xref: /linux/net/tipc/link.c (revision 02000b55850deeadffe433e4b4930a8831f477de)
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_count: # of identical retransmit requests made by peer
110  * @ackers: # of peers that needs to ack each packet before it can be released
111  * @acked: # last packet acked by a certain peer. Used for broadcast.
112  * @rcv_nxt: next sequence number to expect for inbound messages
113  * @deferred_queue: deferred queue saved OOS b'cast message received from node
114  * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
115  * @inputq: buffer queue for messages to be delivered upwards
116  * @namedq: buffer queue for name table messages to be delivered upwards
117  * @next_out: ptr to first unsent outbound message in queue
118  * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
119  * @long_msg_seq_no: next identifier to use for outbound fragmented messages
120  * @reasm_buf: head of partially reassembled inbound message fragments
121  * @bc_rcvr: marks that this is a broadcast receiver link
122  * @stats: collects statistics regarding link activity
123  */
124 struct tipc_link {
125 	u32 addr;
126 	char name[TIPC_MAX_LINK_NAME];
127 	struct net *net;
128 
129 	/* Management and link supervision data */
130 	u32 peer_session;
131 	u32 session;
132 	u32 peer_bearer_id;
133 	u32 bearer_id;
134 	u32 tolerance;
135 	u32 abort_limit;
136 	u32 state;
137 	u16 peer_caps;
138 	bool active;
139 	u32 silent_intv_cnt;
140 	char if_name[TIPC_MAX_IF_NAME];
141 	u32 priority;
142 	char net_plane;
143 	struct tipc_mon_state mon_state;
144 	u16 rst_cnt;
145 
146 	/* Failover/synch */
147 	u16 drop_point;
148 	struct sk_buff *failover_reasm_skb;
149 
150 	/* Max packet negotiation */
151 	u16 mtu;
152 	u16 advertised_mtu;
153 
154 	/* Sending */
155 	struct sk_buff_head transmq;
156 	struct sk_buff_head backlogq;
157 	struct {
158 		u16 len;
159 		u16 limit;
160 	} backlog[5];
161 	u16 snd_nxt;
162 	u16 last_retransm;
163 	u16 window;
164 	u32 stale_count;
165 
166 	/* Reception */
167 	u16 rcv_nxt;
168 	u32 rcv_unacked;
169 	struct sk_buff_head deferdq;
170 	struct sk_buff_head *inputq;
171 	struct sk_buff_head *namedq;
172 
173 	/* Congestion handling */
174 	struct sk_buff_head wakeupq;
175 
176 	/* Fragmentation/reassembly */
177 	struct sk_buff *reasm_buf;
178 
179 	/* Broadcast */
180 	u16 ackers;
181 	u16 acked;
182 	struct tipc_link *bc_rcvlink;
183 	struct tipc_link *bc_sndlink;
184 	unsigned long prev_retr;
185 	u16 prev_from;
186 	u16 prev_to;
187 	u8 nack_state;
188 	bool bc_peer_is_up;
189 
190 	/* Statistics */
191 	struct tipc_stats stats;
192 };
193 
194 /*
195  * Error message prefixes
196  */
197 static const char *link_co_err = "Link tunneling error, ";
198 static const char *link_rst_msg = "Resetting link ";
199 
200 /* Send states for broadcast NACKs
201  */
202 enum {
203 	BC_NACK_SND_CONDITIONAL,
204 	BC_NACK_SND_UNCONDITIONAL,
205 	BC_NACK_SND_SUPPRESS,
206 };
207 
208 #define TIPC_BC_RETR_LIMIT 10   /* [ms] */
209 
210 /*
211  * Interval between NACKs when packets arrive out of order
212  */
213 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
214 
215 /* Wildcard value for link session numbers. When it is known that
216  * peer endpoint is down, any session number must be accepted.
217  */
218 #define ANY_SESSION 0x10000
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 int tipc_link_is_active(struct tipc_link *l)
301 {
302 	return l->active;
303 }
304 
305 void tipc_link_set_active(struct tipc_link *l, bool active)
306 {
307 	l->active = active;
308 }
309 
310 u32 tipc_link_id(struct tipc_link *l)
311 {
312 	return l->peer_bearer_id << 16 | l->bearer_id;
313 }
314 
315 int tipc_link_window(struct tipc_link *l)
316 {
317 	return l->window;
318 }
319 
320 int tipc_link_prio(struct tipc_link *l)
321 {
322 	return l->priority;
323 }
324 
325 unsigned long tipc_link_tolerance(struct tipc_link *l)
326 {
327 	return l->tolerance;
328 }
329 
330 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
331 {
332 	return l->inputq;
333 }
334 
335 char tipc_link_plane(struct tipc_link *l)
336 {
337 	return l->net_plane;
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 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->peer_session = ANY_SESSION;
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 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->peer_session = ANY_SESSION;
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->acked = 0;
861 	l->silent_intv_cnt = 0;
862 	l->rst_cnt = 0;
863 	l->stale_count = 0;
864 	l->bc_peer_is_up = false;
865 	memset(&l->mon_state, 0, sizeof(l->mon_state));
866 	tipc_link_reset_stats(l);
867 }
868 
869 /**
870  * tipc_link_xmit(): enqueue buffer list according to queue situation
871  * @link: link to use
872  * @list: chain of buffers containing message
873  * @xmitq: returned list of packets to be sent by caller
874  *
875  * Consumes the buffer chain.
876  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
877  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
878  */
879 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
880 		   struct sk_buff_head *xmitq)
881 {
882 	struct tipc_msg *hdr = buf_msg(skb_peek(list));
883 	unsigned int maxwin = l->window;
884 	int imp = msg_importance(hdr);
885 	unsigned int mtu = l->mtu;
886 	u16 ack = l->rcv_nxt - 1;
887 	u16 seqno = l->snd_nxt;
888 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
889 	struct sk_buff_head *transmq = &l->transmq;
890 	struct sk_buff_head *backlogq = &l->backlogq;
891 	struct sk_buff *skb, *_skb, *bskb;
892 	int pkt_cnt = skb_queue_len(list);
893 	int rc = 0;
894 
895 	if (unlikely(msg_size(hdr) > mtu)) {
896 		skb_queue_purge(list);
897 		return -EMSGSIZE;
898 	}
899 
900 	/* Allow oversubscription of one data msg per source at congestion */
901 	if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
902 		if (imp == TIPC_SYSTEM_IMPORTANCE) {
903 			pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
904 			return -ENOBUFS;
905 		}
906 		rc = link_schedule_user(l, hdr);
907 	}
908 
909 	if (pkt_cnt > 1) {
910 		l->stats.sent_fragmented++;
911 		l->stats.sent_fragments += pkt_cnt;
912 	}
913 
914 	/* Prepare each packet for sending, and add to relevant queue: */
915 	while (skb_queue_len(list)) {
916 		skb = skb_peek(list);
917 		hdr = buf_msg(skb);
918 		msg_set_seqno(hdr, seqno);
919 		msg_set_ack(hdr, ack);
920 		msg_set_bcast_ack(hdr, bc_ack);
921 
922 		if (likely(skb_queue_len(transmq) < maxwin)) {
923 			_skb = skb_clone(skb, GFP_ATOMIC);
924 			if (!_skb) {
925 				skb_queue_purge(list);
926 				return -ENOBUFS;
927 			}
928 			__skb_dequeue(list);
929 			__skb_queue_tail(transmq, skb);
930 			__skb_queue_tail(xmitq, _skb);
931 			TIPC_SKB_CB(skb)->ackers = l->ackers;
932 			l->rcv_unacked = 0;
933 			l->stats.sent_pkts++;
934 			seqno++;
935 			continue;
936 		}
937 		if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
938 			kfree_skb(__skb_dequeue(list));
939 			l->stats.sent_bundled++;
940 			continue;
941 		}
942 		if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
943 			kfree_skb(__skb_dequeue(list));
944 			__skb_queue_tail(backlogq, bskb);
945 			l->backlog[msg_importance(buf_msg(bskb))].len++;
946 			l->stats.sent_bundled++;
947 			l->stats.sent_bundles++;
948 			continue;
949 		}
950 		l->backlog[imp].len += skb_queue_len(list);
951 		skb_queue_splice_tail_init(list, backlogq);
952 	}
953 	l->snd_nxt = seqno;
954 	return rc;
955 }
956 
957 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
958 {
959 	struct sk_buff *skb, *_skb;
960 	struct tipc_msg *hdr;
961 	u16 seqno = l->snd_nxt;
962 	u16 ack = l->rcv_nxt - 1;
963 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
964 
965 	while (skb_queue_len(&l->transmq) < l->window) {
966 		skb = skb_peek(&l->backlogq);
967 		if (!skb)
968 			break;
969 		_skb = skb_clone(skb, GFP_ATOMIC);
970 		if (!_skb)
971 			break;
972 		__skb_dequeue(&l->backlogq);
973 		hdr = buf_msg(skb);
974 		l->backlog[msg_importance(hdr)].len--;
975 		__skb_queue_tail(&l->transmq, skb);
976 		__skb_queue_tail(xmitq, _skb);
977 		TIPC_SKB_CB(skb)->ackers = l->ackers;
978 		msg_set_seqno(hdr, seqno);
979 		msg_set_ack(hdr, ack);
980 		msg_set_bcast_ack(hdr, bc_ack);
981 		l->rcv_unacked = 0;
982 		l->stats.sent_pkts++;
983 		seqno++;
984 	}
985 	l->snd_nxt = seqno;
986 }
987 
988 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
989 {
990 	struct tipc_msg *hdr = buf_msg(skb);
991 
992 	pr_warn("Retransmission failure on link <%s>\n", l->name);
993 	link_print(l, "State of link ");
994 	pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
995 		msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
996 	pr_info("sqno %u, prev: %x, src: %x\n",
997 		msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
998 }
999 
1000 int tipc_link_retrans(struct tipc_link *l, struct tipc_link *nacker,
1001 		      u16 from, u16 to, struct sk_buff_head *xmitq)
1002 {
1003 	struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1004 	struct tipc_msg *hdr;
1005 	u16 ack = l->rcv_nxt - 1;
1006 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1007 
1008 	if (!skb)
1009 		return 0;
1010 
1011 	/* Detect repeated retransmit failures on same packet */
1012 	if (nacker->last_retransm != buf_seqno(skb)) {
1013 		nacker->last_retransm = buf_seqno(skb);
1014 		nacker->stale_count = 1;
1015 	} else if (++nacker->stale_count > 100) {
1016 		link_retransmit_failure(l, skb);
1017 		nacker->stale_count = 0;
1018 		if (link_is_bc_sndlink(l))
1019 			return TIPC_LINK_DOWN_EVT;
1020 		return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1021 	}
1022 
1023 	/* Move forward to where retransmission should start */
1024 	skb_queue_walk(&l->transmq, skb) {
1025 		if (!less(buf_seqno(skb), from))
1026 			break;
1027 	}
1028 
1029 	skb_queue_walk_from(&l->transmq, skb) {
1030 		if (more(buf_seqno(skb), to))
1031 			break;
1032 		hdr = buf_msg(skb);
1033 		_skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1034 		if (!_skb)
1035 			return 0;
1036 		hdr = buf_msg(_skb);
1037 		msg_set_ack(hdr, ack);
1038 		msg_set_bcast_ack(hdr, bc_ack);
1039 		_skb->priority = TC_PRIO_CONTROL;
1040 		__skb_queue_tail(xmitq, _skb);
1041 		l->stats.retransmitted++;
1042 	}
1043 	return 0;
1044 }
1045 
1046 /* tipc_data_input - deliver data and name distr msgs to upper layer
1047  *
1048  * Consumes buffer if message is of right type
1049  * Node lock must be held
1050  */
1051 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1052 			    struct sk_buff_head *inputq)
1053 {
1054 	struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1055 	struct tipc_msg *hdr = buf_msg(skb);
1056 
1057 	switch (msg_user(hdr)) {
1058 	case TIPC_LOW_IMPORTANCE:
1059 	case TIPC_MEDIUM_IMPORTANCE:
1060 	case TIPC_HIGH_IMPORTANCE:
1061 	case TIPC_CRITICAL_IMPORTANCE:
1062 		if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1063 			skb_queue_tail(mc_inputq, skb);
1064 			return true;
1065 		}
1066 		/* else: fall through */
1067 	case CONN_MANAGER:
1068 		skb_queue_tail(inputq, skb);
1069 		return true;
1070 	case GROUP_PROTOCOL:
1071 		skb_queue_tail(mc_inputq, skb);
1072 		return true;
1073 	case NAME_DISTRIBUTOR:
1074 		l->bc_rcvlink->state = LINK_ESTABLISHED;
1075 		skb_queue_tail(l->namedq, skb);
1076 		return true;
1077 	case MSG_BUNDLER:
1078 	case TUNNEL_PROTOCOL:
1079 	case MSG_FRAGMENTER:
1080 	case BCAST_PROTOCOL:
1081 		return false;
1082 	default:
1083 		pr_warn("Dropping received illegal msg type\n");
1084 		kfree_skb(skb);
1085 		return false;
1086 	};
1087 }
1088 
1089 /* tipc_link_input - process packet that has passed link protocol check
1090  *
1091  * Consumes buffer
1092  */
1093 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1094 			   struct sk_buff_head *inputq)
1095 {
1096 	struct tipc_msg *hdr = buf_msg(skb);
1097 	struct sk_buff **reasm_skb = &l->reasm_buf;
1098 	struct sk_buff *iskb;
1099 	struct sk_buff_head tmpq;
1100 	int usr = msg_user(hdr);
1101 	int rc = 0;
1102 	int pos = 0;
1103 	int ipos = 0;
1104 
1105 	if (unlikely(usr == TUNNEL_PROTOCOL)) {
1106 		if (msg_type(hdr) == SYNCH_MSG) {
1107 			__skb_queue_purge(&l->deferdq);
1108 			goto drop;
1109 		}
1110 		if (!tipc_msg_extract(skb, &iskb, &ipos))
1111 			return rc;
1112 		kfree_skb(skb);
1113 		skb = iskb;
1114 		hdr = buf_msg(skb);
1115 		if (less(msg_seqno(hdr), l->drop_point))
1116 			goto drop;
1117 		if (tipc_data_input(l, skb, inputq))
1118 			return rc;
1119 		usr = msg_user(hdr);
1120 		reasm_skb = &l->failover_reasm_skb;
1121 	}
1122 
1123 	if (usr == MSG_BUNDLER) {
1124 		skb_queue_head_init(&tmpq);
1125 		l->stats.recv_bundles++;
1126 		l->stats.recv_bundled += msg_msgcnt(hdr);
1127 		while (tipc_msg_extract(skb, &iskb, &pos))
1128 			tipc_data_input(l, iskb, &tmpq);
1129 		tipc_skb_queue_splice_tail(&tmpq, inputq);
1130 		return 0;
1131 	} else if (usr == MSG_FRAGMENTER) {
1132 		l->stats.recv_fragments++;
1133 		if (tipc_buf_append(reasm_skb, &skb)) {
1134 			l->stats.recv_fragmented++;
1135 			tipc_data_input(l, skb, inputq);
1136 		} else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1137 			pr_warn_ratelimited("Unable to build fragment list\n");
1138 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1139 		}
1140 		return 0;
1141 	} else if (usr == BCAST_PROTOCOL) {
1142 		tipc_bcast_lock(l->net);
1143 		tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1144 		tipc_bcast_unlock(l->net);
1145 	}
1146 drop:
1147 	kfree_skb(skb);
1148 	return 0;
1149 }
1150 
1151 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1152 {
1153 	bool released = false;
1154 	struct sk_buff *skb, *tmp;
1155 
1156 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1157 		if (more(buf_seqno(skb), acked))
1158 			break;
1159 		__skb_unlink(skb, &l->transmq);
1160 		kfree_skb(skb);
1161 		released = true;
1162 	}
1163 	return released;
1164 }
1165 
1166 /* tipc_link_build_state_msg: prepare link state message for transmission
1167  *
1168  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1169  * risk of ack storms towards the sender
1170  */
1171 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1172 {
1173 	if (!l)
1174 		return 0;
1175 
1176 	/* Broadcast ACK must be sent via a unicast link => defer to caller */
1177 	if (link_is_bc_rcvlink(l)) {
1178 		if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1179 			return 0;
1180 		l->rcv_unacked = 0;
1181 
1182 		/* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1183 		l->snd_nxt = l->rcv_nxt;
1184 		return TIPC_LINK_SND_STATE;
1185 	}
1186 
1187 	/* Unicast ACK */
1188 	l->rcv_unacked = 0;
1189 	l->stats.sent_acks++;
1190 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1191 	return 0;
1192 }
1193 
1194 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1195  */
1196 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1197 {
1198 	int mtyp = RESET_MSG;
1199 	struct sk_buff *skb;
1200 
1201 	if (l->state == LINK_ESTABLISHING)
1202 		mtyp = ACTIVATE_MSG;
1203 
1204 	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1205 
1206 	/* Inform peer that this endpoint is going down if applicable */
1207 	skb = skb_peek_tail(xmitq);
1208 	if (skb && (l->state == LINK_RESET))
1209 		msg_set_peer_stopping(buf_msg(skb), 1);
1210 }
1211 
1212 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1213  * Note that sending of broadcast NACK is coordinated among nodes, to
1214  * reduce the risk of NACK storms towards the sender
1215  */
1216 static int tipc_link_build_nack_msg(struct tipc_link *l,
1217 				    struct sk_buff_head *xmitq)
1218 {
1219 	u32 def_cnt = ++l->stats.deferred_recv;
1220 	int match1, match2;
1221 
1222 	if (link_is_bc_rcvlink(l)) {
1223 		match1 = def_cnt & 0xf;
1224 		match2 = tipc_own_addr(l->net) & 0xf;
1225 		if (match1 == match2)
1226 			return TIPC_LINK_SND_STATE;
1227 		return 0;
1228 	}
1229 
1230 	if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1231 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1232 	return 0;
1233 }
1234 
1235 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1236  * @l: the link that should handle the message
1237  * @skb: TIPC packet
1238  * @xmitq: queue to place packets to be sent after this call
1239  */
1240 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1241 		  struct sk_buff_head *xmitq)
1242 {
1243 	struct sk_buff_head *defq = &l->deferdq;
1244 	struct tipc_msg *hdr;
1245 	u16 seqno, rcv_nxt, win_lim;
1246 	int rc = 0;
1247 
1248 	do {
1249 		hdr = buf_msg(skb);
1250 		seqno = msg_seqno(hdr);
1251 		rcv_nxt = l->rcv_nxt;
1252 		win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1253 
1254 		/* Verify and update link state */
1255 		if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1256 			return tipc_link_proto_rcv(l, skb, xmitq);
1257 
1258 		if (unlikely(!link_is_up(l))) {
1259 			if (l->state == LINK_ESTABLISHING)
1260 				rc = TIPC_LINK_UP_EVT;
1261 			goto drop;
1262 		}
1263 
1264 		/* Don't send probe at next timeout expiration */
1265 		l->silent_intv_cnt = 0;
1266 
1267 		/* Drop if outside receive window */
1268 		if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1269 			l->stats.duplicates++;
1270 			goto drop;
1271 		}
1272 
1273 		/* Forward queues and wake up waiting users */
1274 		if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1275 			tipc_link_advance_backlog(l, xmitq);
1276 			if (unlikely(!skb_queue_empty(&l->wakeupq)))
1277 				link_prepare_wakeup(l);
1278 		}
1279 
1280 		/* Defer delivery if sequence gap */
1281 		if (unlikely(seqno != rcv_nxt)) {
1282 			__tipc_skb_queue_sorted(defq, seqno, skb);
1283 			rc |= tipc_link_build_nack_msg(l, xmitq);
1284 			break;
1285 		}
1286 
1287 		/* Deliver packet */
1288 		l->rcv_nxt++;
1289 		l->stats.recv_pkts++;
1290 		if (!tipc_data_input(l, skb, l->inputq))
1291 			rc |= tipc_link_input(l, skb, l->inputq);
1292 		if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1293 			rc |= tipc_link_build_state_msg(l, xmitq);
1294 		if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1295 			break;
1296 	} while ((skb = __skb_dequeue(defq)));
1297 
1298 	return rc;
1299 drop:
1300 	kfree_skb(skb);
1301 	return rc;
1302 }
1303 
1304 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1305 				      bool probe_reply, u16 rcvgap,
1306 				      int tolerance, int priority,
1307 				      struct sk_buff_head *xmitq)
1308 {
1309 	struct tipc_link *bcl = l->bc_rcvlink;
1310 	struct sk_buff *skb;
1311 	struct tipc_msg *hdr;
1312 	struct sk_buff_head *dfq = &l->deferdq;
1313 	bool node_up = link_is_up(bcl);
1314 	struct tipc_mon_state *mstate = &l->mon_state;
1315 	int dlen = 0;
1316 	void *data;
1317 
1318 	/* Don't send protocol message during reset or link failover */
1319 	if (tipc_link_is_blocked(l))
1320 		return;
1321 
1322 	if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1323 		return;
1324 
1325 	if (!skb_queue_empty(dfq))
1326 		rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1327 
1328 	skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1329 			      tipc_max_domain_size, l->addr,
1330 			      tipc_own_addr(l->net), 0, 0, 0);
1331 	if (!skb)
1332 		return;
1333 
1334 	hdr = buf_msg(skb);
1335 	data = msg_data(hdr);
1336 	msg_set_session(hdr, l->session);
1337 	msg_set_bearer_id(hdr, l->bearer_id);
1338 	msg_set_net_plane(hdr, l->net_plane);
1339 	msg_set_next_sent(hdr, l->snd_nxt);
1340 	msg_set_ack(hdr, l->rcv_nxt - 1);
1341 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1342 	msg_set_bc_ack_invalid(hdr, !node_up);
1343 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1344 	msg_set_link_tolerance(hdr, tolerance);
1345 	msg_set_linkprio(hdr, priority);
1346 	msg_set_redundant_link(hdr, node_up);
1347 	msg_set_seq_gap(hdr, 0);
1348 	msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1349 
1350 	if (mtyp == STATE_MSG) {
1351 		msg_set_seq_gap(hdr, rcvgap);
1352 		msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1353 		msg_set_probe(hdr, probe);
1354 		msg_set_is_keepalive(hdr, probe || probe_reply);
1355 		tipc_mon_prep(l->net, data, &dlen, mstate, l->bearer_id);
1356 		msg_set_size(hdr, INT_H_SIZE + dlen);
1357 		skb_trim(skb, INT_H_SIZE + dlen);
1358 		l->stats.sent_states++;
1359 		l->rcv_unacked = 0;
1360 	} else {
1361 		/* RESET_MSG or ACTIVATE_MSG */
1362 		msg_set_max_pkt(hdr, l->advertised_mtu);
1363 		strcpy(data, l->if_name);
1364 		msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1365 		skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1366 	}
1367 	if (probe)
1368 		l->stats.sent_probes++;
1369 	if (rcvgap)
1370 		l->stats.sent_nacks++;
1371 	skb->priority = TC_PRIO_CONTROL;
1372 	__skb_queue_tail(xmitq, skb);
1373 }
1374 
1375 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1376  * with contents of the link's transmit and backlog queues.
1377  */
1378 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1379 			   int mtyp, struct sk_buff_head *xmitq)
1380 {
1381 	struct sk_buff *skb, *tnlskb;
1382 	struct tipc_msg *hdr, tnlhdr;
1383 	struct sk_buff_head *queue = &l->transmq;
1384 	struct sk_buff_head tmpxq, tnlq;
1385 	u16 pktlen, pktcnt, seqno = l->snd_nxt;
1386 
1387 	if (!tnl)
1388 		return;
1389 
1390 	skb_queue_head_init(&tnlq);
1391 	skb_queue_head_init(&tmpxq);
1392 
1393 	/* At least one packet required for safe algorithm => add dummy */
1394 	skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1395 			      BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1396 			      0, 0, TIPC_ERR_NO_PORT);
1397 	if (!skb) {
1398 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1399 		return;
1400 	}
1401 	skb_queue_tail(&tnlq, skb);
1402 	tipc_link_xmit(l, &tnlq, &tmpxq);
1403 	__skb_queue_purge(&tmpxq);
1404 
1405 	/* Initialize reusable tunnel packet header */
1406 	tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1407 		      mtyp, INT_H_SIZE, l->addr);
1408 	pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1409 	msg_set_msgcnt(&tnlhdr, pktcnt);
1410 	msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1411 tnl:
1412 	/* Wrap each packet into a tunnel packet */
1413 	skb_queue_walk(queue, skb) {
1414 		hdr = buf_msg(skb);
1415 		if (queue == &l->backlogq)
1416 			msg_set_seqno(hdr, seqno++);
1417 		pktlen = msg_size(hdr);
1418 		msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1419 		tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1420 		if (!tnlskb) {
1421 			pr_warn("%sunable to send packet\n", link_co_err);
1422 			return;
1423 		}
1424 		skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1425 		skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1426 		__skb_queue_tail(&tnlq, tnlskb);
1427 	}
1428 	if (queue != &l->backlogq) {
1429 		queue = &l->backlogq;
1430 		goto tnl;
1431 	}
1432 
1433 	tipc_link_xmit(tnl, &tnlq, xmitq);
1434 
1435 	if (mtyp == FAILOVER_MSG) {
1436 		tnl->drop_point = l->rcv_nxt;
1437 		tnl->failover_reasm_skb = l->reasm_buf;
1438 		l->reasm_buf = NULL;
1439 	}
1440 }
1441 
1442 /* tipc_link_proto_rcv(): receive link level protocol message :
1443  * Note that network plane id propagates through the network, and may
1444  * change at any time. The node with lowest numerical id determines
1445  * network plane
1446  */
1447 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1448 			       struct sk_buff_head *xmitq)
1449 {
1450 	struct tipc_msg *hdr = buf_msg(skb);
1451 	u16 rcvgap = 0;
1452 	u16 ack = msg_ack(hdr);
1453 	u16 gap = msg_seq_gap(hdr);
1454 	u16 peers_snd_nxt =  msg_next_sent(hdr);
1455 	u16 peers_tol = msg_link_tolerance(hdr);
1456 	u16 peers_prio = msg_linkprio(hdr);
1457 	u16 rcv_nxt = l->rcv_nxt;
1458 	u16 dlen = msg_data_sz(hdr);
1459 	int mtyp = msg_type(hdr);
1460 	bool reply = msg_probe(hdr);
1461 	void *data;
1462 	char *if_name;
1463 	int rc = 0;
1464 
1465 	if (tipc_link_is_blocked(l) || !xmitq)
1466 		goto exit;
1467 
1468 	if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1469 		l->net_plane = msg_net_plane(hdr);
1470 
1471 	skb_linearize(skb);
1472 	hdr = buf_msg(skb);
1473 	data = msg_data(hdr);
1474 
1475 	switch (mtyp) {
1476 	case RESET_MSG:
1477 
1478 		/* Ignore duplicate RESET with old session number */
1479 		if ((less_eq(msg_session(hdr), l->peer_session)) &&
1480 		    (l->peer_session != ANY_SESSION))
1481 			break;
1482 		/* fall thru' */
1483 
1484 	case ACTIVATE_MSG:
1485 
1486 		/* Complete own link name with peer's interface name */
1487 		if_name =  strrchr(l->name, ':') + 1;
1488 		if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1489 			break;
1490 		if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1491 			break;
1492 		strncpy(if_name, data, TIPC_MAX_IF_NAME);
1493 
1494 		/* Update own tolerance if peer indicates a non-zero value */
1495 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1496 			l->tolerance = peers_tol;
1497 
1498 		/* Update own priority if peer's priority is higher */
1499 		if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1500 			l->priority = peers_prio;
1501 
1502 		/* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1503 		if (msg_peer_stopping(hdr))
1504 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1505 		else if ((mtyp == RESET_MSG) || !link_is_up(l))
1506 			rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1507 
1508 		/* ACTIVATE_MSG takes up link if it was already locally reset */
1509 		if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1510 			rc = TIPC_LINK_UP_EVT;
1511 
1512 		l->peer_session = msg_session(hdr);
1513 		l->peer_bearer_id = msg_bearer_id(hdr);
1514 		if (l->mtu > msg_max_pkt(hdr))
1515 			l->mtu = msg_max_pkt(hdr);
1516 		break;
1517 
1518 	case STATE_MSG:
1519 
1520 		/* Update own tolerance if peer indicates a non-zero value */
1521 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1522 			l->tolerance = peers_tol;
1523 
1524 		/* Update own prio if peer indicates a different value */
1525 		if ((peers_prio != l->priority) &&
1526 		    in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
1527 			l->priority = peers_prio;
1528 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1529 		}
1530 
1531 		l->silent_intv_cnt = 0;
1532 		l->stats.recv_states++;
1533 		if (msg_probe(hdr))
1534 			l->stats.recv_probes++;
1535 
1536 		if (!link_is_up(l)) {
1537 			if (l->state == LINK_ESTABLISHING)
1538 				rc = TIPC_LINK_UP_EVT;
1539 			break;
1540 		}
1541 		tipc_mon_rcv(l->net, data, dlen, l->addr,
1542 			     &l->mon_state, l->bearer_id);
1543 
1544 		/* Send NACK if peer has sent pkts we haven't received yet */
1545 		if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1546 			rcvgap = peers_snd_nxt - l->rcv_nxt;
1547 		if (rcvgap || reply)
1548 			tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
1549 						  rcvgap, 0, 0, xmitq);
1550 		tipc_link_release_pkts(l, ack);
1551 
1552 		/* If NACK, retransmit will now start at right position */
1553 		if (gap) {
1554 			rc = tipc_link_retrans(l, l, ack + 1, ack + gap, xmitq);
1555 			l->stats.recv_nacks++;
1556 		}
1557 
1558 		tipc_link_advance_backlog(l, xmitq);
1559 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
1560 			link_prepare_wakeup(l);
1561 	}
1562 exit:
1563 	kfree_skb(skb);
1564 	return rc;
1565 }
1566 
1567 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1568  */
1569 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1570 					 u16 peers_snd_nxt,
1571 					 struct sk_buff_head *xmitq)
1572 {
1573 	struct sk_buff *skb;
1574 	struct tipc_msg *hdr;
1575 	struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1576 	u16 ack = l->rcv_nxt - 1;
1577 	u16 gap_to = peers_snd_nxt - 1;
1578 
1579 	skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1580 			      0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1581 	if (!skb)
1582 		return false;
1583 	hdr = buf_msg(skb);
1584 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1585 	msg_set_bcast_ack(hdr, ack);
1586 	msg_set_bcgap_after(hdr, ack);
1587 	if (dfrd_skb)
1588 		gap_to = buf_seqno(dfrd_skb) - 1;
1589 	msg_set_bcgap_to(hdr, gap_to);
1590 	msg_set_non_seq(hdr, bcast);
1591 	__skb_queue_tail(xmitq, skb);
1592 	return true;
1593 }
1594 
1595 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1596  *
1597  * Give a newly added peer node the sequence number where it should
1598  * start receiving and acking broadcast packets.
1599  */
1600 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1601 					struct sk_buff_head *xmitq)
1602 {
1603 	struct sk_buff_head list;
1604 
1605 	__skb_queue_head_init(&list);
1606 	if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1607 		return;
1608 	msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
1609 	tipc_link_xmit(l, &list, xmitq);
1610 }
1611 
1612 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1613  */
1614 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1615 {
1616 	int mtyp = msg_type(hdr);
1617 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1618 
1619 	if (link_is_up(l))
1620 		return;
1621 
1622 	if (msg_user(hdr) == BCAST_PROTOCOL) {
1623 		l->rcv_nxt = peers_snd_nxt;
1624 		l->state = LINK_ESTABLISHED;
1625 		return;
1626 	}
1627 
1628 	if (l->peer_caps & TIPC_BCAST_SYNCH)
1629 		return;
1630 
1631 	if (msg_peer_node_is_up(hdr))
1632 		return;
1633 
1634 	/* Compatibility: accept older, less safe initial synch data */
1635 	if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1636 		l->rcv_nxt = peers_snd_nxt;
1637 }
1638 
1639 /* link_bc_retr eval()- check if the indicated range can be retransmitted now
1640  * - Adjust permitted range if there is overlap with previous retransmission
1641  */
1642 static bool link_bc_retr_eval(struct tipc_link *l, u16 *from, u16 *to)
1643 {
1644 	unsigned long elapsed = jiffies_to_msecs(jiffies - l->prev_retr);
1645 
1646 	if (less(*to, *from))
1647 		return false;
1648 
1649 	/* New retransmission request */
1650 	if ((elapsed > TIPC_BC_RETR_LIMIT) ||
1651 	    less(*to, l->prev_from) || more(*from, l->prev_to)) {
1652 		l->prev_from = *from;
1653 		l->prev_to = *to;
1654 		l->prev_retr = jiffies;
1655 		return true;
1656 	}
1657 
1658 	/* Inside range of previous retransmit */
1659 	if (!less(*from, l->prev_from) && !more(*to, l->prev_to))
1660 		return false;
1661 
1662 	/* Fully or partially outside previous range => exclude overlap */
1663 	if (less(*from, l->prev_from)) {
1664 		*to = l->prev_from - 1;
1665 		l->prev_from = *from;
1666 	}
1667 	if (more(*to, l->prev_to)) {
1668 		*from = l->prev_to + 1;
1669 		l->prev_to = *to;
1670 	}
1671 	l->prev_retr = jiffies;
1672 	return true;
1673 }
1674 
1675 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1676  */
1677 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1678 			  struct sk_buff_head *xmitq)
1679 {
1680 	struct tipc_link *snd_l = l->bc_sndlink;
1681 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1682 	u16 from = msg_bcast_ack(hdr) + 1;
1683 	u16 to = from + msg_bc_gap(hdr) - 1;
1684 	int rc = 0;
1685 
1686 	if (!link_is_up(l))
1687 		return rc;
1688 
1689 	if (!msg_peer_node_is_up(hdr))
1690 		return rc;
1691 
1692 	/* Open when peer ackowledges our bcast init msg (pkt #1) */
1693 	if (msg_ack(hdr))
1694 		l->bc_peer_is_up = true;
1695 
1696 	if (!l->bc_peer_is_up)
1697 		return rc;
1698 
1699 	l->stats.recv_nacks++;
1700 
1701 	/* Ignore if peers_snd_nxt goes beyond receive window */
1702 	if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1703 		return rc;
1704 
1705 	if (link_bc_retr_eval(snd_l, &from, &to))
1706 		rc = tipc_link_retrans(snd_l, l, from, to, xmitq);
1707 
1708 	l->snd_nxt = peers_snd_nxt;
1709 	if (link_bc_rcv_gap(l))
1710 		rc |= TIPC_LINK_SND_STATE;
1711 
1712 	/* Return now if sender supports nack via STATE messages */
1713 	if (l->peer_caps & TIPC_BCAST_STATE_NACK)
1714 		return rc;
1715 
1716 	/* Otherwise, be backwards compatible */
1717 
1718 	if (!more(peers_snd_nxt, l->rcv_nxt)) {
1719 		l->nack_state = BC_NACK_SND_CONDITIONAL;
1720 		return 0;
1721 	}
1722 
1723 	/* Don't NACK if one was recently sent or peeked */
1724 	if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1725 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1726 		return 0;
1727 	}
1728 
1729 	/* Conditionally delay NACK sending until next synch rcv */
1730 	if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1731 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1732 		if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1733 			return 0;
1734 	}
1735 
1736 	/* Send NACK now but suppress next one */
1737 	tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1738 	l->nack_state = BC_NACK_SND_SUPPRESS;
1739 	return 0;
1740 }
1741 
1742 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1743 			  struct sk_buff_head *xmitq)
1744 {
1745 	struct sk_buff *skb, *tmp;
1746 	struct tipc_link *snd_l = l->bc_sndlink;
1747 
1748 	if (!link_is_up(l) || !l->bc_peer_is_up)
1749 		return;
1750 
1751 	if (!more(acked, l->acked))
1752 		return;
1753 
1754 	/* Skip over packets peer has already acked */
1755 	skb_queue_walk(&snd_l->transmq, skb) {
1756 		if (more(buf_seqno(skb), l->acked))
1757 			break;
1758 	}
1759 
1760 	/* Update/release the packets peer is acking now */
1761 	skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1762 		if (more(buf_seqno(skb), acked))
1763 			break;
1764 		if (!--TIPC_SKB_CB(skb)->ackers) {
1765 			__skb_unlink(skb, &snd_l->transmq);
1766 			kfree_skb(skb);
1767 		}
1768 	}
1769 	l->acked = acked;
1770 	tipc_link_advance_backlog(snd_l, xmitq);
1771 	if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1772 		link_prepare_wakeup(snd_l);
1773 }
1774 
1775 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1776  * This function is here for backwards compatibility, since
1777  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
1778  */
1779 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1780 			  struct sk_buff_head *xmitq)
1781 {
1782 	struct tipc_msg *hdr = buf_msg(skb);
1783 	u32 dnode = msg_destnode(hdr);
1784 	int mtyp = msg_type(hdr);
1785 	u16 acked = msg_bcast_ack(hdr);
1786 	u16 from = acked + 1;
1787 	u16 to = msg_bcgap_to(hdr);
1788 	u16 peers_snd_nxt = to + 1;
1789 	int rc = 0;
1790 
1791 	kfree_skb(skb);
1792 
1793 	if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1794 		return 0;
1795 
1796 	if (mtyp != STATE_MSG)
1797 		return 0;
1798 
1799 	if (dnode == tipc_own_addr(l->net)) {
1800 		tipc_link_bc_ack_rcv(l, acked, xmitq);
1801 		rc = tipc_link_retrans(l->bc_sndlink, l, from, to, xmitq);
1802 		l->stats.recv_nacks++;
1803 		return rc;
1804 	}
1805 
1806 	/* Msg for other node => suppress own NACK at next sync if applicable */
1807 	if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1808 		l->nack_state = BC_NACK_SND_SUPPRESS;
1809 
1810 	return 0;
1811 }
1812 
1813 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1814 {
1815 	int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
1816 
1817 	l->window = win;
1818 	l->backlog[TIPC_LOW_IMPORTANCE].limit      = max_t(u16, 50, win);
1819 	l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = max_t(u16, 100, win * 2);
1820 	l->backlog[TIPC_HIGH_IMPORTANCE].limit     = max_t(u16, 150, win * 3);
1821 	l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
1822 	l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
1823 }
1824 
1825 /**
1826  * link_reset_stats - reset link statistics
1827  * @l: pointer to link
1828  */
1829 void tipc_link_reset_stats(struct tipc_link *l)
1830 {
1831 	memset(&l->stats, 0, sizeof(l->stats));
1832 }
1833 
1834 static void link_print(struct tipc_link *l, const char *str)
1835 {
1836 	struct sk_buff *hskb = skb_peek(&l->transmq);
1837 	u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1838 	u16 tail = l->snd_nxt - 1;
1839 
1840 	pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1841 	pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1842 		skb_queue_len(&l->transmq), head, tail,
1843 		skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1844 }
1845 
1846 /* Parse and validate nested (link) properties valid for media, bearer and link
1847  */
1848 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1849 {
1850 	int err;
1851 
1852 	err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1853 			       tipc_nl_prop_policy, NULL);
1854 	if (err)
1855 		return err;
1856 
1857 	if (props[TIPC_NLA_PROP_PRIO]) {
1858 		u32 prio;
1859 
1860 		prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1861 		if (prio > TIPC_MAX_LINK_PRI)
1862 			return -EINVAL;
1863 	}
1864 
1865 	if (props[TIPC_NLA_PROP_TOL]) {
1866 		u32 tol;
1867 
1868 		tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1869 		if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1870 			return -EINVAL;
1871 	}
1872 
1873 	if (props[TIPC_NLA_PROP_WIN]) {
1874 		u32 win;
1875 
1876 		win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1877 		if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1878 			return -EINVAL;
1879 	}
1880 
1881 	return 0;
1882 }
1883 
1884 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1885 {
1886 	int i;
1887 	struct nlattr *stats;
1888 
1889 	struct nla_map {
1890 		u32 key;
1891 		u32 val;
1892 	};
1893 
1894 	struct nla_map map[] = {
1895 		{TIPC_NLA_STATS_RX_INFO, 0},
1896 		{TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1897 		{TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1898 		{TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1899 		{TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1900 		{TIPC_NLA_STATS_TX_INFO, 0},
1901 		{TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1902 		{TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1903 		{TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1904 		{TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1905 		{TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1906 			s->msg_length_counts : 1},
1907 		{TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1908 		{TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1909 		{TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1910 		{TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1911 		{TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1912 		{TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1913 		{TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1914 		{TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1915 		{TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1916 		{TIPC_NLA_STATS_RX_STATES, s->recv_states},
1917 		{TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1918 		{TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1919 		{TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1920 		{TIPC_NLA_STATS_TX_STATES, s->sent_states},
1921 		{TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1922 		{TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1923 		{TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1924 		{TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1925 		{TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1926 		{TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1927 		{TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1928 		{TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1929 			(s->accu_queue_sz / s->queue_sz_counts) : 0}
1930 	};
1931 
1932 	stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1933 	if (!stats)
1934 		return -EMSGSIZE;
1935 
1936 	for (i = 0; i <  ARRAY_SIZE(map); i++)
1937 		if (nla_put_u32(skb, map[i].key, map[i].val))
1938 			goto msg_full;
1939 
1940 	nla_nest_end(skb, stats);
1941 
1942 	return 0;
1943 msg_full:
1944 	nla_nest_cancel(skb, stats);
1945 
1946 	return -EMSGSIZE;
1947 }
1948 
1949 /* Caller should hold appropriate locks to protect the link */
1950 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1951 		       struct tipc_link *link, int nlflags)
1952 {
1953 	u32 self = tipc_own_addr(net);
1954 	struct nlattr *attrs;
1955 	struct nlattr *prop;
1956 	void *hdr;
1957 	int err;
1958 
1959 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1960 			  nlflags, TIPC_NL_LINK_GET);
1961 	if (!hdr)
1962 		return -EMSGSIZE;
1963 
1964 	attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1965 	if (!attrs)
1966 		goto msg_full;
1967 
1968 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1969 		goto attr_msg_full;
1970 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
1971 		goto attr_msg_full;
1972 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1973 		goto attr_msg_full;
1974 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
1975 		goto attr_msg_full;
1976 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
1977 		goto attr_msg_full;
1978 
1979 	if (tipc_link_is_up(link))
1980 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1981 			goto attr_msg_full;
1982 	if (link->active)
1983 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
1984 			goto attr_msg_full;
1985 
1986 	prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1987 	if (!prop)
1988 		goto attr_msg_full;
1989 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1990 		goto prop_msg_full;
1991 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
1992 		goto prop_msg_full;
1993 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
1994 			link->window))
1995 		goto prop_msg_full;
1996 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1997 		goto prop_msg_full;
1998 	nla_nest_end(msg->skb, prop);
1999 
2000 	err = __tipc_nl_add_stats(msg->skb, &link->stats);
2001 	if (err)
2002 		goto attr_msg_full;
2003 
2004 	nla_nest_end(msg->skb, attrs);
2005 	genlmsg_end(msg->skb, hdr);
2006 
2007 	return 0;
2008 
2009 prop_msg_full:
2010 	nla_nest_cancel(msg->skb, prop);
2011 attr_msg_full:
2012 	nla_nest_cancel(msg->skb, attrs);
2013 msg_full:
2014 	genlmsg_cancel(msg->skb, hdr);
2015 
2016 	return -EMSGSIZE;
2017 }
2018 
2019 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2020 				      struct tipc_stats *stats)
2021 {
2022 	int i;
2023 	struct nlattr *nest;
2024 
2025 	struct nla_map {
2026 		__u32 key;
2027 		__u32 val;
2028 	};
2029 
2030 	struct nla_map map[] = {
2031 		{TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2032 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2033 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2034 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2035 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2036 		{TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2037 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2038 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2039 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2040 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2041 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2042 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2043 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2044 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2045 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2046 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2047 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2048 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2049 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2050 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2051 	};
2052 
2053 	nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
2054 	if (!nest)
2055 		return -EMSGSIZE;
2056 
2057 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2058 		if (nla_put_u32(skb, map[i].key, map[i].val))
2059 			goto msg_full;
2060 
2061 	nla_nest_end(skb, nest);
2062 
2063 	return 0;
2064 msg_full:
2065 	nla_nest_cancel(skb, nest);
2066 
2067 	return -EMSGSIZE;
2068 }
2069 
2070 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2071 {
2072 	int err;
2073 	void *hdr;
2074 	struct nlattr *attrs;
2075 	struct nlattr *prop;
2076 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2077 	struct tipc_link *bcl = tn->bcl;
2078 
2079 	if (!bcl)
2080 		return 0;
2081 
2082 	tipc_bcast_lock(net);
2083 
2084 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2085 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
2086 	if (!hdr) {
2087 		tipc_bcast_unlock(net);
2088 		return -EMSGSIZE;
2089 	}
2090 
2091 	attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
2092 	if (!attrs)
2093 		goto msg_full;
2094 
2095 	/* The broadcast link is always up */
2096 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2097 		goto attr_msg_full;
2098 
2099 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2100 		goto attr_msg_full;
2101 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2102 		goto attr_msg_full;
2103 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2104 		goto attr_msg_full;
2105 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2106 		goto attr_msg_full;
2107 
2108 	prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2109 	if (!prop)
2110 		goto attr_msg_full;
2111 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2112 		goto prop_msg_full;
2113 	nla_nest_end(msg->skb, prop);
2114 
2115 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2116 	if (err)
2117 		goto attr_msg_full;
2118 
2119 	tipc_bcast_unlock(net);
2120 	nla_nest_end(msg->skb, attrs);
2121 	genlmsg_end(msg->skb, hdr);
2122 
2123 	return 0;
2124 
2125 prop_msg_full:
2126 	nla_nest_cancel(msg->skb, prop);
2127 attr_msg_full:
2128 	nla_nest_cancel(msg->skb, attrs);
2129 msg_full:
2130 	tipc_bcast_unlock(net);
2131 	genlmsg_cancel(msg->skb, hdr);
2132 
2133 	return -EMSGSIZE;
2134 }
2135 
2136 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2137 			     struct sk_buff_head *xmitq)
2138 {
2139 	l->tolerance = tol;
2140 	if (link_is_up(l))
2141 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2142 }
2143 
2144 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2145 			struct sk_buff_head *xmitq)
2146 {
2147 	l->priority = prio;
2148 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2149 }
2150 
2151 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2152 {
2153 	l->abort_limit = limit;
2154 }
2155