xref: /linux/net/tipc/link.c (revision f6d08d9d8543c8ee494b307804b28e2750ffedb9)
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  * @prev_from: sequence number of most previous retransmission request
110  * @stale_cnt: counter for number of identical retransmit attempts
111  * @stale_limit: time when repeated identical retransmits must force link reset
112  * @ackers: # of peers that needs to ack each packet before it can be released
113  * @acked: # last packet acked by a certain peer. Used for broadcast.
114  * @rcv_nxt: next sequence number to expect for inbound messages
115  * @deferred_queue: deferred queue saved OOS b'cast message received from node
116  * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
117  * @inputq: buffer queue for messages to be delivered upwards
118  * @namedq: buffer queue for name table messages to be delivered upwards
119  * @next_out: ptr to first unsent outbound message in queue
120  * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
121  * @long_msg_seq_no: next identifier to use for outbound fragmented messages
122  * @reasm_buf: head of partially reassembled inbound message fragments
123  * @bc_rcvr: marks that this is a broadcast receiver link
124  * @stats: collects statistics regarding link activity
125  */
126 struct tipc_link {
127 	u32 addr;
128 	char name[TIPC_MAX_LINK_NAME];
129 	struct net *net;
130 
131 	/* Management and link supervision data */
132 	u16 peer_session;
133 	u16 session;
134 	u16 snd_nxt_state;
135 	u16 rcv_nxt_state;
136 	u32 peer_bearer_id;
137 	u32 bearer_id;
138 	u32 tolerance;
139 	u32 abort_limit;
140 	u32 state;
141 	u16 peer_caps;
142 	bool in_session;
143 	bool active;
144 	u32 silent_intv_cnt;
145 	char if_name[TIPC_MAX_IF_NAME];
146 	u32 priority;
147 	char net_plane;
148 	struct tipc_mon_state mon_state;
149 	u16 rst_cnt;
150 
151 	/* Failover/synch */
152 	u16 drop_point;
153 	struct sk_buff *failover_reasm_skb;
154 	struct sk_buff_head failover_deferdq;
155 
156 	/* Max packet negotiation */
157 	u16 mtu;
158 	u16 advertised_mtu;
159 
160 	/* Sending */
161 	struct sk_buff_head transmq;
162 	struct sk_buff_head backlogq;
163 	struct {
164 		u16 len;
165 		u16 limit;
166 	} backlog[5];
167 	u16 snd_nxt;
168 	u16 prev_from;
169 	u16 window;
170 	u16 stale_cnt;
171 	unsigned long stale_limit;
172 
173 	/* Reception */
174 	u16 rcv_nxt;
175 	u32 rcv_unacked;
176 	struct sk_buff_head deferdq;
177 	struct sk_buff_head *inputq;
178 	struct sk_buff_head *namedq;
179 
180 	/* Congestion handling */
181 	struct sk_buff_head wakeupq;
182 
183 	/* Fragmentation/reassembly */
184 	struct sk_buff *reasm_buf;
185 
186 	/* Broadcast */
187 	u16 ackers;
188 	u16 acked;
189 	struct tipc_link *bc_rcvlink;
190 	struct tipc_link *bc_sndlink;
191 	u8 nack_state;
192 	bool bc_peer_is_up;
193 
194 	/* Statistics */
195 	struct tipc_stats stats;
196 };
197 
198 /*
199  * Error message prefixes
200  */
201 static const char *link_co_err = "Link tunneling error, ";
202 static const char *link_rst_msg = "Resetting link ";
203 
204 /* Send states for broadcast NACKs
205  */
206 enum {
207 	BC_NACK_SND_CONDITIONAL,
208 	BC_NACK_SND_UNCONDITIONAL,
209 	BC_NACK_SND_SUPPRESS,
210 };
211 
212 #define TIPC_BC_RETR_LIM msecs_to_jiffies(10)   /* [ms] */
213 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
214 
215 /*
216  * Interval between NACKs when packets arrive out of order
217  */
218 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
219 
220 /* Link FSM states:
221  */
222 enum {
223 	LINK_ESTABLISHED     = 0xe,
224 	LINK_ESTABLISHING    = 0xe  << 4,
225 	LINK_RESET           = 0x1  << 8,
226 	LINK_RESETTING       = 0x2  << 12,
227 	LINK_PEER_RESET      = 0xd  << 16,
228 	LINK_FAILINGOVER     = 0xf  << 20,
229 	LINK_SYNCHING        = 0xc  << 24
230 };
231 
232 /* Link FSM state checking routines
233  */
234 static int link_is_up(struct tipc_link *l)
235 {
236 	return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
237 }
238 
239 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
240 			       struct sk_buff_head *xmitq);
241 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
242 				      bool probe_reply, u16 rcvgap,
243 				      int tolerance, int priority,
244 				      struct sk_buff_head *xmitq);
245 static void link_print(struct tipc_link *l, const char *str);
246 static int tipc_link_build_nack_msg(struct tipc_link *l,
247 				    struct sk_buff_head *xmitq);
248 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
249 					struct sk_buff_head *xmitq);
250 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
251 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data);
252 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
253 				     struct tipc_gap_ack_blks *ga,
254 				     struct sk_buff_head *xmitq);
255 
256 /*
257  *  Simple non-static link routines (i.e. referenced outside this file)
258  */
259 bool tipc_link_is_up(struct tipc_link *l)
260 {
261 	return link_is_up(l);
262 }
263 
264 bool tipc_link_peer_is_down(struct tipc_link *l)
265 {
266 	return l->state == LINK_PEER_RESET;
267 }
268 
269 bool tipc_link_is_reset(struct tipc_link *l)
270 {
271 	return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
272 }
273 
274 bool tipc_link_is_establishing(struct tipc_link *l)
275 {
276 	return l->state == LINK_ESTABLISHING;
277 }
278 
279 bool tipc_link_is_synching(struct tipc_link *l)
280 {
281 	return l->state == LINK_SYNCHING;
282 }
283 
284 bool tipc_link_is_failingover(struct tipc_link *l)
285 {
286 	return l->state == LINK_FAILINGOVER;
287 }
288 
289 bool tipc_link_is_blocked(struct tipc_link *l)
290 {
291 	return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
292 }
293 
294 static bool link_is_bc_sndlink(struct tipc_link *l)
295 {
296 	return !l->bc_sndlink;
297 }
298 
299 static bool link_is_bc_rcvlink(struct tipc_link *l)
300 {
301 	return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
302 }
303 
304 void tipc_link_set_active(struct tipc_link *l, bool active)
305 {
306 	l->active = active;
307 }
308 
309 u32 tipc_link_id(struct tipc_link *l)
310 {
311 	return l->peer_bearer_id << 16 | l->bearer_id;
312 }
313 
314 int tipc_link_window(struct tipc_link *l)
315 {
316 	return l->window;
317 }
318 
319 int tipc_link_prio(struct tipc_link *l)
320 {
321 	return l->priority;
322 }
323 
324 unsigned long tipc_link_tolerance(struct tipc_link *l)
325 {
326 	return l->tolerance;
327 }
328 
329 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
330 {
331 	return l->inputq;
332 }
333 
334 char tipc_link_plane(struct tipc_link *l)
335 {
336 	return l->net_plane;
337 }
338 
339 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
340 {
341 	l->peer_caps = capabilities;
342 }
343 
344 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
345 			   struct tipc_link *uc_l,
346 			   struct sk_buff_head *xmitq)
347 {
348 	struct tipc_link *rcv_l = uc_l->bc_rcvlink;
349 
350 	snd_l->ackers++;
351 	rcv_l->acked = snd_l->snd_nxt - 1;
352 	snd_l->state = LINK_ESTABLISHED;
353 	tipc_link_build_bc_init_msg(uc_l, xmitq);
354 }
355 
356 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
357 			      struct tipc_link *rcv_l,
358 			      struct sk_buff_head *xmitq)
359 {
360 	u16 ack = snd_l->snd_nxt - 1;
361 
362 	snd_l->ackers--;
363 	rcv_l->bc_peer_is_up = true;
364 	rcv_l->state = LINK_ESTABLISHED;
365 	tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
366 	trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
367 	tipc_link_reset(rcv_l);
368 	rcv_l->state = LINK_RESET;
369 	if (!snd_l->ackers) {
370 		trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
371 		tipc_link_reset(snd_l);
372 		snd_l->state = LINK_RESET;
373 		__skb_queue_purge(xmitq);
374 	}
375 }
376 
377 int tipc_link_bc_peers(struct tipc_link *l)
378 {
379 	return l->ackers;
380 }
381 
382 static u16 link_bc_rcv_gap(struct tipc_link *l)
383 {
384 	struct sk_buff *skb = skb_peek(&l->deferdq);
385 	u16 gap = 0;
386 
387 	if (more(l->snd_nxt, l->rcv_nxt))
388 		gap = l->snd_nxt - l->rcv_nxt;
389 	if (skb)
390 		gap = buf_seqno(skb) - l->rcv_nxt;
391 	return gap;
392 }
393 
394 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
395 {
396 	l->mtu = mtu;
397 }
398 
399 int tipc_link_mtu(struct tipc_link *l)
400 {
401 	return l->mtu;
402 }
403 
404 u16 tipc_link_rcv_nxt(struct tipc_link *l)
405 {
406 	return l->rcv_nxt;
407 }
408 
409 u16 tipc_link_acked(struct tipc_link *l)
410 {
411 	return l->acked;
412 }
413 
414 char *tipc_link_name(struct tipc_link *l)
415 {
416 	return l->name;
417 }
418 
419 u32 tipc_link_state(struct tipc_link *l)
420 {
421 	return l->state;
422 }
423 
424 /**
425  * tipc_link_create - create a new link
426  * @n: pointer to associated node
427  * @if_name: associated interface name
428  * @bearer_id: id (index) of associated bearer
429  * @tolerance: link tolerance to be used by link
430  * @net_plane: network plane (A,B,c..) this link belongs to
431  * @mtu: mtu to be advertised by link
432  * @priority: priority to be used by link
433  * @window: send window to be used by link
434  * @session: session to be used by link
435  * @ownnode: identity of own node
436  * @peer: node id of peer node
437  * @peer_caps: bitmap describing peer node capabilities
438  * @bc_sndlink: the namespace global link used for broadcast sending
439  * @bc_rcvlink: the peer specific link used for broadcast reception
440  * @inputq: queue to put messages ready for delivery
441  * @namedq: queue to put binding table update messages ready for delivery
442  * @link: return value, pointer to put the created link
443  *
444  * Returns true if link was created, otherwise false
445  */
446 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
447 		      int tolerance, char net_plane, u32 mtu, int priority,
448 		      int window, u32 session, u32 self,
449 		      u32 peer, u8 *peer_id, u16 peer_caps,
450 		      struct tipc_link *bc_sndlink,
451 		      struct tipc_link *bc_rcvlink,
452 		      struct sk_buff_head *inputq,
453 		      struct sk_buff_head *namedq,
454 		      struct tipc_link **link)
455 {
456 	char peer_str[NODE_ID_STR_LEN] = {0,};
457 	char self_str[NODE_ID_STR_LEN] = {0,};
458 	struct tipc_link *l;
459 
460 	l = kzalloc(sizeof(*l), GFP_ATOMIC);
461 	if (!l)
462 		return false;
463 	*link = l;
464 	l->session = session;
465 
466 	/* Set link name for unicast links only */
467 	if (peer_id) {
468 		tipc_nodeid2string(self_str, tipc_own_id(net));
469 		if (strlen(self_str) > 16)
470 			sprintf(self_str, "%x", self);
471 		tipc_nodeid2string(peer_str, peer_id);
472 		if (strlen(peer_str) > 16)
473 			sprintf(peer_str, "%x", peer);
474 	}
475 	/* Peer i/f name will be completed by reset/activate message */
476 	snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
477 		 self_str, if_name, peer_str);
478 
479 	strcpy(l->if_name, if_name);
480 	l->addr = peer;
481 	l->peer_caps = peer_caps;
482 	l->net = net;
483 	l->in_session = false;
484 	l->bearer_id = bearer_id;
485 	l->tolerance = tolerance;
486 	if (bc_rcvlink)
487 		bc_rcvlink->tolerance = tolerance;
488 	l->net_plane = net_plane;
489 	l->advertised_mtu = mtu;
490 	l->mtu = mtu;
491 	l->priority = priority;
492 	tipc_link_set_queue_limits(l, window);
493 	l->ackers = 1;
494 	l->bc_sndlink = bc_sndlink;
495 	l->bc_rcvlink = bc_rcvlink;
496 	l->inputq = inputq;
497 	l->namedq = namedq;
498 	l->state = LINK_RESETTING;
499 	__skb_queue_head_init(&l->transmq);
500 	__skb_queue_head_init(&l->backlogq);
501 	__skb_queue_head_init(&l->deferdq);
502 	__skb_queue_head_init(&l->failover_deferdq);
503 	skb_queue_head_init(&l->wakeupq);
504 	skb_queue_head_init(l->inputq);
505 	return true;
506 }
507 
508 /**
509  * tipc_link_bc_create - create new link to be used for broadcast
510  * @n: pointer to associated node
511  * @mtu: mtu to be used initially if no peers
512  * @window: send window to be used
513  * @inputq: queue to put messages ready for delivery
514  * @namedq: queue to put binding table update messages ready for delivery
515  * @link: return value, pointer to put the created link
516  *
517  * Returns true if link was created, otherwise false
518  */
519 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
520 			 int mtu, int window, u16 peer_caps,
521 			 struct sk_buff_head *inputq,
522 			 struct sk_buff_head *namedq,
523 			 struct tipc_link *bc_sndlink,
524 			 struct tipc_link **link)
525 {
526 	struct tipc_link *l;
527 
528 	if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
529 			      0, ownnode, peer, NULL, peer_caps, bc_sndlink,
530 			      NULL, inputq, namedq, link))
531 		return false;
532 
533 	l = *link;
534 	strcpy(l->name, tipc_bclink_name);
535 	trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
536 	tipc_link_reset(l);
537 	l->state = LINK_RESET;
538 	l->ackers = 0;
539 	l->bc_rcvlink = l;
540 
541 	/* Broadcast send link is always up */
542 	if (link_is_bc_sndlink(l))
543 		l->state = LINK_ESTABLISHED;
544 
545 	/* Disable replicast if even a single peer doesn't support it */
546 	if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
547 		tipc_bcast_disable_rcast(net);
548 
549 	return true;
550 }
551 
552 /**
553  * tipc_link_fsm_evt - link finite state machine
554  * @l: pointer to link
555  * @evt: state machine event to be processed
556  */
557 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
558 {
559 	int rc = 0;
560 	int old_state = l->state;
561 
562 	switch (l->state) {
563 	case LINK_RESETTING:
564 		switch (evt) {
565 		case LINK_PEER_RESET_EVT:
566 			l->state = LINK_PEER_RESET;
567 			break;
568 		case LINK_RESET_EVT:
569 			l->state = LINK_RESET;
570 			break;
571 		case LINK_FAILURE_EVT:
572 		case LINK_FAILOVER_BEGIN_EVT:
573 		case LINK_ESTABLISH_EVT:
574 		case LINK_FAILOVER_END_EVT:
575 		case LINK_SYNCH_BEGIN_EVT:
576 		case LINK_SYNCH_END_EVT:
577 		default:
578 			goto illegal_evt;
579 		}
580 		break;
581 	case LINK_RESET:
582 		switch (evt) {
583 		case LINK_PEER_RESET_EVT:
584 			l->state = LINK_ESTABLISHING;
585 			break;
586 		case LINK_FAILOVER_BEGIN_EVT:
587 			l->state = LINK_FAILINGOVER;
588 		case LINK_FAILURE_EVT:
589 		case LINK_RESET_EVT:
590 		case LINK_ESTABLISH_EVT:
591 		case LINK_FAILOVER_END_EVT:
592 			break;
593 		case LINK_SYNCH_BEGIN_EVT:
594 		case LINK_SYNCH_END_EVT:
595 		default:
596 			goto illegal_evt;
597 		}
598 		break;
599 	case LINK_PEER_RESET:
600 		switch (evt) {
601 		case LINK_RESET_EVT:
602 			l->state = LINK_ESTABLISHING;
603 			break;
604 		case LINK_PEER_RESET_EVT:
605 		case LINK_ESTABLISH_EVT:
606 		case LINK_FAILURE_EVT:
607 			break;
608 		case LINK_SYNCH_BEGIN_EVT:
609 		case LINK_SYNCH_END_EVT:
610 		case LINK_FAILOVER_BEGIN_EVT:
611 		case LINK_FAILOVER_END_EVT:
612 		default:
613 			goto illegal_evt;
614 		}
615 		break;
616 	case LINK_FAILINGOVER:
617 		switch (evt) {
618 		case LINK_FAILOVER_END_EVT:
619 			l->state = LINK_RESET;
620 			break;
621 		case LINK_PEER_RESET_EVT:
622 		case LINK_RESET_EVT:
623 		case LINK_ESTABLISH_EVT:
624 		case LINK_FAILURE_EVT:
625 			break;
626 		case LINK_FAILOVER_BEGIN_EVT:
627 		case LINK_SYNCH_BEGIN_EVT:
628 		case LINK_SYNCH_END_EVT:
629 		default:
630 			goto illegal_evt;
631 		}
632 		break;
633 	case LINK_ESTABLISHING:
634 		switch (evt) {
635 		case LINK_ESTABLISH_EVT:
636 			l->state = LINK_ESTABLISHED;
637 			break;
638 		case LINK_FAILOVER_BEGIN_EVT:
639 			l->state = LINK_FAILINGOVER;
640 			break;
641 		case LINK_RESET_EVT:
642 			l->state = LINK_RESET;
643 			break;
644 		case LINK_FAILURE_EVT:
645 		case LINK_PEER_RESET_EVT:
646 		case LINK_SYNCH_BEGIN_EVT:
647 		case LINK_FAILOVER_END_EVT:
648 			break;
649 		case LINK_SYNCH_END_EVT:
650 		default:
651 			goto illegal_evt;
652 		}
653 		break;
654 	case LINK_ESTABLISHED:
655 		switch (evt) {
656 		case LINK_PEER_RESET_EVT:
657 			l->state = LINK_PEER_RESET;
658 			rc |= TIPC_LINK_DOWN_EVT;
659 			break;
660 		case LINK_FAILURE_EVT:
661 			l->state = LINK_RESETTING;
662 			rc |= TIPC_LINK_DOWN_EVT;
663 			break;
664 		case LINK_RESET_EVT:
665 			l->state = LINK_RESET;
666 			break;
667 		case LINK_ESTABLISH_EVT:
668 		case LINK_SYNCH_END_EVT:
669 			break;
670 		case LINK_SYNCH_BEGIN_EVT:
671 			l->state = LINK_SYNCHING;
672 			break;
673 		case LINK_FAILOVER_BEGIN_EVT:
674 		case LINK_FAILOVER_END_EVT:
675 		default:
676 			goto illegal_evt;
677 		}
678 		break;
679 	case LINK_SYNCHING:
680 		switch (evt) {
681 		case LINK_PEER_RESET_EVT:
682 			l->state = LINK_PEER_RESET;
683 			rc |= TIPC_LINK_DOWN_EVT;
684 			break;
685 		case LINK_FAILURE_EVT:
686 			l->state = LINK_RESETTING;
687 			rc |= TIPC_LINK_DOWN_EVT;
688 			break;
689 		case LINK_RESET_EVT:
690 			l->state = LINK_RESET;
691 			break;
692 		case LINK_ESTABLISH_EVT:
693 		case LINK_SYNCH_BEGIN_EVT:
694 			break;
695 		case LINK_SYNCH_END_EVT:
696 			l->state = LINK_ESTABLISHED;
697 			break;
698 		case LINK_FAILOVER_BEGIN_EVT:
699 		case LINK_FAILOVER_END_EVT:
700 		default:
701 			goto illegal_evt;
702 		}
703 		break;
704 	default:
705 		pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
706 	}
707 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
708 	return rc;
709 illegal_evt:
710 	pr_err("Illegal FSM event %x in state %x on link %s\n",
711 	       evt, l->state, l->name);
712 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
713 	return rc;
714 }
715 
716 /* link_profile_stats - update statistical profiling of traffic
717  */
718 static void link_profile_stats(struct tipc_link *l)
719 {
720 	struct sk_buff *skb;
721 	struct tipc_msg *msg;
722 	int length;
723 
724 	/* Update counters used in statistical profiling of send traffic */
725 	l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
726 	l->stats.queue_sz_counts++;
727 
728 	skb = skb_peek(&l->transmq);
729 	if (!skb)
730 		return;
731 	msg = buf_msg(skb);
732 	length = msg_size(msg);
733 
734 	if (msg_user(msg) == MSG_FRAGMENTER) {
735 		if (msg_type(msg) != FIRST_FRAGMENT)
736 			return;
737 		length = msg_size(msg_get_wrapped(msg));
738 	}
739 	l->stats.msg_lengths_total += length;
740 	l->stats.msg_length_counts++;
741 	if (length <= 64)
742 		l->stats.msg_length_profile[0]++;
743 	else if (length <= 256)
744 		l->stats.msg_length_profile[1]++;
745 	else if (length <= 1024)
746 		l->stats.msg_length_profile[2]++;
747 	else if (length <= 4096)
748 		l->stats.msg_length_profile[3]++;
749 	else if (length <= 16384)
750 		l->stats.msg_length_profile[4]++;
751 	else if (length <= 32768)
752 		l->stats.msg_length_profile[5]++;
753 	else
754 		l->stats.msg_length_profile[6]++;
755 }
756 
757 /**
758  * tipc_link_too_silent - check if link is "too silent"
759  * @l: tipc link to be checked
760  *
761  * Returns true if the link 'silent_intv_cnt' is about to reach the
762  * 'abort_limit' value, otherwise false
763  */
764 bool tipc_link_too_silent(struct tipc_link *l)
765 {
766 	return (l->silent_intv_cnt + 2 > l->abort_limit);
767 }
768 
769 /* tipc_link_timeout - perform periodic task as instructed from node timeout
770  */
771 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
772 {
773 	int mtyp = 0;
774 	int rc = 0;
775 	bool state = false;
776 	bool probe = false;
777 	bool setup = false;
778 	u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
779 	u16 bc_acked = l->bc_rcvlink->acked;
780 	struct tipc_mon_state *mstate = &l->mon_state;
781 
782 	trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
783 	trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
784 	switch (l->state) {
785 	case LINK_ESTABLISHED:
786 	case LINK_SYNCHING:
787 		mtyp = STATE_MSG;
788 		link_profile_stats(l);
789 		tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
790 		if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
791 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
792 		state = bc_acked != bc_snt;
793 		state |= l->bc_rcvlink->rcv_unacked;
794 		state |= l->rcv_unacked;
795 		state |= !skb_queue_empty(&l->transmq);
796 		state |= !skb_queue_empty(&l->deferdq);
797 		probe = mstate->probing;
798 		probe |= l->silent_intv_cnt;
799 		if (probe || mstate->monitoring)
800 			l->silent_intv_cnt++;
801 		break;
802 	case LINK_RESET:
803 		setup = l->rst_cnt++ <= 4;
804 		setup |= !(l->rst_cnt % 16);
805 		mtyp = RESET_MSG;
806 		break;
807 	case LINK_ESTABLISHING:
808 		setup = true;
809 		mtyp = ACTIVATE_MSG;
810 		break;
811 	case LINK_PEER_RESET:
812 	case LINK_RESETTING:
813 	case LINK_FAILINGOVER:
814 		break;
815 	default:
816 		break;
817 	}
818 
819 	if (state || probe || setup)
820 		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
821 
822 	return rc;
823 }
824 
825 /**
826  * link_schedule_user - schedule a message sender for wakeup after congestion
827  * @l: congested link
828  * @hdr: header of message that is being sent
829  * Create pseudo msg to send back to user when congestion abates
830  */
831 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
832 {
833 	u32 dnode = tipc_own_addr(l->net);
834 	u32 dport = msg_origport(hdr);
835 	struct sk_buff *skb;
836 
837 	/* Create and schedule wakeup pseudo message */
838 	skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
839 			      dnode, l->addr, dport, 0, 0);
840 	if (!skb)
841 		return -ENOBUFS;
842 	msg_set_dest_droppable(buf_msg(skb), true);
843 	TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
844 	skb_queue_tail(&l->wakeupq, skb);
845 	l->stats.link_congs++;
846 	trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
847 	return -ELINKCONG;
848 }
849 
850 /**
851  * link_prepare_wakeup - prepare users for wakeup after congestion
852  * @l: congested link
853  * Wake up a number of waiting users, as permitted by available space
854  * in the send queue
855  */
856 static void link_prepare_wakeup(struct tipc_link *l)
857 {
858 	struct sk_buff *skb, *tmp;
859 	int imp, i = 0;
860 
861 	skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
862 		imp = TIPC_SKB_CB(skb)->chain_imp;
863 		if (l->backlog[imp].len < l->backlog[imp].limit) {
864 			skb_unlink(skb, &l->wakeupq);
865 			skb_queue_tail(l->inputq, skb);
866 		} else if (i++ > 10) {
867 			break;
868 		}
869 	}
870 }
871 
872 void tipc_link_reset(struct tipc_link *l)
873 {
874 	struct sk_buff_head list;
875 
876 	__skb_queue_head_init(&list);
877 
878 	l->in_session = false;
879 	/* Force re-synch of peer session number before establishing */
880 	l->peer_session--;
881 	l->session++;
882 	l->mtu = l->advertised_mtu;
883 
884 	spin_lock_bh(&l->wakeupq.lock);
885 	skb_queue_splice_init(&l->wakeupq, &list);
886 	spin_unlock_bh(&l->wakeupq.lock);
887 
888 	spin_lock_bh(&l->inputq->lock);
889 	skb_queue_splice_init(&list, l->inputq);
890 	spin_unlock_bh(&l->inputq->lock);
891 
892 	__skb_queue_purge(&l->transmq);
893 	__skb_queue_purge(&l->deferdq);
894 	__skb_queue_purge(&l->backlogq);
895 	__skb_queue_purge(&l->failover_deferdq);
896 	l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
897 	l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
898 	l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
899 	l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
900 	l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
901 	kfree_skb(l->reasm_buf);
902 	kfree_skb(l->failover_reasm_skb);
903 	l->reasm_buf = NULL;
904 	l->failover_reasm_skb = NULL;
905 	l->rcv_unacked = 0;
906 	l->snd_nxt = 1;
907 	l->rcv_nxt = 1;
908 	l->snd_nxt_state = 1;
909 	l->rcv_nxt_state = 1;
910 	l->acked = 0;
911 	l->silent_intv_cnt = 0;
912 	l->rst_cnt = 0;
913 	l->stale_cnt = 0;
914 	l->bc_peer_is_up = false;
915 	memset(&l->mon_state, 0, sizeof(l->mon_state));
916 	tipc_link_reset_stats(l);
917 }
918 
919 /**
920  * tipc_link_xmit(): enqueue buffer list according to queue situation
921  * @link: link to use
922  * @list: chain of buffers containing message
923  * @xmitq: returned list of packets to be sent by caller
924  *
925  * Consumes the buffer chain.
926  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
927  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
928  */
929 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
930 		   struct sk_buff_head *xmitq)
931 {
932 	struct tipc_msg *hdr = buf_msg(skb_peek(list));
933 	unsigned int maxwin = l->window;
934 	int imp = msg_importance(hdr);
935 	unsigned int mtu = l->mtu;
936 	u16 ack = l->rcv_nxt - 1;
937 	u16 seqno = l->snd_nxt;
938 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
939 	struct sk_buff_head *transmq = &l->transmq;
940 	struct sk_buff_head *backlogq = &l->backlogq;
941 	struct sk_buff *skb, *_skb, *bskb;
942 	int pkt_cnt = skb_queue_len(list);
943 	int rc = 0;
944 
945 	if (unlikely(msg_size(hdr) > mtu)) {
946 		skb_queue_purge(list);
947 		return -EMSGSIZE;
948 	}
949 
950 	/* Allow oversubscription of one data msg per source at congestion */
951 	if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
952 		if (imp == TIPC_SYSTEM_IMPORTANCE) {
953 			pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
954 			return -ENOBUFS;
955 		}
956 		rc = link_schedule_user(l, hdr);
957 	}
958 
959 	if (pkt_cnt > 1) {
960 		l->stats.sent_fragmented++;
961 		l->stats.sent_fragments += pkt_cnt;
962 	}
963 
964 	/* Prepare each packet for sending, and add to relevant queue: */
965 	while (skb_queue_len(list)) {
966 		skb = skb_peek(list);
967 		hdr = buf_msg(skb);
968 		msg_set_seqno(hdr, seqno);
969 		msg_set_ack(hdr, ack);
970 		msg_set_bcast_ack(hdr, bc_ack);
971 
972 		if (likely(skb_queue_len(transmq) < maxwin)) {
973 			_skb = skb_clone(skb, GFP_ATOMIC);
974 			if (!_skb) {
975 				skb_queue_purge(list);
976 				return -ENOBUFS;
977 			}
978 			__skb_dequeue(list);
979 			__skb_queue_tail(transmq, skb);
980 			/* next retransmit attempt */
981 			if (link_is_bc_sndlink(l))
982 				TIPC_SKB_CB(skb)->nxt_retr =
983 					jiffies + TIPC_BC_RETR_LIM;
984 			__skb_queue_tail(xmitq, _skb);
985 			TIPC_SKB_CB(skb)->ackers = l->ackers;
986 			l->rcv_unacked = 0;
987 			l->stats.sent_pkts++;
988 			seqno++;
989 			continue;
990 		}
991 		if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
992 			kfree_skb(__skb_dequeue(list));
993 			l->stats.sent_bundled++;
994 			continue;
995 		}
996 		if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
997 			kfree_skb(__skb_dequeue(list));
998 			__skb_queue_tail(backlogq, bskb);
999 			l->backlog[msg_importance(buf_msg(bskb))].len++;
1000 			l->stats.sent_bundled++;
1001 			l->stats.sent_bundles++;
1002 			continue;
1003 		}
1004 		l->backlog[imp].len += skb_queue_len(list);
1005 		skb_queue_splice_tail_init(list, backlogq);
1006 	}
1007 	l->snd_nxt = seqno;
1008 	return rc;
1009 }
1010 
1011 static void tipc_link_advance_backlog(struct tipc_link *l,
1012 				      struct sk_buff_head *xmitq)
1013 {
1014 	struct sk_buff *skb, *_skb;
1015 	struct tipc_msg *hdr;
1016 	u16 seqno = l->snd_nxt;
1017 	u16 ack = l->rcv_nxt - 1;
1018 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1019 
1020 	while (skb_queue_len(&l->transmq) < l->window) {
1021 		skb = skb_peek(&l->backlogq);
1022 		if (!skb)
1023 			break;
1024 		_skb = skb_clone(skb, GFP_ATOMIC);
1025 		if (!_skb)
1026 			break;
1027 		__skb_dequeue(&l->backlogq);
1028 		hdr = buf_msg(skb);
1029 		l->backlog[msg_importance(hdr)].len--;
1030 		__skb_queue_tail(&l->transmq, skb);
1031 		/* next retransmit attempt */
1032 		if (link_is_bc_sndlink(l))
1033 			TIPC_SKB_CB(skb)->nxt_retr = jiffies + TIPC_BC_RETR_LIM;
1034 
1035 		__skb_queue_tail(xmitq, _skb);
1036 		TIPC_SKB_CB(skb)->ackers = l->ackers;
1037 		msg_set_seqno(hdr, seqno);
1038 		msg_set_ack(hdr, ack);
1039 		msg_set_bcast_ack(hdr, bc_ack);
1040 		l->rcv_unacked = 0;
1041 		l->stats.sent_pkts++;
1042 		seqno++;
1043 	}
1044 	l->snd_nxt = seqno;
1045 }
1046 
1047 /**
1048  * link_retransmit_failure() - Detect repeated retransmit failures
1049  * @l: tipc link sender
1050  * @r: tipc link receiver (= l in case of unicast)
1051  * @from: seqno of the 1st packet in retransmit request
1052  * @rc: returned code
1053  *
1054  * Return: true if the repeated retransmit failures happens, otherwise
1055  * false
1056  */
1057 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1058 				    u16 from, int *rc)
1059 {
1060 	struct sk_buff *skb = skb_peek(&l->transmq);
1061 	struct tipc_msg *hdr;
1062 
1063 	if (!skb)
1064 		return false;
1065 	hdr = buf_msg(skb);
1066 
1067 	/* Detect repeated retransmit failures on same packet */
1068 	if (r->prev_from != from) {
1069 		r->prev_from = from;
1070 		r->stale_limit = jiffies + msecs_to_jiffies(r->tolerance);
1071 		r->stale_cnt = 0;
1072 	} else if (++r->stale_cnt > 99 && time_after(jiffies, r->stale_limit)) {
1073 		pr_warn("Retransmission failure on link <%s>\n", l->name);
1074 		link_print(l, "State of link ");
1075 		pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1076 			msg_user(hdr), msg_type(hdr), msg_size(hdr),
1077 			msg_errcode(hdr));
1078 		pr_info("sqno %u, prev: %x, src: %x\n",
1079 			msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
1080 
1081 		trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1082 		trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1083 		trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1084 
1085 		if (link_is_bc_sndlink(l))
1086 			*rc = TIPC_LINK_DOWN_EVT;
1087 
1088 		*rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1089 		return true;
1090 	}
1091 
1092 	return false;
1093 }
1094 
1095 /* tipc_link_bc_retrans() - retransmit zero or more packets
1096  * @l: the link to transmit on
1097  * @r: the receiving link ordering the retransmit. Same as l if unicast
1098  * @from: retransmit from (inclusive) this sequence number
1099  * @to: retransmit to (inclusive) this sequence number
1100  * xmitq: queue for accumulating the retransmitted packets
1101  */
1102 static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r,
1103 				u16 from, u16 to, struct sk_buff_head *xmitq)
1104 {
1105 	struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1106 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1107 	u16 ack = l->rcv_nxt - 1;
1108 	struct tipc_msg *hdr;
1109 	int rc = 0;
1110 
1111 	if (!skb)
1112 		return 0;
1113 	if (less(to, from))
1114 		return 0;
1115 
1116 	trace_tipc_link_retrans(r, from, to, &l->transmq);
1117 
1118 	if (link_retransmit_failure(l, r, from, &rc))
1119 		return rc;
1120 
1121 	skb_queue_walk(&l->transmq, skb) {
1122 		hdr = buf_msg(skb);
1123 		if (less(msg_seqno(hdr), from))
1124 			continue;
1125 		if (more(msg_seqno(hdr), to))
1126 			break;
1127 		if (link_is_bc_sndlink(l)) {
1128 			if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1129 				continue;
1130 			TIPC_SKB_CB(skb)->nxt_retr = jiffies + TIPC_BC_RETR_LIM;
1131 		}
1132 		_skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1133 		if (!_skb)
1134 			return 0;
1135 		hdr = buf_msg(_skb);
1136 		msg_set_ack(hdr, ack);
1137 		msg_set_bcast_ack(hdr, bc_ack);
1138 		_skb->priority = TC_PRIO_CONTROL;
1139 		__skb_queue_tail(xmitq, _skb);
1140 		l->stats.retransmitted++;
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 	u16 seqno, n = 0;
1366 	int rc = 0;
1367 
1368 	if (gap && link_retransmit_failure(l, l, acked + 1, &rc))
1369 		return rc;
1370 
1371 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1372 		seqno = buf_seqno(skb);
1373 
1374 next_gap_ack:
1375 		if (less_eq(seqno, acked)) {
1376 			/* release skb */
1377 			__skb_unlink(skb, &l->transmq);
1378 			kfree_skb(skb);
1379 		} else if (less_eq(seqno, acked + gap)) {
1380 			/* retransmit skb */
1381 			if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1382 				continue;
1383 			TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
1384 
1385 			_skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1386 			if (!_skb)
1387 				continue;
1388 			hdr = buf_msg(_skb);
1389 			msg_set_ack(hdr, ack);
1390 			msg_set_bcast_ack(hdr, bc_ack);
1391 			_skb->priority = TC_PRIO_CONTROL;
1392 			__skb_queue_tail(xmitq, _skb);
1393 			l->stats.retransmitted++;
1394 		} else {
1395 			/* retry with Gap ACK blocks if any */
1396 			if (!ga || n >= ga->gack_cnt)
1397 				break;
1398 			acked = ntohs(ga->gacks[n].ack);
1399 			gap = ntohs(ga->gacks[n].gap);
1400 			n++;
1401 			goto next_gap_ack;
1402 		}
1403 	}
1404 
1405 	return 0;
1406 }
1407 
1408 /* tipc_link_build_state_msg: prepare link state message for transmission
1409  *
1410  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1411  * risk of ack storms towards the sender
1412  */
1413 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1414 {
1415 	if (!l)
1416 		return 0;
1417 
1418 	/* Broadcast ACK must be sent via a unicast link => defer to caller */
1419 	if (link_is_bc_rcvlink(l)) {
1420 		if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1421 			return 0;
1422 		l->rcv_unacked = 0;
1423 
1424 		/* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1425 		l->snd_nxt = l->rcv_nxt;
1426 		return TIPC_LINK_SND_STATE;
1427 	}
1428 
1429 	/* Unicast ACK */
1430 	l->rcv_unacked = 0;
1431 	l->stats.sent_acks++;
1432 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1433 	return 0;
1434 }
1435 
1436 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1437  */
1438 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1439 {
1440 	int mtyp = RESET_MSG;
1441 	struct sk_buff *skb;
1442 
1443 	if (l->state == LINK_ESTABLISHING)
1444 		mtyp = ACTIVATE_MSG;
1445 
1446 	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1447 
1448 	/* Inform peer that this endpoint is going down if applicable */
1449 	skb = skb_peek_tail(xmitq);
1450 	if (skb && (l->state == LINK_RESET))
1451 		msg_set_peer_stopping(buf_msg(skb), 1);
1452 }
1453 
1454 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1455  * Note that sending of broadcast NACK is coordinated among nodes, to
1456  * reduce the risk of NACK storms towards the sender
1457  */
1458 static int tipc_link_build_nack_msg(struct tipc_link *l,
1459 				    struct sk_buff_head *xmitq)
1460 {
1461 	u32 def_cnt = ++l->stats.deferred_recv;
1462 	u32 defq_len = skb_queue_len(&l->deferdq);
1463 	int match1, match2;
1464 
1465 	if (link_is_bc_rcvlink(l)) {
1466 		match1 = def_cnt & 0xf;
1467 		match2 = tipc_own_addr(l->net) & 0xf;
1468 		if (match1 == match2)
1469 			return TIPC_LINK_SND_STATE;
1470 		return 0;
1471 	}
1472 
1473 	if (defq_len >= 3 && !((defq_len - 3) % 16))
1474 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1475 	return 0;
1476 }
1477 
1478 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1479  * @l: the link that should handle the message
1480  * @skb: TIPC packet
1481  * @xmitq: queue to place packets to be sent after this call
1482  */
1483 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1484 		  struct sk_buff_head *xmitq)
1485 {
1486 	struct sk_buff_head *defq = &l->deferdq;
1487 	struct tipc_msg *hdr = buf_msg(skb);
1488 	u16 seqno, rcv_nxt, win_lim;
1489 	int rc = 0;
1490 
1491 	/* Verify and update link state */
1492 	if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1493 		return tipc_link_proto_rcv(l, skb, xmitq);
1494 
1495 	/* Don't send probe at next timeout expiration */
1496 	l->silent_intv_cnt = 0;
1497 
1498 	do {
1499 		hdr = buf_msg(skb);
1500 		seqno = msg_seqno(hdr);
1501 		rcv_nxt = l->rcv_nxt;
1502 		win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1503 
1504 		if (unlikely(!link_is_up(l))) {
1505 			if (l->state == LINK_ESTABLISHING)
1506 				rc = TIPC_LINK_UP_EVT;
1507 			goto drop;
1508 		}
1509 
1510 		/* Drop if outside receive window */
1511 		if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1512 			l->stats.duplicates++;
1513 			goto drop;
1514 		}
1515 
1516 		/* Forward queues and wake up waiting users */
1517 		if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1518 			l->stale_cnt = 0;
1519 			tipc_link_advance_backlog(l, xmitq);
1520 			if (unlikely(!skb_queue_empty(&l->wakeupq)))
1521 				link_prepare_wakeup(l);
1522 		}
1523 
1524 		/* Defer delivery if sequence gap */
1525 		if (unlikely(seqno != rcv_nxt)) {
1526 			__tipc_skb_queue_sorted(defq, seqno, skb);
1527 			rc |= tipc_link_build_nack_msg(l, xmitq);
1528 			break;
1529 		}
1530 
1531 		/* Deliver packet */
1532 		l->rcv_nxt++;
1533 		l->stats.recv_pkts++;
1534 
1535 		if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1536 			rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1537 		else if (!tipc_data_input(l, skb, l->inputq))
1538 			rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1539 		if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1540 			rc |= tipc_link_build_state_msg(l, xmitq);
1541 		if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1542 			break;
1543 	} while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1544 
1545 	return rc;
1546 drop:
1547 	kfree_skb(skb);
1548 	return rc;
1549 }
1550 
1551 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1552 				      bool probe_reply, u16 rcvgap,
1553 				      int tolerance, int priority,
1554 				      struct sk_buff_head *xmitq)
1555 {
1556 	struct tipc_link *bcl = l->bc_rcvlink;
1557 	struct sk_buff *skb;
1558 	struct tipc_msg *hdr;
1559 	struct sk_buff_head *dfq = &l->deferdq;
1560 	bool node_up = link_is_up(bcl);
1561 	struct tipc_mon_state *mstate = &l->mon_state;
1562 	int dlen = 0;
1563 	void *data;
1564 	u16 glen = 0;
1565 
1566 	/* Don't send protocol message during reset or link failover */
1567 	if (tipc_link_is_blocked(l))
1568 		return;
1569 
1570 	if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1571 		return;
1572 
1573 	if (!skb_queue_empty(dfq))
1574 		rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1575 
1576 	skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1577 			      tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1578 			      l->addr, tipc_own_addr(l->net), 0, 0, 0);
1579 	if (!skb)
1580 		return;
1581 
1582 	hdr = buf_msg(skb);
1583 	data = msg_data(hdr);
1584 	msg_set_session(hdr, l->session);
1585 	msg_set_bearer_id(hdr, l->bearer_id);
1586 	msg_set_net_plane(hdr, l->net_plane);
1587 	msg_set_next_sent(hdr, l->snd_nxt);
1588 	msg_set_ack(hdr, l->rcv_nxt - 1);
1589 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1590 	msg_set_bc_ack_invalid(hdr, !node_up);
1591 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1592 	msg_set_link_tolerance(hdr, tolerance);
1593 	msg_set_linkprio(hdr, priority);
1594 	msg_set_redundant_link(hdr, node_up);
1595 	msg_set_seq_gap(hdr, 0);
1596 	msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1597 
1598 	if (mtyp == STATE_MSG) {
1599 		if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1600 			msg_set_seqno(hdr, l->snd_nxt_state++);
1601 		msg_set_seq_gap(hdr, rcvgap);
1602 		msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1603 		msg_set_probe(hdr, probe);
1604 		msg_set_is_keepalive(hdr, probe || probe_reply);
1605 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1606 			glen = tipc_build_gap_ack_blks(l, data);
1607 		tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1608 		msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1609 		skb_trim(skb, INT_H_SIZE + glen + dlen);
1610 		l->stats.sent_states++;
1611 		l->rcv_unacked = 0;
1612 	} else {
1613 		/* RESET_MSG or ACTIVATE_MSG */
1614 		if (mtyp == ACTIVATE_MSG) {
1615 			msg_set_dest_session_valid(hdr, 1);
1616 			msg_set_dest_session(hdr, l->peer_session);
1617 		}
1618 		msg_set_max_pkt(hdr, l->advertised_mtu);
1619 		strcpy(data, l->if_name);
1620 		msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1621 		skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1622 	}
1623 	if (probe)
1624 		l->stats.sent_probes++;
1625 	if (rcvgap)
1626 		l->stats.sent_nacks++;
1627 	skb->priority = TC_PRIO_CONTROL;
1628 	__skb_queue_tail(xmitq, skb);
1629 	trace_tipc_proto_build(skb, false, l->name);
1630 }
1631 
1632 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1633 				    struct sk_buff_head *xmitq)
1634 {
1635 	u32 onode = tipc_own_addr(l->net);
1636 	struct tipc_msg *hdr, *ihdr;
1637 	struct sk_buff_head tnlq;
1638 	struct sk_buff *skb;
1639 	u32 dnode = l->addr;
1640 
1641 	skb_queue_head_init(&tnlq);
1642 	skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1643 			      INT_H_SIZE, BASIC_H_SIZE,
1644 			      dnode, onode, 0, 0, 0);
1645 	if (!skb) {
1646 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1647 		return;
1648 	}
1649 
1650 	hdr = buf_msg(skb);
1651 	msg_set_msgcnt(hdr, 1);
1652 	msg_set_bearer_id(hdr, l->peer_bearer_id);
1653 
1654 	ihdr = (struct tipc_msg *)msg_data(hdr);
1655 	tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1656 		      BASIC_H_SIZE, dnode);
1657 	msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1658 	__skb_queue_tail(&tnlq, skb);
1659 	tipc_link_xmit(l, &tnlq, xmitq);
1660 }
1661 
1662 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1663  * with contents of the link's transmit and backlog queues.
1664  */
1665 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1666 			   int mtyp, struct sk_buff_head *xmitq)
1667 {
1668 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1669 	struct sk_buff *skb, *tnlskb;
1670 	struct tipc_msg *hdr, tnlhdr;
1671 	struct sk_buff_head *queue = &l->transmq;
1672 	struct sk_buff_head tmpxq, tnlq;
1673 	u16 pktlen, pktcnt, seqno = l->snd_nxt;
1674 
1675 	if (!tnl)
1676 		return;
1677 
1678 	skb_queue_head_init(&tnlq);
1679 	skb_queue_head_init(&tmpxq);
1680 
1681 	/* At least one packet required for safe algorithm => add dummy */
1682 	skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1683 			      BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1684 			      0, 0, TIPC_ERR_NO_PORT);
1685 	if (!skb) {
1686 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1687 		return;
1688 	}
1689 	skb_queue_tail(&tnlq, skb);
1690 	tipc_link_xmit(l, &tnlq, &tmpxq);
1691 	__skb_queue_purge(&tmpxq);
1692 
1693 	/* Initialize reusable tunnel packet header */
1694 	tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1695 		      mtyp, INT_H_SIZE, l->addr);
1696 	if (mtyp == SYNCH_MSG)
1697 		pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
1698 	else
1699 		pktcnt = skb_queue_len(&l->transmq);
1700 	pktcnt += skb_queue_len(&l->backlogq);
1701 	msg_set_msgcnt(&tnlhdr, pktcnt);
1702 	msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1703 tnl:
1704 	/* Wrap each packet into a tunnel packet */
1705 	skb_queue_walk(queue, skb) {
1706 		hdr = buf_msg(skb);
1707 		if (queue == &l->backlogq)
1708 			msg_set_seqno(hdr, seqno++);
1709 		pktlen = msg_size(hdr);
1710 		msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1711 		tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1712 		if (!tnlskb) {
1713 			pr_warn("%sunable to send packet\n", link_co_err);
1714 			return;
1715 		}
1716 		skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1717 		skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1718 		__skb_queue_tail(&tnlq, tnlskb);
1719 	}
1720 	if (queue != &l->backlogq) {
1721 		queue = &l->backlogq;
1722 		goto tnl;
1723 	}
1724 
1725 	tipc_link_xmit(tnl, &tnlq, xmitq);
1726 
1727 	if (mtyp == FAILOVER_MSG) {
1728 		tnl->drop_point = l->rcv_nxt;
1729 		tnl->failover_reasm_skb = l->reasm_buf;
1730 		l->reasm_buf = NULL;
1731 
1732 		/* Failover the link's deferdq */
1733 		if (unlikely(!skb_queue_empty(fdefq))) {
1734 			pr_warn("Link failover deferdq not empty: %d!\n",
1735 				skb_queue_len(fdefq));
1736 			__skb_queue_purge(fdefq);
1737 		}
1738 		skb_queue_splice_init(&l->deferdq, fdefq);
1739 	}
1740 }
1741 
1742 /**
1743  * tipc_link_failover_prepare() - prepare tnl for link failover
1744  *
1745  * This is a special version of the precursor - tipc_link_tnl_prepare(),
1746  * see the tipc_node_link_failover() for details
1747  *
1748  * @l: failover link
1749  * @tnl: tunnel link
1750  * @xmitq: queue for messages to be xmited
1751  */
1752 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
1753 				struct sk_buff_head *xmitq)
1754 {
1755 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1756 
1757 	tipc_link_create_dummy_tnl_msg(tnl, xmitq);
1758 
1759 	/* This failover link enpoint was never established before,
1760 	 * so it has not received anything from peer.
1761 	 * Otherwise, it must be a normal failover situation or the
1762 	 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
1763 	 * would have to start over from scratch instead.
1764 	 */
1765 	tnl->drop_point = 1;
1766 	tnl->failover_reasm_skb = NULL;
1767 
1768 	/* Initiate the link's failover deferdq */
1769 	if (unlikely(!skb_queue_empty(fdefq))) {
1770 		pr_warn("Link failover deferdq not empty: %d!\n",
1771 			skb_queue_len(fdefq));
1772 		__skb_queue_purge(fdefq);
1773 	}
1774 }
1775 
1776 /* tipc_link_validate_msg(): validate message against current link state
1777  * Returns true if message should be accepted, otherwise false
1778  */
1779 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
1780 {
1781 	u16 curr_session = l->peer_session;
1782 	u16 session = msg_session(hdr);
1783 	int mtyp = msg_type(hdr);
1784 
1785 	if (msg_user(hdr) != LINK_PROTOCOL)
1786 		return true;
1787 
1788 	switch (mtyp) {
1789 	case RESET_MSG:
1790 		if (!l->in_session)
1791 			return true;
1792 		/* Accept only RESET with new session number */
1793 		return more(session, curr_session);
1794 	case ACTIVATE_MSG:
1795 		if (!l->in_session)
1796 			return true;
1797 		/* Accept only ACTIVATE with new or current session number */
1798 		return !less(session, curr_session);
1799 	case STATE_MSG:
1800 		/* Accept only STATE with current session number */
1801 		if (!l->in_session)
1802 			return false;
1803 		if (session != curr_session)
1804 			return false;
1805 		/* Extra sanity check */
1806 		if (!link_is_up(l) && msg_ack(hdr))
1807 			return false;
1808 		if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
1809 			return true;
1810 		/* Accept only STATE with new sequence number */
1811 		return !less(msg_seqno(hdr), l->rcv_nxt_state);
1812 	default:
1813 		return false;
1814 	}
1815 }
1816 
1817 /* tipc_link_proto_rcv(): receive link level protocol message :
1818  * Note that network plane id propagates through the network, and may
1819  * change at any time. The node with lowest numerical id determines
1820  * network plane
1821  */
1822 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1823 			       struct sk_buff_head *xmitq)
1824 {
1825 	struct tipc_msg *hdr = buf_msg(skb);
1826 	struct tipc_gap_ack_blks *ga = NULL;
1827 	u16 rcvgap = 0;
1828 	u16 ack = msg_ack(hdr);
1829 	u16 gap = msg_seq_gap(hdr);
1830 	u16 peers_snd_nxt =  msg_next_sent(hdr);
1831 	u16 peers_tol = msg_link_tolerance(hdr);
1832 	u16 peers_prio = msg_linkprio(hdr);
1833 	u16 rcv_nxt = l->rcv_nxt;
1834 	u16 dlen = msg_data_sz(hdr);
1835 	int mtyp = msg_type(hdr);
1836 	bool reply = msg_probe(hdr);
1837 	u16 glen = 0;
1838 	void *data;
1839 	char *if_name;
1840 	int rc = 0;
1841 
1842 	trace_tipc_proto_rcv(skb, false, l->name);
1843 	if (tipc_link_is_blocked(l) || !xmitq)
1844 		goto exit;
1845 
1846 	if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1847 		l->net_plane = msg_net_plane(hdr);
1848 
1849 	skb_linearize(skb);
1850 	hdr = buf_msg(skb);
1851 	data = msg_data(hdr);
1852 
1853 	if (!tipc_link_validate_msg(l, hdr)) {
1854 		trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
1855 		trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
1856 		goto exit;
1857 	}
1858 
1859 	switch (mtyp) {
1860 	case RESET_MSG:
1861 	case ACTIVATE_MSG:
1862 		/* Complete own link name with peer's interface name */
1863 		if_name =  strrchr(l->name, ':') + 1;
1864 		if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1865 			break;
1866 		if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1867 			break;
1868 		strncpy(if_name, data, TIPC_MAX_IF_NAME);
1869 
1870 		/* Update own tolerance if peer indicates a non-zero value */
1871 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1872 			l->tolerance = peers_tol;
1873 			l->bc_rcvlink->tolerance = peers_tol;
1874 		}
1875 		/* Update own priority if peer's priority is higher */
1876 		if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1877 			l->priority = peers_prio;
1878 
1879 		/* If peer is going down we want full re-establish cycle */
1880 		if (msg_peer_stopping(hdr)) {
1881 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1882 			break;
1883 		}
1884 
1885 		/* If this endpoint was re-created while peer was ESTABLISHING
1886 		 * it doesn't know current session number. Force re-synch.
1887 		 */
1888 		if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
1889 		    l->session != msg_dest_session(hdr)) {
1890 			if (less(l->session, msg_dest_session(hdr)))
1891 				l->session = msg_dest_session(hdr) + 1;
1892 			break;
1893 		}
1894 
1895 		/* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1896 		if (mtyp == RESET_MSG || !link_is_up(l))
1897 			rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1898 
1899 		/* ACTIVATE_MSG takes up link if it was already locally reset */
1900 		if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
1901 			rc = TIPC_LINK_UP_EVT;
1902 
1903 		l->peer_session = msg_session(hdr);
1904 		l->in_session = true;
1905 		l->peer_bearer_id = msg_bearer_id(hdr);
1906 		if (l->mtu > msg_max_pkt(hdr))
1907 			l->mtu = msg_max_pkt(hdr);
1908 		break;
1909 
1910 	case STATE_MSG:
1911 		l->rcv_nxt_state = msg_seqno(hdr) + 1;
1912 
1913 		/* Update own tolerance if peer indicates a non-zero value */
1914 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1915 			l->tolerance = peers_tol;
1916 			l->bc_rcvlink->tolerance = peers_tol;
1917 		}
1918 		/* Update own prio if peer indicates a different value */
1919 		if ((peers_prio != l->priority) &&
1920 		    in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
1921 			l->priority = peers_prio;
1922 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1923 		}
1924 
1925 		l->silent_intv_cnt = 0;
1926 		l->stats.recv_states++;
1927 		if (msg_probe(hdr))
1928 			l->stats.recv_probes++;
1929 
1930 		if (!link_is_up(l)) {
1931 			if (l->state == LINK_ESTABLISHING)
1932 				rc = TIPC_LINK_UP_EVT;
1933 			break;
1934 		}
1935 
1936 		/* Receive Gap ACK blocks from peer if any */
1937 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
1938 			ga = (struct tipc_gap_ack_blks *)data;
1939 			glen = ntohs(ga->len);
1940 			/* sanity check: if failed, ignore Gap ACK blocks */
1941 			if (glen != tipc_gap_ack_blks_sz(ga->gack_cnt))
1942 				ga = NULL;
1943 		}
1944 
1945 		tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
1946 			     &l->mon_state, l->bearer_id);
1947 
1948 		/* Send NACK if peer has sent pkts we haven't received yet */
1949 		if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1950 			rcvgap = peers_snd_nxt - l->rcv_nxt;
1951 		if (rcvgap || reply)
1952 			tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
1953 						  rcvgap, 0, 0, xmitq);
1954 
1955 		rc |= tipc_link_advance_transmq(l, ack, gap, ga, xmitq);
1956 
1957 		/* If NACK, retransmit will now start at right position */
1958 		if (gap)
1959 			l->stats.recv_nacks++;
1960 
1961 		tipc_link_advance_backlog(l, xmitq);
1962 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
1963 			link_prepare_wakeup(l);
1964 	}
1965 exit:
1966 	kfree_skb(skb);
1967 	return rc;
1968 }
1969 
1970 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1971  */
1972 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1973 					 u16 peers_snd_nxt,
1974 					 struct sk_buff_head *xmitq)
1975 {
1976 	struct sk_buff *skb;
1977 	struct tipc_msg *hdr;
1978 	struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1979 	u16 ack = l->rcv_nxt - 1;
1980 	u16 gap_to = peers_snd_nxt - 1;
1981 
1982 	skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1983 			      0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1984 	if (!skb)
1985 		return false;
1986 	hdr = buf_msg(skb);
1987 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1988 	msg_set_bcast_ack(hdr, ack);
1989 	msg_set_bcgap_after(hdr, ack);
1990 	if (dfrd_skb)
1991 		gap_to = buf_seqno(dfrd_skb) - 1;
1992 	msg_set_bcgap_to(hdr, gap_to);
1993 	msg_set_non_seq(hdr, bcast);
1994 	__skb_queue_tail(xmitq, skb);
1995 	return true;
1996 }
1997 
1998 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1999  *
2000  * Give a newly added peer node the sequence number where it should
2001  * start receiving and acking broadcast packets.
2002  */
2003 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2004 					struct sk_buff_head *xmitq)
2005 {
2006 	struct sk_buff_head list;
2007 
2008 	__skb_queue_head_init(&list);
2009 	if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2010 		return;
2011 	msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2012 	tipc_link_xmit(l, &list, xmitq);
2013 }
2014 
2015 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2016  */
2017 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2018 {
2019 	int mtyp = msg_type(hdr);
2020 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2021 
2022 	if (link_is_up(l))
2023 		return;
2024 
2025 	if (msg_user(hdr) == BCAST_PROTOCOL) {
2026 		l->rcv_nxt = peers_snd_nxt;
2027 		l->state = LINK_ESTABLISHED;
2028 		return;
2029 	}
2030 
2031 	if (l->peer_caps & TIPC_BCAST_SYNCH)
2032 		return;
2033 
2034 	if (msg_peer_node_is_up(hdr))
2035 		return;
2036 
2037 	/* Compatibility: accept older, less safe initial synch data */
2038 	if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2039 		l->rcv_nxt = peers_snd_nxt;
2040 }
2041 
2042 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2043  */
2044 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2045 			  struct sk_buff_head *xmitq)
2046 {
2047 	struct tipc_link *snd_l = l->bc_sndlink;
2048 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2049 	u16 from = msg_bcast_ack(hdr) + 1;
2050 	u16 to = from + msg_bc_gap(hdr) - 1;
2051 	int rc = 0;
2052 
2053 	if (!link_is_up(l))
2054 		return rc;
2055 
2056 	if (!msg_peer_node_is_up(hdr))
2057 		return rc;
2058 
2059 	/* Open when peer ackowledges our bcast init msg (pkt #1) */
2060 	if (msg_ack(hdr))
2061 		l->bc_peer_is_up = true;
2062 
2063 	if (!l->bc_peer_is_up)
2064 		return rc;
2065 
2066 	l->stats.recv_nacks++;
2067 
2068 	/* Ignore if peers_snd_nxt goes beyond receive window */
2069 	if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2070 		return rc;
2071 
2072 	rc = tipc_link_bc_retrans(snd_l, l, from, to, xmitq);
2073 
2074 	l->snd_nxt = peers_snd_nxt;
2075 	if (link_bc_rcv_gap(l))
2076 		rc |= TIPC_LINK_SND_STATE;
2077 
2078 	/* Return now if sender supports nack via STATE messages */
2079 	if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2080 		return rc;
2081 
2082 	/* Otherwise, be backwards compatible */
2083 
2084 	if (!more(peers_snd_nxt, l->rcv_nxt)) {
2085 		l->nack_state = BC_NACK_SND_CONDITIONAL;
2086 		return 0;
2087 	}
2088 
2089 	/* Don't NACK if one was recently sent or peeked */
2090 	if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2091 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2092 		return 0;
2093 	}
2094 
2095 	/* Conditionally delay NACK sending until next synch rcv */
2096 	if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2097 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2098 		if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2099 			return 0;
2100 	}
2101 
2102 	/* Send NACK now but suppress next one */
2103 	tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2104 	l->nack_state = BC_NACK_SND_SUPPRESS;
2105 	return 0;
2106 }
2107 
2108 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
2109 			  struct sk_buff_head *xmitq)
2110 {
2111 	struct sk_buff *skb, *tmp;
2112 	struct tipc_link *snd_l = l->bc_sndlink;
2113 
2114 	if (!link_is_up(l) || !l->bc_peer_is_up)
2115 		return;
2116 
2117 	if (!more(acked, l->acked))
2118 		return;
2119 
2120 	trace_tipc_link_bc_ack(l, l->acked, acked, &snd_l->transmq);
2121 	/* Skip over packets peer has already acked */
2122 	skb_queue_walk(&snd_l->transmq, skb) {
2123 		if (more(buf_seqno(skb), l->acked))
2124 			break;
2125 	}
2126 
2127 	/* Update/release the packets peer is acking now */
2128 	skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
2129 		if (more(buf_seqno(skb), acked))
2130 			break;
2131 		if (!--TIPC_SKB_CB(skb)->ackers) {
2132 			__skb_unlink(skb, &snd_l->transmq);
2133 			kfree_skb(skb);
2134 		}
2135 	}
2136 	l->acked = acked;
2137 	tipc_link_advance_backlog(snd_l, xmitq);
2138 	if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
2139 		link_prepare_wakeup(snd_l);
2140 }
2141 
2142 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2143  * This function is here for backwards compatibility, since
2144  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2145  */
2146 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2147 			  struct sk_buff_head *xmitq)
2148 {
2149 	struct tipc_msg *hdr = buf_msg(skb);
2150 	u32 dnode = msg_destnode(hdr);
2151 	int mtyp = msg_type(hdr);
2152 	u16 acked = msg_bcast_ack(hdr);
2153 	u16 from = acked + 1;
2154 	u16 to = msg_bcgap_to(hdr);
2155 	u16 peers_snd_nxt = to + 1;
2156 	int rc = 0;
2157 
2158 	kfree_skb(skb);
2159 
2160 	if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2161 		return 0;
2162 
2163 	if (mtyp != STATE_MSG)
2164 		return 0;
2165 
2166 	if (dnode == tipc_own_addr(l->net)) {
2167 		tipc_link_bc_ack_rcv(l, acked, xmitq);
2168 		rc = tipc_link_bc_retrans(l->bc_sndlink, l, from, to, xmitq);
2169 		l->stats.recv_nacks++;
2170 		return rc;
2171 	}
2172 
2173 	/* Msg for other node => suppress own NACK at next sync if applicable */
2174 	if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2175 		l->nack_state = BC_NACK_SND_SUPPRESS;
2176 
2177 	return 0;
2178 }
2179 
2180 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
2181 {
2182 	int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2183 
2184 	l->window = win;
2185 	l->backlog[TIPC_LOW_IMPORTANCE].limit      = max_t(u16, 50, win);
2186 	l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = max_t(u16, 100, win * 2);
2187 	l->backlog[TIPC_HIGH_IMPORTANCE].limit     = max_t(u16, 150, win * 3);
2188 	l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
2189 	l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
2190 }
2191 
2192 /**
2193  * link_reset_stats - reset link statistics
2194  * @l: pointer to link
2195  */
2196 void tipc_link_reset_stats(struct tipc_link *l)
2197 {
2198 	memset(&l->stats, 0, sizeof(l->stats));
2199 }
2200 
2201 static void link_print(struct tipc_link *l, const char *str)
2202 {
2203 	struct sk_buff *hskb = skb_peek(&l->transmq);
2204 	u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2205 	u16 tail = l->snd_nxt - 1;
2206 
2207 	pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2208 	pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2209 		skb_queue_len(&l->transmq), head, tail,
2210 		skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2211 }
2212 
2213 /* Parse and validate nested (link) properties valid for media, bearer and link
2214  */
2215 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2216 {
2217 	int err;
2218 
2219 	err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2220 					  tipc_nl_prop_policy, NULL);
2221 	if (err)
2222 		return err;
2223 
2224 	if (props[TIPC_NLA_PROP_PRIO]) {
2225 		u32 prio;
2226 
2227 		prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2228 		if (prio > TIPC_MAX_LINK_PRI)
2229 			return -EINVAL;
2230 	}
2231 
2232 	if (props[TIPC_NLA_PROP_TOL]) {
2233 		u32 tol;
2234 
2235 		tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2236 		if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2237 			return -EINVAL;
2238 	}
2239 
2240 	if (props[TIPC_NLA_PROP_WIN]) {
2241 		u32 win;
2242 
2243 		win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2244 		if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
2245 			return -EINVAL;
2246 	}
2247 
2248 	return 0;
2249 }
2250 
2251 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2252 {
2253 	int i;
2254 	struct nlattr *stats;
2255 
2256 	struct nla_map {
2257 		u32 key;
2258 		u32 val;
2259 	};
2260 
2261 	struct nla_map map[] = {
2262 		{TIPC_NLA_STATS_RX_INFO, 0},
2263 		{TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2264 		{TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2265 		{TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2266 		{TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2267 		{TIPC_NLA_STATS_TX_INFO, 0},
2268 		{TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2269 		{TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2270 		{TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2271 		{TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2272 		{TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2273 			s->msg_length_counts : 1},
2274 		{TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2275 		{TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2276 		{TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2277 		{TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2278 		{TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2279 		{TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2280 		{TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2281 		{TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2282 		{TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2283 		{TIPC_NLA_STATS_RX_STATES, s->recv_states},
2284 		{TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2285 		{TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2286 		{TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2287 		{TIPC_NLA_STATS_TX_STATES, s->sent_states},
2288 		{TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2289 		{TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2290 		{TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2291 		{TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2292 		{TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2293 		{TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2294 		{TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2295 		{TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2296 			(s->accu_queue_sz / s->queue_sz_counts) : 0}
2297 	};
2298 
2299 	stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2300 	if (!stats)
2301 		return -EMSGSIZE;
2302 
2303 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2304 		if (nla_put_u32(skb, map[i].key, map[i].val))
2305 			goto msg_full;
2306 
2307 	nla_nest_end(skb, stats);
2308 
2309 	return 0;
2310 msg_full:
2311 	nla_nest_cancel(skb, stats);
2312 
2313 	return -EMSGSIZE;
2314 }
2315 
2316 /* Caller should hold appropriate locks to protect the link */
2317 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2318 		       struct tipc_link *link, int nlflags)
2319 {
2320 	u32 self = tipc_own_addr(net);
2321 	struct nlattr *attrs;
2322 	struct nlattr *prop;
2323 	void *hdr;
2324 	int err;
2325 
2326 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2327 			  nlflags, TIPC_NL_LINK_GET);
2328 	if (!hdr)
2329 		return -EMSGSIZE;
2330 
2331 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2332 	if (!attrs)
2333 		goto msg_full;
2334 
2335 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2336 		goto attr_msg_full;
2337 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2338 		goto attr_msg_full;
2339 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2340 		goto attr_msg_full;
2341 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2342 		goto attr_msg_full;
2343 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2344 		goto attr_msg_full;
2345 
2346 	if (tipc_link_is_up(link))
2347 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2348 			goto attr_msg_full;
2349 	if (link->active)
2350 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2351 			goto attr_msg_full;
2352 
2353 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2354 	if (!prop)
2355 		goto attr_msg_full;
2356 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2357 		goto prop_msg_full;
2358 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2359 		goto prop_msg_full;
2360 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2361 			link->window))
2362 		goto prop_msg_full;
2363 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2364 		goto prop_msg_full;
2365 	nla_nest_end(msg->skb, prop);
2366 
2367 	err = __tipc_nl_add_stats(msg->skb, &link->stats);
2368 	if (err)
2369 		goto attr_msg_full;
2370 
2371 	nla_nest_end(msg->skb, attrs);
2372 	genlmsg_end(msg->skb, hdr);
2373 
2374 	return 0;
2375 
2376 prop_msg_full:
2377 	nla_nest_cancel(msg->skb, prop);
2378 attr_msg_full:
2379 	nla_nest_cancel(msg->skb, attrs);
2380 msg_full:
2381 	genlmsg_cancel(msg->skb, hdr);
2382 
2383 	return -EMSGSIZE;
2384 }
2385 
2386 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2387 				      struct tipc_stats *stats)
2388 {
2389 	int i;
2390 	struct nlattr *nest;
2391 
2392 	struct nla_map {
2393 		__u32 key;
2394 		__u32 val;
2395 	};
2396 
2397 	struct nla_map map[] = {
2398 		{TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2399 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2400 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2401 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2402 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2403 		{TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2404 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2405 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2406 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2407 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2408 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2409 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2410 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2411 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2412 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2413 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2414 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2415 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2416 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2417 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2418 	};
2419 
2420 	nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2421 	if (!nest)
2422 		return -EMSGSIZE;
2423 
2424 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2425 		if (nla_put_u32(skb, map[i].key, map[i].val))
2426 			goto msg_full;
2427 
2428 	nla_nest_end(skb, nest);
2429 
2430 	return 0;
2431 msg_full:
2432 	nla_nest_cancel(skb, nest);
2433 
2434 	return -EMSGSIZE;
2435 }
2436 
2437 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2438 {
2439 	int err;
2440 	void *hdr;
2441 	struct nlattr *attrs;
2442 	struct nlattr *prop;
2443 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2444 	u32 bc_mode = tipc_bcast_get_broadcast_mode(net);
2445 	u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2446 	struct tipc_link *bcl = tn->bcl;
2447 
2448 	if (!bcl)
2449 		return 0;
2450 
2451 	tipc_bcast_lock(net);
2452 
2453 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2454 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
2455 	if (!hdr) {
2456 		tipc_bcast_unlock(net);
2457 		return -EMSGSIZE;
2458 	}
2459 
2460 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2461 	if (!attrs)
2462 		goto msg_full;
2463 
2464 	/* The broadcast link is always up */
2465 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2466 		goto attr_msg_full;
2467 
2468 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2469 		goto attr_msg_full;
2470 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2471 		goto attr_msg_full;
2472 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2473 		goto attr_msg_full;
2474 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2475 		goto attr_msg_full;
2476 
2477 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2478 	if (!prop)
2479 		goto attr_msg_full;
2480 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2481 		goto prop_msg_full;
2482 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2483 		goto prop_msg_full;
2484 	if (bc_mode & BCLINK_MODE_SEL)
2485 		if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2486 				bc_ratio))
2487 			goto prop_msg_full;
2488 	nla_nest_end(msg->skb, prop);
2489 
2490 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2491 	if (err)
2492 		goto attr_msg_full;
2493 
2494 	tipc_bcast_unlock(net);
2495 	nla_nest_end(msg->skb, attrs);
2496 	genlmsg_end(msg->skb, hdr);
2497 
2498 	return 0;
2499 
2500 prop_msg_full:
2501 	nla_nest_cancel(msg->skb, prop);
2502 attr_msg_full:
2503 	nla_nest_cancel(msg->skb, attrs);
2504 msg_full:
2505 	tipc_bcast_unlock(net);
2506 	genlmsg_cancel(msg->skb, hdr);
2507 
2508 	return -EMSGSIZE;
2509 }
2510 
2511 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2512 			     struct sk_buff_head *xmitq)
2513 {
2514 	l->tolerance = tol;
2515 	if (l->bc_rcvlink)
2516 		l->bc_rcvlink->tolerance = tol;
2517 	if (link_is_up(l))
2518 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2519 }
2520 
2521 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2522 			struct sk_buff_head *xmitq)
2523 {
2524 	l->priority = prio;
2525 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2526 }
2527 
2528 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2529 {
2530 	l->abort_limit = limit;
2531 }
2532 
2533 char *tipc_link_name_ext(struct tipc_link *l, char *buf)
2534 {
2535 	if (!l)
2536 		scnprintf(buf, TIPC_MAX_LINK_NAME, "null");
2537 	else if (link_is_bc_sndlink(l))
2538 		scnprintf(buf, TIPC_MAX_LINK_NAME, "broadcast-sender");
2539 	else if (link_is_bc_rcvlink(l))
2540 		scnprintf(buf, TIPC_MAX_LINK_NAME,
2541 			  "broadcast-receiver, peer %x", l->addr);
2542 	else
2543 		memcpy(buf, l->name, TIPC_MAX_LINK_NAME);
2544 
2545 	return buf;
2546 }
2547 
2548 /**
2549  * tipc_link_dump - dump TIPC link data
2550  * @l: tipc link to be dumped
2551  * @dqueues: bitmask to decide if any link queue to be dumped?
2552  *           - TIPC_DUMP_NONE: don't dump link queues
2553  *           - TIPC_DUMP_TRANSMQ: dump link transmq queue
2554  *           - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2555  *           - TIPC_DUMP_DEFERDQ: dump link deferd queue
2556  *           - TIPC_DUMP_INPUTQ: dump link input queue
2557  *           - TIPC_DUMP_WAKEUP: dump link wakeup queue
2558  *           - TIPC_DUMP_ALL: dump all the link queues above
2559  * @buf: returned buffer of dump data in format
2560  */
2561 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2562 {
2563 	int i = 0;
2564 	size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2565 	struct sk_buff_head *list;
2566 	struct sk_buff *hskb, *tskb;
2567 	u32 len;
2568 
2569 	if (!l) {
2570 		i += scnprintf(buf, sz, "link data: (null)\n");
2571 		return i;
2572 	}
2573 
2574 	i += scnprintf(buf, sz, "link data: %x", l->addr);
2575 	i += scnprintf(buf + i, sz - i, " %x", l->state);
2576 	i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2577 	i += scnprintf(buf + i, sz - i, " %u", l->session);
2578 	i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2579 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2580 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2581 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2582 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2583 	i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2584 	i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2585 	i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2586 	i += scnprintf(buf + i, sz - i, " %u", l->prev_from);
2587 	i += scnprintf(buf + i, sz - i, " %u", l->stale_cnt);
2588 	i += scnprintf(buf + i, sz - i, " %u", l->acked);
2589 
2590 	list = &l->transmq;
2591 	len = skb_queue_len(list);
2592 	hskb = skb_peek(list);
2593 	tskb = skb_peek_tail(list);
2594 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2595 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2596 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2597 
2598 	list = &l->deferdq;
2599 	len = skb_queue_len(list);
2600 	hskb = skb_peek(list);
2601 	tskb = skb_peek_tail(list);
2602 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2603 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2604 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2605 
2606 	list = &l->backlogq;
2607 	len = skb_queue_len(list);
2608 	hskb = skb_peek(list);
2609 	tskb = skb_peek_tail(list);
2610 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2611 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2612 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2613 
2614 	list = l->inputq;
2615 	len = skb_queue_len(list);
2616 	hskb = skb_peek(list);
2617 	tskb = skb_peek_tail(list);
2618 	i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2619 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2620 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2621 
2622 	if (dqueues & TIPC_DUMP_TRANSMQ) {
2623 		i += scnprintf(buf + i, sz - i, "transmq: ");
2624 		i += tipc_list_dump(&l->transmq, false, buf + i);
2625 	}
2626 	if (dqueues & TIPC_DUMP_BACKLOGQ) {
2627 		i += scnprintf(buf + i, sz - i,
2628 			       "backlogq: <%u %u %u %u %u>, ",
2629 			       l->backlog[TIPC_LOW_IMPORTANCE].len,
2630 			       l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2631 			       l->backlog[TIPC_HIGH_IMPORTANCE].len,
2632 			       l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2633 			       l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2634 		i += tipc_list_dump(&l->backlogq, false, buf + i);
2635 	}
2636 	if (dqueues & TIPC_DUMP_DEFERDQ) {
2637 		i += scnprintf(buf + i, sz - i, "deferdq: ");
2638 		i += tipc_list_dump(&l->deferdq, false, buf + i);
2639 	}
2640 	if (dqueues & TIPC_DUMP_INPUTQ) {
2641 		i += scnprintf(buf + i, sz - i, "inputq: ");
2642 		i += tipc_list_dump(l->inputq, false, buf + i);
2643 	}
2644 	if (dqueues & TIPC_DUMP_WAKEUP) {
2645 		i += scnprintf(buf + i, sz - i, "wakeup: ");
2646 		i += tipc_list_dump(&l->wakeupq, false, buf + i);
2647 	}
2648 
2649 	return i;
2650 }
2651