hdr: header of message that is being sent
823  * Create pseudo msg to send back to user when congestion abates
824  */
825 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
826 {
827 	u32 dnode = tipc_own_addr(l->net);
828 	u32 dport = msg_origport(hdr);
829 	struct sk_buff *skb;
830 
831 	/* Create and schedule wakeup pseudo message */
832 	skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
833 			      dnode, l->addr, dport, 0, 0);
834 	if (!skb)
835 		return -ENOBUFS;
836 	msg_set_dest_droppable(buf_msg(skb), true);
837 	TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
838 	skb_queue_tail(&l->wakeupq, skb);
839 	l->stats.link_congs++;
840 	trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
841 	return -ELINKCONG;
842 }
843 
844 /**
845  * link_prepare_wakeup - prepare users for wakeup after congestion
846  * @l: congested link
847  * Wake up a number of waiting users, as permitted by available space
848  * in the send queue
849  */
850 static void link_prepare_wakeup(struct tipc_link *l)
851 {
852 	struct sk_buff *skb, *tmp;
853 	int imp, i = 0;
854 
855 	skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
856 		imp = TIPC_SKB_CB(skb)->chain_imp;
857 		if (l->backlog[imp].len < l->backlog[imp].limit) {
858 			skb_unlink(skb, &l->wakeupq);
859 			skb_queue_tail(l->inputq, skb);
860 		} else if (i++ > 10) {
861 			break;
862 		}
863 	}
864 }
865 
866 void tipc_link_reset(struct tipc_link *l)
867 {
868 	struct sk_buff_head list;
869 
870 	__skb_queue_head_init(&list);
871 
872 	l->in_session = false;
873 	/* Force re-synch of peer session number before establishing */
874 	l->peer_session--;
875 	l->session++;
876 	l->mtu = l->advertised_mtu;
877 
878 	spin_lock_bh(&l->wakeupq.lock);
879 	skb_queue_splice_init(&l->wakeupq, &list);
880 	spin_unlock_bh(&l->wakeupq.lock);
881 
882 	spin_lock_bh(&l->inputq->lock);
883 	skb_queue_splice_init(&list, l->inputq);
884 	spin_unlock_bh(&l->inputq->lock);
885 
886 	__skb_queue_purge(&l->transmq);
887 	__skb_queue_purge(&l->deferdq);
888 	__skb_queue_purge(&l->backlogq);
889 	__skb_queue_purge(&l->failover_deferdq);
890 	l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
891 	l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
892 	l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
893 	l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
894 	l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
895 	kfree_skb(l->reasm_buf);
896 	kfree_skb(l->failover_reasm_skb);
897 	l->reasm_buf = NULL;
898 	l->failover_reasm_skb = NULL;
899 	l->rcv_unacked = 0;
900 	l->snd_nxt = 1;
901 	l->rcv_nxt = 1;
902 	l->snd_nxt_state = 1;
903 	l->rcv_nxt_state = 1;
904 	l->acked = 0;
905 	l->silent_intv_cnt = 0;
906 	l->rst_cnt = 0;
907 	l->bc_peer_is_up = false;
908 	memset(&l->mon_state, 0, sizeof(l->mon_state));
909 	tipc_link_reset_stats(l);
910 }
911 
912 /**
913  * tipc_link_xmit(): enqueue buffer list according to queue situation
914  * @link: link to use
915  * @list: chain of buffers containing message
916  * @xmitq: returned list of packets to be sent by caller
917  *
918  * Consumes the buffer chain.
919  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
920  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
921  */
922 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
923 		   struct sk_buff_head *xmitq)
924 {
925 	struct tipc_msg *hdr = buf_msg(skb_peek(list));
926 	unsigned int maxwin = l->window;
927 	int imp = msg_importance(hdr);
928 	unsigned int mtu = l->mtu;
929 	u16 ack = l->rcv_nxt - 1;
930 	u16 seqno = l->snd_nxt;
931 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
932 	struct sk_buff_head *transmq = &l->transmq;
933 	struct sk_buff_head *backlogq = &l->backlogq;
934 	struct sk_buff *skb, *_skb, *bskb;
935 	int pkt_cnt = skb_queue_len(list);
936 	int rc = 0;
937 
938 	if (unlikely(msg_size(hdr) > mtu)) {
939 		skb_queue_purge(list);
940 		return -EMSGSIZE;
941 	}
942 
943 	/* Allow oversubscription of one data msg per source at congestion */
944 	if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
945 		if (imp == TIPC_SYSTEM_IMPORTANCE) {
946 			pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
947 			return -ENOBUFS;
948 		}
949 		rc = link_schedule_user(l, hdr);
950 	}
951 
952 	if (pkt_cnt > 1) {
953 		l->stats.sent_fragmented++;
954 		l->stats.sent_fragments += pkt_cnt;
955 	}
956 
957 	/* Prepare each packet for sending, and add to relevant queue: */
958 	while (skb_queue_len(list)) {
959 		skb = skb_peek(list);
960 		hdr = buf_msg(skb);
961 		msg_set_seqno(hdr, seqno);
962 		msg_set_ack(hdr, ack);
963 		msg_set_bcast_ack(hdr, bc_ack);
964 
965 		if (likely(skb_queue_len(transmq) < maxwin)) {
966 			_skb = skb_clone(skb, GFP_ATOMIC);
967 			if (!_skb) {
968 				skb_queue_purge(list);
969 				return -ENOBUFS;
970 			}
971 			__skb_dequeue(list);
972 			__skb_queue_tail(transmq, skb);
973 			/* next retransmit attempt */
974 			if (link_is_bc_sndlink(l))
975 				TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
976 			__skb_queue_tail(xmitq, _skb);
977 			TIPC_SKB_CB(skb)->ackers = l->ackers;
978 			l->rcv_unacked = 0;
979 			l->stats.sent_pkts++;
980 			seqno++;
981 			continue;
982 		}
983 		if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
984 			kfree_skb(__skb_dequeue(list));
985 			l->stats.sent_bundled++;
986 			continue;
987 		}
988 		if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
989 			kfree_skb(__skb_dequeue(list));
990 			__skb_queue_tail(backlogq, bskb);
991 			l->backlog[msg_importance(buf_msg(bskb))].len++;
992 			l->stats.sent_bundled++;
993 			l->stats.sent_bundles++;
994 			continue;
995 		}
996 		l->backlog[imp].len += skb_queue_len(list);
997 		skb_queue_splice_tail_init(list, backlogq);
998 	}
999 	l->snd_nxt = seqno;
1000 	return rc;
1001 }
1002 
1003 static void tipc_link_advance_backlog(struct tipc_link *l,
1004 				      struct sk_buff_head *xmitq)
1005 {
1006 	struct sk_buff *skb, *_skb;
1007 	struct tipc_msg *hdr;
1008 	u16 seqno = l->snd_nxt;
1009 	u16 ack = l->rcv_nxt - 1;
1010 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1011 
1012 	while (skb_queue_len(&l->transmq) < l->window) {
1013 		skb = skb_peek(&l->backlogq);
1014 		if (!skb)
1015 			break;
1016 		_skb = skb_clone(skb, GFP_ATOMIC);
1017 		if (!_skb)
1018 			break;
1019 		__skb_dequeue(&l->backlogq);
1020 		hdr = buf_msg(skb);
1021 		l->backlog[msg_importance(hdr)].len--;
1022 		__skb_queue_tail(&l->transmq, skb);
1023 		/* next retransmit attempt */
1024 		if (link_is_bc_sndlink(l))
1025 			TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1026 
1027 		__skb_queue_tail(xmitq, _skb);
1028 		TIPC_SKB_CB(skb)->ackers = l->ackers;
1029 		msg_set_seqno(hdr, seqno);
1030 		msg_set_ack(hdr, ack);
1031 		msg_set_bcast_ack(hdr, bc_ack);
1032 		l->rcv_unacked = 0;
1033 		l->stats.sent_pkts++;
1034 		seqno++;
1035 	}
1036 	l->snd_nxt = seqno;
1037 }
1038 
1039 /**
1040  * link_retransmit_failure() - Detect repeated retransmit failures
1041  * @l: tipc link sender
1042  * @r: tipc link receiver (= l in case of unicast)
1043  * @rc: returned code
1044  *
1045  * Return: true if the repeated retransmit failures happens, otherwise
1046  * false
1047  */
1048 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1049 				    int *rc)
1050 {
1051 	struct sk_buff *skb = skb_peek(&l->transmq);
1052 	struct tipc_msg *hdr;
1053 
1054 	if (!skb)
1055 		return false;
1056 
1057 	if (!TIPC_SKB_CB(skb)->retr_cnt)
1058 		return false;
1059 
1060 	if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1061 			msecs_to_jiffies(r->tolerance)))
1062 		return false;
1063 
1064 	hdr = buf_msg(skb);
1065 	if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1066 		return false;
1067 
1068 	pr_warn("Retransmission failure on link <%s>\n", l->name);
1069 	link_print(l, "State of link ");
1070 	pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1071 		msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1072 	pr_info("sqno %u, prev: %x, dest: %x\n",
1073 		msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1074 	pr_info("retr_stamp %d, retr_cnt %d\n",
1075 		jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1076 		TIPC_SKB_CB(skb)->retr_cnt);
1077 
1078 	trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1079 	trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1080 	trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1081 
1082 	if (link_is_bc_sndlink(l)) {
1083 		r->state = LINK_RESET;
1084 		*rc = TIPC_LINK_DOWN_EVT;
1085 	} else {
1086 		*rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1087 	}
1088 
1089 	return true;
1090 }
1091 
1092 /* tipc_link_bc_retrans() - retransmit zero or more packets
1093  * @l: the link to transmit on
1094  * @r: the receiving link ordering the retransmit. Same as l if unicast
1095  * @from: retransmit from (inclusive) this sequence number
1096  * @to: retransmit to (inclusive) this sequence number
1097  * xmitq: queue for accumulating the retransmitted packets
1098  */
1099 static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r,
1100 				u16 from, u16 to, struct sk_buff_head *xmitq)
1101 {
1102 	struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1103 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1104 	u16 ack = l->rcv_nxt - 1;
1105 	struct tipc_msg *hdr;
1106 	int rc = 0;
1107 
1108 	if (!skb)
1109 		return 0;
1110 	if (less(to, from))
1111 		return 0;
1112 
1113 	trace_tipc_link_retrans(r, from, to, &l->transmq);
1114 
1115 	if (link_retransmit_failure(l, r, &rc))
1116 		return rc;
1117 
1118 	skb_queue_walk(&l->transmq, skb) {
1119 		hdr = buf_msg(skb);
1120 		if (less(msg_seqno(hdr), from))
1121 			continue;
1122 		if (more(msg_seqno(hdr), to))
1123 			break;
1124 
1125 		if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1126 			continue;
1127 		TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1128 		_skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE, GFP_ATOMIC);
1129 		if (!_skb)
1130 			return 0;
1131 		hdr = buf_msg(_skb);
1132 		msg_set_ack(hdr, ack);
1133 		msg_set_bcast_ack(hdr, bc_ack);
1134 		_skb->priority = TC_PRIO_CONTROL;
1135 		__skb_queue_tail(xmitq, _skb);
1136 		l->stats.retransmitted++;
1137 
1138 		/* Increase actual retrans counter & mark first time */
1139 		if (!TIPC_SKB_CB(skb)->retr_cnt++)
1140 			TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1141 	}
1142 	return 0;
1143 }
1144 
1145 /* tipc_data_input - deliver data and name distr msgs to upper layer
1146  *
1147  * Consumes buffer if message is of right type
1148  * Node lock must be held
1149  */
1150 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1151 			    struct sk_buff_head *inputq)
1152 {
1153 	struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1154 	struct tipc_msg *hdr = buf_msg(skb);
1155 
1156 	switch (msg_user(hdr)) {
1157 	case TIPC_LOW_IMPORTANCE:
1158 	case TIPC_MEDIUM_IMPORTANCE:
1159 	case TIPC_HIGH_IMPORTANCE:
1160 	case TIPC_CRITICAL_IMPORTANCE:
1161 		if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1162 			skb_queue_tail(mc_inputq, skb);
1163 			return true;
1164 		}
1165 		/* fall through */
1166 	case CONN_MANAGER:
1167 		skb_queue_tail(inputq, skb);
1168 		return true;
1169 	case GROUP_PROTOCOL:
1170 		skb_queue_tail(mc_inputq, skb);
1171 		return true;
1172 	case NAME_DISTRIBUTOR:
1173 		l->bc_rcvlink->state = LINK_ESTABLISHED;
1174 		skb_queue_tail(l->namedq, skb);
1175 		return true;
1176 	case MSG_BUNDLER:
1177 	case TUNNEL_PROTOCOL:
1178 	case MSG_FRAGMENTER:
1179 	case BCAST_PROTOCOL:
1180 		return false;
1181 	default:
1182 		pr_warn("Dropping received illegal msg type\n");
1183 		kfree_skb(skb);
1184 		return true;
1185 	};
1186 }
1187 
1188 /* tipc_link_input - process packet that has passed link protocol check
1189  *
1190  * Consumes buffer
1191  */
1192 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1193 			   struct sk_buff_head *inputq,
1194 			   struct sk_buff **reasm_skb)
1195 {
1196 	struct tipc_msg *hdr = buf_msg(skb);
1197 	struct sk_buff *iskb;
1198 	struct sk_buff_head tmpq;
1199 	int usr = msg_user(hdr);
1200 	int pos = 0;
1201 
1202 	if (usr == MSG_BUNDLER) {
1203 		skb_queue_head_init(&tmpq);
1204 		l->stats.recv_bundles++;
1205 		l->stats.recv_bundled += msg_msgcnt(hdr);
1206 		while (tipc_msg_extract(skb, &iskb, &pos))
1207 			tipc_data_input(l, iskb, &tmpq);
1208 		tipc_skb_queue_splice_tail(&tmpq, inputq);
1209 		return 0;
1210 	} else if (usr == MSG_FRAGMENTER) {
1211 		l->stats.recv_fragments++;
1212 		if (tipc_buf_append(reasm_skb, &skb)) {
1213 			l->stats.recv_fragmented++;
1214 			tipc_data_input(l, skb, inputq);
1215 		} else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1216 			pr_warn_ratelimited("Unable to build fragment list\n");
1217 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1218 		}
1219 		return 0;
1220 	} else if (usr == BCAST_PROTOCOL) {
1221 		tipc_bcast_lock(l->net);
1222 		tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1223 		tipc_bcast_unlock(l->net);
1224 	}
1225 
1226 	kfree_skb(skb);
1227 	return 0;
1228 }
1229 
1230 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1231  *			 inner message along with the ones in the old link's
1232  *			 deferdq
1233  * @l: tunnel link
1234  * @skb: TUNNEL_PROTOCOL message
1235  * @inputq: queue to put messages ready for delivery
1236  */
1237 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1238 			     struct sk_buff_head *inputq)
1239 {
1240 	struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1241 	struct sk_buff_head *fdefq = &l->failover_deferdq;
1242 	struct tipc_msg *hdr = buf_msg(skb);
1243 	struct sk_buff *iskb;
1244 	int ipos = 0;
1245 	int rc = 0;
1246 	u16 seqno;
1247 
1248 	/* SYNCH_MSG */
1249 	if (msg_type(hdr) == SYNCH_MSG)
1250 		goto drop;
1251 
1252 	/* FAILOVER_MSG */
1253 	if (!tipc_msg_extract(skb, &iskb, &ipos)) {
1254 		pr_warn_ratelimited("Cannot extract FAILOVER_MSG, defq: %d\n",
1255 				    skb_queue_len(fdefq));
1256 		return rc;
1257 	}
1258 
1259 	do {
1260 		seqno = buf_seqno(iskb);
1261 
1262 		if (unlikely(less(seqno, l->drop_point))) {
1263 			kfree_skb(iskb);
1264 			continue;
1265 		}
1266 
1267 		if (unlikely(seqno != l->drop_point)) {
1268 			__tipc_skb_queue_sorted(fdefq, seqno, iskb);
1269 			continue;
1270 		}
1271 
1272 		l->drop_point++;
1273 
1274 		if (!tipc_data_input(l, iskb, inputq))
1275 			rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1276 		if (unlikely(rc))
1277 			break;
1278 	} while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1279 
1280 drop:
1281 	kfree_skb(skb);
1282 	return rc;
1283 }
1284 
1285 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1286 {
1287 	bool released = false;
1288 	struct sk_buff *skb, *tmp;
1289 
1290 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1291 		if (more(buf_seqno(skb), acked))
1292 			break;
1293 		__skb_unlink(skb, &l->transmq);
1294 		kfree_skb(skb);
1295 		released = true;
1296 	}
1297 	return released;
1298 }
1299 
1300 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1301  * @l: tipc link that data have come with gaps in sequence if any
1302  * @data: data buffer to store the Gap ACK blocks after built
1303  *
1304  * returns the actual allocated memory size
1305  */
1306 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data)
1307 {
1308 	struct sk_buff *skb = skb_peek(&l->deferdq);
1309 	struct tipc_gap_ack_blks *ga = data;
1310 	u16 len, expect, seqno = 0;
1311 	u8 n = 0;
1312 
1313 	if (!skb)
1314 		goto exit;
1315 
1316 	expect = buf_seqno(skb);
1317 	skb_queue_walk(&l->deferdq, skb) {
1318 		seqno = buf_seqno(skb);
1319 		if (unlikely(more(seqno, expect))) {
1320 			ga->gacks[n].ack = htons(expect - 1);
1321 			ga->gacks[n].gap = htons(seqno - expect);
1322 			if (++n >= MAX_GAP_ACK_BLKS) {
1323 				pr_info_ratelimited("Too few Gap ACK blocks!\n");
1324 				goto exit;
1325 			}
1326 		} else if (unlikely(less(seqno, expect))) {
1327 			pr_warn("Unexpected skb in deferdq!\n");
1328 			continue;
1329 		}
1330 		expect = seqno + 1;
1331 	}
1332 
1333 	/* last block */
1334 	ga->gacks[n].ack = htons(seqno);
1335 	ga->gacks[n].gap = 0;
1336 	n++;
1337 
1338 exit:
1339 	len = tipc_gap_ack_blks_sz(n);
1340 	ga->len = htons(len);
1341 	ga->gack_cnt = n;
1342 	return len;
1343 }
1344 
1345 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1346  *			       acked packets, also doing retransmissions if
1347  *			       gaps found
1348  * @l: tipc link with transmq queue to be advanced
1349  * @acked: seqno of last packet acked by peer without any gaps before
1350  * @gap: # of gap packets
1351  * @ga: buffer pointer to Gap ACK blocks from peer
1352  * @xmitq: queue for accumulating the retransmitted packets if any
1353  *
1354  * In case of a repeated retransmit failures, the call will return shortly
1355  * with a returned code (e.g. TIPC_LINK_DOWN_EVT)
1356  */
1357 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
1358 				     struct tipc_gap_ack_blks *ga,
1359 				     struct sk_buff_head *xmitq)
1360 {
1361 	struct sk_buff *skb, *_skb, *tmp;
1362 	struct tipc_msg *hdr;
1363 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1364 	u16 ack = l->rcv_nxt - 1;
1365 	bool passed = false;
1366 	u16 seqno, n = 0;
1367 	int rc = 0;
1368 
1369 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1370 		seqno = buf_seqno(skb);
1371 
1372 next_gap_ack:
1373 		if (less_eq(seqno, acked)) {
1374 			/* release skb */
1375 			__skb_unlink(skb, &l->transmq);
1376 			kfree_skb(skb);
1377 		} else if (less_eq(seqno, acked + gap)) {
1378 			/* First, check if repeated retrans failures occurs? */
1379 			if (!passed && link_retransmit_failure(l, l, &rc))
1380 				return rc;
1381 			passed = true;
1382 
1383 			/* retransmit skb if unrestricted*/
1384 			if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1385 				continue;
1386 			TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
1387 			_skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE,
1388 					   GFP_ATOMIC);
1389 			if (!_skb)
1390 				continue;
1391 			hdr = buf_msg(_skb);
1392 			msg_set_ack(hdr, ack);
1393 			msg_set_bcast_ack(hdr, bc_ack);
1394 			_skb->priority = TC_PRIO_CONTROL;
1395 			__skb_queue_tail(xmitq, _skb);
1396 			l->stats.retransmitted++;
1397 
1398 			/* Increase actual retrans counter & mark first time */
1399 			if (!TIPC_SKB_CB(skb)->retr_cnt++)
1400 				TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1401 		} else {
1402 			/* retry with Gap ACK blocks if any */
1403 			if (!ga || n >= ga->gack_cnt)
1404 				break;
1405 			acked = ntohs(ga->gacks[n].ack);
1406 			gap = ntohs(ga->gacks[n].gap);
1407 			n++;
1408 			goto next_gap_ack;
1409 		}
1410 	}
1411 
1412 	return 0;
1413 }
1414 
1415 /* tipc_link_build_state_msg: prepare link state message for transmission
1416  *
1417  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1418  * risk of ack storms towards the sender
1419  */
1420 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1421 {
1422 	if (!l)
1423 		return 0;
1424 
1425 	/* Broadcast ACK must be sent via a unicast link => defer to caller */
1426 	if (link_is_bc_rcvlink(l)) {
1427 		if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1428 			return 0;
1429 		l->rcv_unacked = 0;
1430 
1431 		/* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1432 		l->snd_nxt = l->rcv_nxt;
1433 		return TIPC_LINK_SND_STATE;
1434 	}
1435 
1436 	/* Unicast ACK */
1437 	l->rcv_unacked = 0;
1438 	l->stats.sent_acks++;
1439 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1440 	return 0;
1441 }
1442 
1443 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1444  */
1445 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1446 {
1447 	int mtyp = RESET_MSG;
1448 	struct sk_buff *skb;
1449 
1450 	if (l->state == LINK_ESTABLISHING)
1451 		mtyp = ACTIVATE_MSG;
1452 
1453 	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1454 
1455 	/* Inform peer that this endpoint is going down if applicable */
1456 	skb = skb_peek_tail(xmitq);
1457 	if (skb && (l->state == LINK_RESET))
1458 		msg_set_peer_stopping(buf_msg(skb), 1);
1459 }
1460 
1461 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1462  * Note that sending of broadcast NACK is coordinated among nodes, to
1463  * reduce the risk of NACK storms towards the sender
1464  */
1465 static int tipc_link_build_nack_msg(struct tipc_link *l,
1466 				    struct sk_buff_head *xmitq)
1467 {
1468 	u32 def_cnt = ++l->stats.deferred_recv;
1469 	u32 defq_len = skb_queue_len(&l->deferdq);
1470 	int match1, match2;
1471 
1472 	if (link_is_bc_rcvlink(l)) {
1473 		match1 = def_cnt & 0xf;
1474 		match2 = tipc_own_addr(l->net) & 0xf;
1475 		if (match1 == match2)
1476 			return TIPC_LINK_SND_STATE;
1477 		return 0;
1478 	}
1479 
1480 	if (defq_len >= 3 && !((defq_len - 3) % 16))
1481 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1482 	return 0;
1483 }
1484 
1485 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1486  * @l: the link that should handle the message
1487  * @skb: TIPC packet
1488  * @xmitq: queue to place packets to be sent after this call
1489  */
1490 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1491 		  struct sk_buff_head *xmitq)
1492 {
1493 	struct sk_buff_head *defq = &l->deferdq;
1494 	struct tipc_msg *hdr = buf_msg(skb);
1495 	u16 seqno, rcv_nxt, win_lim;
1496 	int rc = 0;
1497 
1498 	/* Verify and update link state */
1499 	if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1500 		return tipc_link_proto_rcv(l, skb, xmitq);
1501 
1502 	/* Don't send probe at next timeout expiration */
1503 	l->silent_intv_cnt = 0;
1504 
1505 	do {
1506 		hdr = buf_msg(skb);
1507 		seqno = msg_seqno(hdr);
1508 		rcv_nxt = l->rcv_nxt;
1509 		win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1510 
1511 		if (unlikely(!link_is_up(l))) {
1512 			if (l->state == LINK_ESTABLISHING)
1513 				rc = TIPC_LINK_UP_EVT;
1514 			goto drop;
1515 		}
1516 
1517 		/* Drop if outside receive window */
1518 		if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1519 			l->stats.duplicates++;
1520 			goto drop;
1521 		}
1522 
1523 		/* Forward queues and wake up waiting users */
1524 		if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1525 			tipc_link_advance_backlog(l, xmitq);
1526 			if (unlikely(!skb_queue_empty(&l->wakeupq)))
1527 				link_prepare_wakeup(l);
1528 		}
1529 
1530 		/* Defer delivery if sequence gap */
1531 		if (unlikely(seqno != rcv_nxt)) {
1532 			__tipc_skb_queue_sorted(defq, seqno, skb);
1533 			rc |= tipc_link_build_nack_msg(l, xmitq);
1534 			break;
1535 		}
1536 
1537 		/* Deliver packet */
1538 		l->rcv_nxt++;
1539 		l->stats.recv_pkts++;
1540 
1541 		if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1542 			rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1543 		else if (!tipc_data_input(l, skb, l->inputq))
1544 			rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1545 		if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1546 			rc |= tipc_link_build_state_msg(l, xmitq);
1547 		if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1548 			break;
1549 	} while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1550 
1551 	return rc;
1552 drop:
1553 	kfree_skb(skb);
1554 	return rc;
1555 }
1556 
1557 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1558 				      bool probe_reply, u16 rcvgap,
1559 				      int tolerance, int priority,
1560 				      struct sk_buff_head *xmitq)
1561 {
1562 	struct tipc_link *bcl = l->bc_rcvlink;
1563 	struct sk_buff *skb;
1564 	struct tipc_msg *hdr;
1565 	struct sk_buff_head *dfq = &l->deferdq;
1566 	bool node_up = link_is_up(bcl);
1567 	struct tipc_mon_state *mstate = &l->mon_state;
1568 	int dlen = 0;
1569 	void *data;
1570 	u16 glen = 0;
1571 
1572 	/* Don't send protocol message during reset or link failover */
1573 	if (tipc_link_is_blocked(l))
1574 		return;
1575 
1576 	if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1577 		return;
1578 
1579 	if (!skb_queue_empty(dfq))
1580 		rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1581 
1582 	skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1583 			      tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1584 			      l->addr, tipc_own_addr(l->net), 0, 0, 0);
1585 	if (!skb)
1586 		return;
1587 
1588 	hdr = buf_msg(skb);
1589 	data = msg_data(hdr);
1590 	msg_set_session(hdr, l->session);
1591 	msg_set_bearer_id(hdr, l->bearer_id);
1592 	msg_set_net_plane(hdr, l->net_plane);
1593 	msg_set_next_sent(hdr, l->snd_nxt);
1594 	msg_set_ack(hdr, l->rcv_nxt - 1);
1595 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1596 	msg_set_bc_ack_invalid(hdr, !node_up);
1597 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1598 	msg_set_link_tolerance(hdr, tolerance);
1599 	msg_set_linkprio(hdr, priority);
1600 	msg_set_redundant_link(hdr, node_up);
1601 	msg_set_seq_gap(hdr, 0);
1602 	msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1603 
1604 	if (mtyp == STATE_MSG) {
1605 		if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1606 			msg_set_seqno(hdr, l->snd_nxt_state++);
1607 		msg_set_seq_gap(hdr, rcvgap);
1608 		msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1609 		msg_set_probe(hdr, probe);
1610 		msg_set_is_keepalive(hdr, probe || probe_reply);
1611 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1612 			glen = tipc_build_gap_ack_blks(l, data);
1613 		tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1614 		msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1615 		skb_trim(skb, INT_H_SIZE + glen + dlen);
1616 		l->stats.sent_states++;
1617 		l->rcv_unacked = 0;
1618 	} else {
1619 		/* RESET_MSG or ACTIVATE_MSG */
1620 		if (mtyp == ACTIVATE_MSG) {
1621 			msg_set_dest_session_valid(hdr, 1);
1622 			msg_set_dest_session(hdr, l->peer_session);
1623 		}
1624 		msg_set_max_pkt(hdr, l->advertised_mtu);
1625 		strcpy(data, l->if_name);
1626 		msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1627 		skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1628 	}
1629 	if (probe)
1630 		l->stats.sent_probes++;
1631 	if (rcvgap)
1632 		l->stats.sent_nacks++;
1633 	skb->priority = TC_PRIO_CONTROL;
1634 	__skb_queue_tail(xmitq, skb);
1635 	trace_tipc_proto_build(skb, false, l->name);
1636 }
1637 
1638 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1639 				    struct sk_buff_head *xmitq)
1640 {
1641 	u32 onode = tipc_own_addr(l->net);
1642 	struct tipc_msg *hdr, *ihdr;
1643 	struct sk_buff_head tnlq;
1644 	struct sk_buff *skb;
1645 	u32 dnode = l->addr;
1646 
1647 	skb_queue_head_init(&tnlq);
1648 	skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1649 			      INT_H_SIZE, BASIC_H_SIZE,
1650 			      dnode, onode, 0, 0, 0);
1651 	if (!skb) {
1652 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1653 		return;
1654 	}
1655 
1656 	hdr = buf_msg(skb);
1657 	msg_set_msgcnt(hdr, 1);
1658 	msg_set_bearer_id(hdr, l->peer_bearer_id);
1659 
1660 	ihdr = (struct tipc_msg *)msg_data(hdr);
1661 	tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1662 		      BASIC_H_SIZE, dnode);
1663 	msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1664 	__skb_queue_tail(&tnlq, skb);
1665 	tipc_link_xmit(l, &tnlq, xmitq);
1666 }
1667 
1668 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1669  * with contents of the link's transmit and backlog queues.
1670  */
1671 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1672 			   int mtyp, struct sk_buff_head *xmitq)
1673 {
1674 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1675 	struct sk_buff *skb, *tnlskb;
1676 	struct tipc_msg *hdr, tnlhdr;
1677 	struct sk_buff_head *queue = &l->transmq;
1678 	struct sk_buff_head tmpxq, tnlq;
1679 	u16 pktlen, pktcnt, seqno = l->snd_nxt;
1680 
1681 	if (!tnl)
1682 		return;
1683 
1684 	skb_queue_head_init(&tnlq);
1685 	skb_queue_head_init(&tmpxq);
1686 
1687 	/* At least one packet required for safe algorithm => add dummy */
1688 	skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1689 			      BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1690 			      0, 0, TIPC_ERR_NO_PORT);
1691 	if (!skb) {
1692 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1693 		return;
1694 	}
1695 	skb_queue_tail(&tnlq, skb);
1696 	tipc_link_xmit(l, &tnlq, &tmpxq);
1697 	__skb_queue_purge(&tmpxq);
1698 
1699 	/* Initialize reusable tunnel packet header */
1700 	tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1701 		      mtyp, INT_H_SIZE, l->addr);
1702 	if (mtyp == SYNCH_MSG)
1703 		pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
1704 	else
1705 		pktcnt = skb_queue_len(&l->transmq);
1706 	pktcnt += skb_queue_len(&l->backlogq);
1707 	msg_set_msgcnt(&tnlhdr, pktcnt);
1708 	msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1709 tnl:
1710 	/* Wrap each packet into a tunnel packet */
1711 	skb_queue_walk(queue, skb) {
1712 		hdr = buf_msg(skb);
1713 		if (queue == &l->backlogq)
1714 			msg_set_seqno(hdr, seqno++);
1715 		pktlen = msg_size(hdr);
1716 		msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1717 		tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1718 		if (!tnlskb) {
1719 			pr_warn("%sunable to send packet\n", link_co_err);
1720 			return;
1721 		}
1722 		skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1723 		skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1724 		__skb_queue_tail(&tnlq, tnlskb);
1725 	}
1726 	if (queue != &l->backlogq) {
1727 		queue = &l->backlogq;
1728 		goto tnl;
1729 	}
1730 
1731 	tipc_link_xmit(tnl, &tnlq, xmitq);
1732 
1733 	if (mtyp == FAILOVER_MSG) {
1734 		tnl->drop_point = l->rcv_nxt;
1735 		tnl->failover_reasm_skb = l->reasm_buf;
1736 		l->reasm_buf = NULL;
1737 
1738 		/* Failover the link's deferdq */
1739 		if (unlikely(!skb_queue_empty(fdefq))) {
1740 			pr_warn("Link failover deferdq not empty: %d!\n",
1741 				skb_queue_len(fdefq));
1742 			__skb_queue_purge(fdefq);
1743 		}
1744 		skb_queue_splice_init(&l->deferdq, fdefq);
1745 	}
1746 }
1747 
1748 /**
1749  * tipc_link_failover_prepare() - prepare tnl for link failover
1750  *
1751  * This is a special version of the precursor - tipc_link_tnl_prepare(),
1752  * see the tipc_node_link_failover() for details
1753  *
1754  * @l: failover link
1755  * @tnl: tunnel link
1756  * @xmitq: queue for messages to be xmited
1757  */
1758 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
1759 				struct sk_buff_head *xmitq)
1760 {
1761 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1762 
1763 	tipc_link_create_dummy_tnl_msg(tnl, xmitq);
1764 
1765 	/* This failover link enpoint was never established before,
1766 	 * so it has not received anything from peer.
1767 	 * Otherwise, it must be a normal failover situation or the
1768 	 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
1769 	 * would have to start over from scratch instead.
1770 	 */
1771 	tnl->drop_point = 1;
1772 	tnl->failover_reasm_skb = NULL;
1773 
1774 	/* Initiate the link's failover deferdq */
1775 	if (unlikely(!skb_queue_empty(fdefq))) {
1776 		pr_warn("Link failover deferdq not empty: %d!\n",
1777 			skb_queue_len(fdefq));
1778 		__skb_queue_purge(fdefq);
1779 	}
1780 }
1781 
1782 /* tipc_link_validate_msg(): validate message against current link state
1783  * Returns true if message should be accepted, otherwise false
1784  */
1785 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
1786 {
1787 	u16 curr_session = l->peer_session;
1788 	u16 session = msg_session(hdr);
1789 	int mtyp = msg_type(hdr);
1790 
1791 	if (msg_user(hdr) != LINK_PROTOCOL)
1792 		return true;
1793 
1794 	switch (mtyp) {
1795 	case RESET_MSG:
1796 		if (!l->in_session)
1797 			return true;
1798 		/* Accept only RESET with new session number */
1799 		return more(session, curr_session);
1800 	case ACTIVATE_MSG:
1801 		if (!l->in_session)
1802 			return true;
1803 		/* Accept only ACTIVATE with new or current session number */
1804 		return !less(session, curr_session);
1805 	case STATE_MSG:
1806 		/* Accept only STATE with current session number */
1807 		if (!l->in_session)
1808 			return false;
1809 		if (session != curr_session)
1810 			return false;
1811 		/* Extra sanity check */
1812 		if (!link_is_up(l) && msg_ack(hdr))
1813 			return false;
1814 		if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
1815 			return true;
1816 		/* Accept only STATE with new sequence number */
1817 		return !less(msg_seqno(hdr), l->rcv_nxt_state);
1818 	default:
1819 		return false;
1820 	}
1821 }
1822 
1823 /* tipc_link_proto_rcv(): receive link level protocol message :
1824  * Note that network plane id propagates through the network, and may
1825  * change at any time. The node with lowest numerical id determines
1826  * network plane
1827  */
1828 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1829 			       struct sk_buff_head *xmitq)
1830 {
1831 	struct tipc_msg *hdr = buf_msg(skb);
1832 	struct tipc_gap_ack_blks *ga = NULL;
1833 	u16 rcvgap = 0;
1834 	u16 ack = msg_ack(hdr);
1835 	u16 gap = msg_seq_gap(hdr);
1836 	u16 peers_snd_nxt =  msg_next_sent(hdr);
1837 	u16 peers_tol = msg_link_tolerance(hdr);
1838 	u16 peers_prio = msg_linkprio(hdr);
1839 	u16 rcv_nxt = l->rcv_nxt;
1840 	u16 dlen = msg_data_sz(hdr);
1841 	int mtyp = msg_type(hdr);
1842 	bool reply = msg_probe(hdr);
1843 	u16 glen = 0;
1844 	void *data;
1845 	char *if_name;
1846 	int rc = 0;
1847 
1848 	trace_tipc_proto_rcv(skb, false, l->name);
1849 	if (tipc_link_is_blocked(l) || !xmitq)
1850 		goto exit;
1851 
1852 	if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1853 		l->net_plane = msg_net_plane(hdr);
1854 
1855 	skb_linearize(skb);
1856 	hdr = buf_msg(skb);
1857 	data = msg_data(hdr);
1858 
1859 	if (!tipc_link_validate_msg(l, hdr)) {
1860 		trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
1861 		trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
1862 		goto exit;
1863 	}
1864 
1865 	switch (mtyp) {
1866 	case RESET_MSG:
1867 	case ACTIVATE_MSG:
1868 		/* Complete own link name with peer's interface name */
1869 		if_name =  strrchr(l->name, ':') + 1;
1870 		if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1871 			break;
1872 		if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1873 			break;
1874 		strncpy(if_name, data, TIPC_MAX_IF_NAME);
1875 
1876 		/* Update own tolerance if peer indicates a non-zero value */
1877 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1878 			l->tolerance = peers_tol;
1879 			l->bc_rcvlink->tolerance = peers_tol;
1880 		}
1881 		/* Update own priority if peer's priority is higher */
1882 		if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1883 			l->priority = peers_prio;
1884 
1885 		/* If peer is going down we want full re-establish cycle */
1886 		if (msg_peer_stopping(hdr)) {
1887 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1888 			break;
1889 		}
1890 
1891 		/* If this endpoint was re-created while peer was ESTABLISHING
1892 		 * it doesn't know current session number. Force re-synch.
1893 		 */
1894 		if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
1895 		    l->session != msg_dest_session(hdr)) {
1896 			if (less(l->session, msg_dest_session(hdr)))
1897 				l->session = msg_dest_session(hdr) + 1;
1898 			break;
1899 		}
1900 
1901 		/* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1902 		if (mtyp == RESET_MSG || !link_is_up(l))
1903 			rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1904 
1905 		/* ACTIVATE_MSG takes up link if it was already locally reset */
1906 		if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
1907 			rc = TIPC_LINK_UP_EVT;
1908 
1909 		l->peer_session = msg_session(hdr);
1910 		l->in_session = true;
1911 		l->peer_bearer_id = msg_bearer_id(hdr);
1912 		if (l->mtu > msg_max_pkt(hdr))
1913 			l->mtu = msg_max_pkt(hdr);
1914 		break;
1915 
1916 	case STATE_MSG:
1917 		l->rcv_nxt_state = msg_seqno(hdr) + 1;
1918 
1919 		/* Update own tolerance if peer indicates a non-zero value */
1920 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1921 			l->tolerance = peers_tol;
1922 			l->bc_rcvlink->tolerance = peers_tol;
1923 		}
1924 		/* Update own prio if peer indicates a different value */
1925 		if ((peers_prio != l->priority) &&
1926 		    in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
1927 			l->priority = peers_prio;
1928 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1929 		}
1930 
1931 		l->silent_intv_cnt = 0;
1932 		l->stats.recv_states++;
1933 		if (msg_probe(hdr))
1934 			l->stats.recv_probes++;
1935 
1936 		if (!link_is_up(l)) {
1937 			if (l->state == LINK_ESTABLISHING)
1938 				rc = TIPC_LINK_UP_EVT;
1939 			break;
1940 		}
1941 
1942 		/* Receive Gap ACK blocks from peer if any */
1943 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
1944 			ga = (struct tipc_gap_ack_blks *)data;
1945 			glen = ntohs(ga->len);
1946 			/* sanity check: if failed, ignore Gap ACK blocks */
1947 			if (glen != tipc_gap_ack_blks_sz(ga->gack_cnt))
1948 				ga = NULL;
1949 		}
1950 
1951 		tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
1952 			     &l->mon_state, l->bearer_id);
1953 
1954 		/* Send NACK if peer has sent pkts we haven't received yet */
1955 		if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1956 			rcvgap = peers_snd_nxt - l->rcv_nxt;
1957 		if (rcvgap || reply)
1958 			tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
1959 						  rcvgap, 0, 0, xmitq);
1960 
1961 		rc |= tipc_link_advance_transmq(l, ack, gap, ga, xmitq);
1962 
1963 		/* If NACK, retransmit will now start at right position */
1964 		if (gap)
1965 			l->stats.recv_nacks++;
1966 
1967 		tipc_link_advance_backlog(l, xmitq);
1968 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
1969 			link_prepare_wakeup(l);
1970 	}
1971 exit:
1972 	kfree_skb(skb);
1973 	return rc;
1974 }
1975 
1976 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1977  */
1978 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1979 					 u16 peers_snd_nxt,
1980 					 struct sk_buff_head *xmitq)
1981 {
1982 	struct sk_buff *skb;
1983 	struct tipc_msg *hdr;
1984 	struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1985 	u16 ack = l->rcv_nxt - 1;
1986 	u16 gap_to = peers_snd_nxt - 1;
1987 
1988 	skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1989 			      0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1990 	if (!skb)
1991 		return false;
1992 	hdr = buf_msg(skb);
1993 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1994 	msg_set_bcast_ack(hdr, ack);
1995 	msg_set_bcgap_after(hdr, ack);
1996 	if (dfrd_skb)
1997 		gap_to = buf_seqno(dfrd_skb) - 1;
1998 	msg_set_bcgap_to(hdr, gap_to);
1999 	msg_set_non_seq(hdr, bcast);
2000 	__skb_queue_tail(xmitq, skb);
2001 	return true;
2002 }
2003 
2004 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2005  *
2006  * Give a newly added peer node the sequence number where it should
2007  * start receiving and acking broadcast packets.
2008  */
2009 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2010 					struct sk_buff_head *xmitq)
2011 {
2012 	struct sk_buff_head list;
2013 
2014 	__skb_queue_head_init(&list);
2015 	if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2016 		return;
2017 	msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2018 	tipc_link_xmit(l, &list, xmitq);
2019 }
2020 
2021 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2022  */
2023 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2024 {
2025 	int mtyp = msg_type(hdr);
2026 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2027 
2028 	if (link_is_up(l))
2029 		return;
2030 
2031 	if (msg_user(hdr) == BCAST_PROTOCOL) {
2032 		l->rcv_nxt = peers_snd_nxt;
2033 		l->state = LINK_ESTABLISHED;
2034 		return;
2035 	}
2036 
2037 	if (l->peer_caps & TIPC_BCAST_SYNCH)
2038 		return;
2039 
2040 	if (msg_peer_node_is_up(hdr))
2041 		return;
2042 
2043 	/* Compatibility: accept older, less safe initial synch data */
2044 	if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2045 		l->rcv_nxt = peers_snd_nxt;
2046 }
2047 
2048 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2049  */
2050 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2051 			  struct sk_buff_head *xmitq)
2052 {
2053 	struct tipc_link *snd_l = l->bc_sndlink;
2054 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2055 	u16 from = msg_bcast_ack(hdr) + 1;
2056 	u16 to = from + msg_bc_gap(hdr) - 1;
2057 	int rc = 0;
2058 
2059 	if (!link_is_up(l))
2060 		return rc;
2061 
2062 	if (!msg_peer_node_is_up(hdr))
2063 		return rc;
2064 
2065 	/* Open when peer ackowledges our bcast init msg (pkt #1) */
2066 	if (msg_ack(hdr))
2067 		l->bc_peer_is_up = true;
2068 
2069 	if (!l->bc_peer_is_up)
2070 		return rc;
2071 
2072 	l->stats.recv_nacks++;
2073 
2074 	/* Ignore if peers_snd_nxt goes beyond receive window */
2075 	if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2076 		return rc;
2077 
2078 	rc = tipc_link_bc_retrans(snd_l, l, from, to, xmitq);
2079 
2080 	l->snd_nxt = peers_snd_nxt;
2081 	if (link_bc_rcv_gap(l))
2082 		rc |= TIPC_LINK_SND_STATE;
2083 
2084 	/* Return now if sender supports nack via STATE messages */
2085 	if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2086 		return rc;
2087 
2088 	/* Otherwise, be backwards compatible */
2089 
2090 	if (!more(peers_snd_nxt, l->rcv_nxt)) {
2091 		l->nack_state = BC_NACK_SND_CONDITIONAL;
2092 		return 0;
2093 	}
2094 
2095 	/* Don't NACK if one was recently sent or peeked */
2096 	if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2097 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2098 		return 0;
2099 	}
2100 
2101 	/* Conditionally delay NACK sending until next synch rcv */
2102 	if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2103 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2104 		if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2105 			return 0;
2106 	}
2107 
2108 	/* Send NACK now but suppress next one */
2109 	tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2110 	l->nack_state = BC_NACK_SND_SUPPRESS;
2111 	return 0;
2112 }
2113 
2114 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
2115 			  struct sk_buff_head *xmitq)
2116 {
2117 	struct sk_buff *skb, *tmp;
2118 	struct tipc_link *snd_l = l->bc_sndlink;
2119 
2120 	if (!link_is_up(l) || !l->bc_peer_is_up)
2121 		return;
2122 
2123 	if (!more(acked, l->acked))
2124 		return;
2125 
2126 	trace_tipc_link_bc_ack(l, l->acked, acked, &snd_l->transmq);
2127 	/* Skip over packets peer has already acked */
2128 	skb_queue_walk(&snd_l->transmq, skb) {
2129 		if (more(buf_seqno(skb), l->acked))
2130 			break;
2131 	}
2132 
2133 	/* Update/release the packets peer is acking now */
2134 	skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
2135 		if (more(buf_seqno(skb), acked))
2136 			break;
2137 		if (!--TIPC_SKB_CB(skb)->ackers) {
2138 			__skb_unlink(skb, &snd_l->transmq);
2139 			kfree_skb(skb);
2140 		}
2141 	}
2142 	l->acked = acked;
2143 	tipc_link_advance_backlog(snd_l, xmitq);
2144 	if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
2145 		link_prepare_wakeup(snd_l);
2146 }
2147 
2148 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2149  * This function is here for backwards compatibility, since
2150  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2151  */
2152 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2153 			  struct sk_buff_head *xmitq)
2154 {
2155 	struct tipc_msg *hdr = buf_msg(skb);
2156 	u32 dnode = msg_destnode(hdr);
2157 	int mtyp = msg_type(hdr);
2158 	u16 acked = msg_bcast_ack(hdr);
2159 	u16 from = acked + 1;
2160 	u16 to = msg_bcgap_to(hdr);
2161 	u16 peers_snd_nxt = to + 1;
2162 	int rc = 0;
2163 
2164 	kfree_skb(skb);
2165 
2166 	if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2167 		return 0;
2168 
2169 	if (mtyp != STATE_MSG)
2170 		return 0;
2171 
2172 	if (dnode == tipc_own_addr(l->net)) {
2173 		tipc_link_bc_ack_rcv(l, acked, xmitq);
2174 		rc = tipc_link_bc_retrans(l->bc_sndlink, l, from, to, xmitq);
2175 		l->stats.recv_nacks++;
2176 		return rc;
2177 	}
2178 
2179 	/* Msg for other node => suppress own NACK at next sync if applicable */
2180 	if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2181 		l->nack_state = BC_NACK_SND_SUPPRESS;
2182 
2183 	return 0;
2184 }
2185 
2186 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
2187 {
2188 	int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2189 
2190 	l->window = win;
2191 	l->backlog[TIPC_LOW_IMPORTANCE].limit      = max_t(u16, 50, win);
2192 	l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = max_t(u16, 100, win * 2);
2193 	l->backlog[TIPC_HIGH_IMPORTANCE].limit     = max_t(u16, 150, win * 3);
2194 	l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
2195 	l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
2196 }
2197 
2198 /**
2199  * link_reset_stats - reset link statistics
2200  * @l: pointer to link
2201  */
2202 void tipc_link_reset_stats(struct tipc_link *l)
2203 {
2204 	memset(&l->stats, 0, sizeof(l->stats));
2205 }
2206 
2207 static void link_print(struct tipc_link *l, const char *str)
2208 {
2209 	struct sk_buff *hskb = skb_peek(&l->transmq);
2210 	u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2211 	u16 tail = l->snd_nxt - 1;
2212 
2213 	pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2214 	pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2215 		skb_queue_len(&l->transmq), head, tail,
2216 		skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2217 }
2218 
2219 /* Parse and validate nested (link) properties valid for media, bearer and link
2220  */
2221 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2222 {
2223 	int err;
2224 
2225 	err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2226 					  tipc_nl_prop_policy, NULL);
2227 	if (err)
2228 		return err;
2229 
2230 	if (props[TIPC_NLA_PROP_PRIO]) {
2231 		u32 prio;
2232 
2233 		prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2234 		if (prio > TIPC_MAX_LINK_PRI)
2235 			return -EINVAL;
2236 	}
2237 
2238 	if (props[TIPC_NLA_PROP_TOL]) {
2239 		u32 tol;
2240 
2241 		tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2242 		if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2243 			return -EINVAL;
2244 	}
2245 
2246 	if (props[TIPC_NLA_PROP_WIN]) {
2247 		u32 win;
2248 
2249 		win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2250 		if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
2251 			return -EINVAL;
2252 	}
2253 
2254 	return 0;
2255 }
2256 
2257 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2258 {
2259 	int i;
2260 	struct nlattr *stats;
2261 
2262 	struct nla_map {
2263 		u32 key;
2264 		u32 val;
2265 	};
2266 
2267 	struct nla_map map[] = {
2268 		{TIPC_NLA_STATS_RX_INFO, 0},
2269 		{TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2270 		{TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2271 		{TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2272 		{TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2273 		{TIPC_NLA_STATS_TX_INFO, 0},
2274 		{TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2275 		{TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2276 		{TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2277 		{TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2278 		{TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2279 			s->msg_length_counts : 1},
2280 		{TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2281 		{TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2282 		{TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2283 		{TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2284 		{TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2285 		{TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2286 		{TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2287 		{TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2288 		{TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2289 		{TIPC_NLA_STATS_RX_STATES, s->recv_states},
2290 		{TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2291 		{TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2292 		{TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2293 		{TIPC_NLA_STATS_TX_STATES, s->sent_states},
2294 		{TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2295 		{TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2296 		{TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2297 		{TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2298 		{TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2299 		{TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2300 		{TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2301 		{TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2302 			(s->accu_queue_sz / s->queue_sz_counts) : 0}
2303 	};
2304 
2305 	stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2306 	if (!stats)
2307 		return -EMSGSIZE;
2308 
2309 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2310 		if (nla_put_u32(skb, map[i].key, map[i].val))
2311 			goto msg_full;
2312 
2313 	nla_nest_end(skb, stats);
2314 
2315 	return 0;
2316 msg_full:
2317 	nla_nest_cancel(skb, stats);
2318 
2319 	return -EMSGSIZE;
2320 }
2321 
2322 /* Caller should hold appropriate locks to protect the link */
2323 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2324 		       struct tipc_link *link, int nlflags)
2325 {
2326 	u32 self = tipc_own_addr(net);
2327 	struct nlattr *attrs;
2328 	struct nlattr *prop;
2329 	void *hdr;
2330 	int err;
2331 
2332 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2333 			  nlflags, TIPC_NL_LINK_GET);
2334 	if (!hdr)
2335 		return -EMSGSIZE;
2336 
2337 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2338 	if (!attrs)
2339 		goto msg_full;
2340 
2341 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2342 		goto attr_msg_full;
2343 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2344 		goto attr_msg_full;
2345 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2346 		goto attr_msg_full;
2347 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2348 		goto attr_msg_full;
2349 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2350 		goto attr_msg_full;
2351 
2352 	if (tipc_link_is_up(link))
2353 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2354 			goto attr_msg_full;
2355 	if (link->active)
2356 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2357 			goto attr_msg_full;
2358 
2359 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2360 	if (!prop)
2361 		goto attr_msg_full;
2362 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2363 		goto prop_msg_full;
2364 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2365 		goto prop_msg_full;
2366 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2367 			link->window))
2368 		goto prop_msg_full;
2369 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2370 		goto prop_msg_full;
2371 	nla_nest_end(msg->skb, prop);
2372 
2373 	err = __tipc_nl_add_stats(msg->skb, &link->stats);
2374 	if (err)
2375 		goto attr_msg_full;
2376 
2377 	nla_nest_end(msg->skb, attrs);
2378 	genlmsg_end(msg->skb, hdr);
2379 
2380 	return 0;
2381 
2382 prop_msg_full:
2383 	nla_nest_cancel(msg->skb, prop);
2384 attr_msg_full:
2385 	nla_nest_cancel(msg->skb, attrs);
2386 msg_full:
2387 	genlmsg_cancel(msg->skb, hdr);
2388 
2389 	return -EMSGSIZE;
2390 }
2391 
2392 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2393 				      struct tipc_stats *stats)
2394 {
2395 	int i;
2396 	struct nlattr *nest;
2397 
2398 	struct nla_map {
2399 		__u32 key;
2400 		__u32 val;
2401 	};
2402 
2403 	struct nla_map map[] = {
2404 		{TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2405 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2406 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2407 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2408 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2409 		{TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2410 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2411 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2412 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2413 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2414 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2415 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2416 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2417 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2418 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2419 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2420 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2421 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2422 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2423 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2424 	};
2425 
2426 	nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2427 	if (!nest)
2428 		return -EMSGSIZE;
2429 
2430 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2431 		if (nla_put_u32(skb, map[i].key, map[i].val))
2432 			goto msg_full;
2433 
2434 	nla_nest_end(skb, nest);
2435 
2436 	return 0;
2437 msg_full:
2438 	nla_nest_cancel(skb, nest);
2439 
2440 	return -EMSGSIZE;
2441 }
2442 
2443 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2444 {
2445 	int err;
2446 	void *hdr;
2447 	struct nlattr *attrs;
2448 	struct nlattr *prop;
2449 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2450 	u32 bc_mode = tipc_bcast_get_broadcast_mode(net);
2451 	u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2452 	struct tipc_link *bcl = tn->bcl;
2453 
2454 	if (!bcl)
2455 		return 0;
2456 
2457 	tipc_bcast_lock(net);
2458 
2459 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2460 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
2461 	if (!hdr) {
2462 		tipc_bcast_unlock(net);
2463 		return -EMSGSIZE;
2464 	}
2465 
2466 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2467 	if (!attrs)
2468 		goto msg_full;
2469 
2470 	/* The broadcast link is always up */
2471 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2472 		goto attr_msg_full;
2473 
2474 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2475 		goto attr_msg_full;
2476 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2477 		goto attr_msg_full;
2478 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2479 		goto attr_msg_full;
2480 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2481 		goto attr_msg_full;
2482 
2483 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2484 	if (!prop)
2485 		goto attr_msg_full;
2486 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2487 		goto prop_msg_full;
2488 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2489 		goto prop_msg_full;
2490 	if (bc_mode & BCLINK_MODE_SEL)
2491 		if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2492 				bc_ratio))
2493 			goto prop_msg_full;
2494 	nla_nest_end(msg->skb, prop);
2495 
2496 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2497 	if (err)
2498 		goto attr_msg_full;
2499 
2500 	tipc_bcast_unlock(net);
2501 	nla_nest_end(msg->skb, attrs);
2502 	genlmsg_end(msg->skb, hdr);
2503 
2504 	return 0;
2505 
2506 prop_msg_full:
2507 	nla_nest_cancel(msg->skb, prop);
2508 attr_msg_full:
2509 	nla_nest_cancel(msg->skb, attrs);
2510 msg_full:
2511 	tipc_bcast_unlock(net);
2512 	genlmsg_cancel(msg->skb, hdr);
2513 
2514 	return -EMSGSIZE;
2515 }
2516 
2517 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2518 			     struct sk_buff_head *xmitq)
2519 {
2520 	l->tolerance = tol;
2521 	if (l->bc_rcvlink)
2522 		l->bc_rcvlink->tolerance = tol;
2523 	if (link_is_up(l))
2524 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2525 }
2526 
2527 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2528 			struct sk_buff_head *xmitq)
2529 {
2530 	l->priority = prio;
2531 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2532 }
2533 
2534 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2535 {
2536 	l->abort_limit = limit;
2537 }
2538 
2539 char *tipc_link_name_ext(struct tipc_link *l, char *buf)
2540 {
2541 	if (!l)
2542 		scnprintf(buf, TIPC_MAX_LINK_NAME, "null");
2543 	else if (link_is_bc_sndlink(l))
2544 		scnprintf(buf, TIPC_MAX_LINK_NAME, "broadcast-sender");
2545 	else if (link_is_bc_rcvlink(l))
2546 		scnprintf(buf, TIPC_MAX_LINK_NAME,
2547 			  "broadcast-receiver, peer %x", l->addr);
2548 	else
2549 		memcpy(buf, l->name, TIPC_MAX_LINK_NAME);
2550 
2551 	return buf;
2552 }
2553 
2554 /**
2555  * tipc_link_dump - dump TIPC link data
2556  * @l: tipc link to be dumped
2557  * @dqueues: bitmask to decide if any link queue to be dumped?
2558  *           - TIPC_DUMP_NONE: don't dump link queues
2559  *           - TIPC_DUMP_TRANSMQ: dump link transmq queue
2560  *           - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2561  *           - TIPC_DUMP_DEFERDQ: dump link deferd queue
2562  *           - TIPC_DUMP_INPUTQ: dump link input queue
2563  *           - TIPC_DUMP_WAKEUP: dump link wakeup queue
2564  *           - TIPC_DUMP_ALL: dump all the link queues above
2565  * @buf: returned buffer of dump data in format
2566  */
2567 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2568 {
2569 	int i = 0;
2570 	size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2571 	struct sk_buff_head *list;
2572 	struct sk_buff *hskb, *tskb;
2573 	u32 len;
2574 
2575 	if (!l) {
2576 		i += scnprintf(buf, sz, "link data: (null)\n");
2577 		return i;
2578 	}
2579 
2580 	i += scnprintf(buf, sz, "link data: %x", l->addr);
2581 	i += scnprintf(buf + i, sz - i, " %x", l->state);
2582 	i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2583 	i += scnprintf(buf + i, sz - i, " %u", l->session);
2584 	i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2585 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2586 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2587 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2588 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2589 	i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2590 	i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2591 	i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2592 	i += scnprintf(buf + i, sz - i, " %u", 0);
2593 	i += scnprintf(buf + i, sz - i, " %u", 0);
2594 	i += scnprintf(buf + i, sz - i, " %u", l->acked);
2595 
2596 	list = &l->transmq;
2597 	len = skb_queue_len(list);
2598 	hskb = skb_peek(list);
2599 	tskb = skb_peek_tail(list);
2600 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2601 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2602 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2603 
2604 	list = &l->deferdq;
2605 	len = skb_queue_len(list);
2606 	hskb = skb_peek(list);
2607 	tskb = skb_peek_tail(list);
2608 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2609 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2610 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2611 
2612 	list = &l->backlogq;
2613 	len = skb_queue_len(list);
2614 	hskb = skb_peek(list);
2615 	tskb = skb_peek_tail(list);
2616 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2617 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2618 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2619 
2620 	list = l->inputq;
2621 	len = skb_queue_len(list);
2622 	hskb = skb_peek(list);
2623 	tskb = skb_peek_tail(list);
2624 	i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2625 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2626 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2627 
2628 	if (dqueues & TIPC_DUMP_TRANSMQ) {
2629 		i += scnprintf(buf + i, sz - i, "transmq: ");
2630 		i += tipc_list_dump(&l->transmq, false, buf + i);
2631 	}
2632 	if (dqueues & TIPC_DUMP_BACKLOGQ) {
2633 		i += scnprintf(buf + i, sz - i,
2634 			       "backlogq: <%u %u %u %u %u>, ",
2635 			       l->backlog[TIPC_LOW_IMPORTANCE].len,
2636 			       l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2637 			       l->backlog[TIPC_HIGH_IMPORTANCE].len,
2638 			       l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2639 			       l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2640 		i += tipc_list_dump(&l->backlogq, false, buf + i);
2641 	}
2642 	if (dqueues & TIPC_DUMP_DEFERDQ) {
2643 		i += scnprintf(buf + i, sz - i, "deferdq: ");
2644 		i += tipc_list_dump(&l->deferdq, false, buf + i);
2645 	}
2646 	if (dqueues & TIPC_DUMP_INPUTQ) {
2647 		i += scnprintf(buf + i, sz - i, "inputq: ");
2648 		i += tipc_list_dump(l->inputq, false, buf + i);
2649 	}
2650 	if (dqueues & TIPC_DUMP_WAKEUP) {
2651 		i += scnprintf(buf + i, sz - i, "wakeup: ");
2652 		i += tipc_list_dump(&l->wakeupq, false, buf + i);
2653 	}
2654 
2655 	return i;
2656 }
2657