xref: /linux/net/tipc/link.c (revision 2c63221cd9e5c0dad0424029aeb1c40faada8330)
1 /*
2  * net/tipc/link.c: TIPC link code
3  *
4  * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5  * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #include "core.h"
38 #include "subscr.h"
39 #include "link.h"
40 #include "bcast.h"
41 #include "socket.h"
42 #include "name_distr.h"
43 #include "discover.h"
44 #include "netlink.h"
45 #include "monitor.h"
46 #include "trace.h"
47 
48 #include <linux/pkt_sched.h>
49 
50 struct tipc_stats {
51 	u32 sent_pkts;
52 	u32 recv_pkts;
53 	u32 sent_states;
54 	u32 recv_states;
55 	u32 sent_probes;
56 	u32 recv_probes;
57 	u32 sent_nacks;
58 	u32 recv_nacks;
59 	u32 sent_acks;
60 	u32 sent_bundled;
61 	u32 sent_bundles;
62 	u32 recv_bundled;
63 	u32 recv_bundles;
64 	u32 retransmitted;
65 	u32 sent_fragmented;
66 	u32 sent_fragments;
67 	u32 recv_fragmented;
68 	u32 recv_fragments;
69 	u32 link_congs;		/* # port sends blocked by congestion */
70 	u32 deferred_recv;
71 	u32 duplicates;
72 	u32 max_queue_sz;	/* send queue size high water mark */
73 	u32 accu_queue_sz;	/* used for send queue size profiling */
74 	u32 queue_sz_counts;	/* used for send queue size profiling */
75 	u32 msg_length_counts;	/* used for message length profiling */
76 	u32 msg_lengths_total;	/* used for message length profiling */
77 	u32 msg_length_profile[7]; /* used for msg. length profiling */
78 };
79 
80 /**
81  * struct tipc_link - TIPC link data structure
82  * @addr: network address of link's peer node
83  * @name: link name character string
84  * @media_addr: media address to use when sending messages over link
85  * @timer: link timer
86  * @net: pointer to namespace struct
87  * @refcnt: reference counter for permanent references (owner node & timer)
88  * @peer_session: link session # being used by peer end of link
89  * @peer_bearer_id: bearer id used by link's peer endpoint
90  * @bearer_id: local bearer id used by link
91  * @tolerance: minimum link continuity loss needed to reset link [in ms]
92  * @abort_limit: # of unacknowledged continuity probes needed to reset link
93  * @state: current state of link FSM
94  * @peer_caps: bitmap describing capabilities of peer node
95  * @silent_intv_cnt: # of timer intervals without any reception from peer
96  * @proto_msg: template for control messages generated by link
97  * @pmsg: convenience pointer to "proto_msg" field
98  * @priority: current link priority
99  * @net_plane: current link network plane ('A' through 'H')
100  * @mon_state: cookie with information needed by link monitor
101  * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
102  * @exp_msg_count: # of tunnelled messages expected during link changeover
103  * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
104  * @mtu: current maximum packet size for this link
105  * @advertised_mtu: advertised own mtu when link is being established
106  * @transmitq: queue for sent, non-acked messages
107  * @backlogq: queue for messages waiting to be sent
108  * @snt_nxt: next sequence number to use for outbound messages
109  * @ackers: # of peers that needs to ack each packet before it can be released
110  * @acked: # last packet acked by a certain peer. Used for broadcast.
111  * @rcv_nxt: next sequence number to expect for inbound messages
112  * @deferred_queue: deferred queue saved OOS b'cast message received from node
113  * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
114  * @inputq: buffer queue for messages to be delivered upwards
115  * @namedq: buffer queue for name table messages to be delivered upwards
116  * @next_out: ptr to first unsent outbound message in queue
117  * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
118  * @long_msg_seq_no: next identifier to use for outbound fragmented messages
119  * @reasm_buf: head of partially reassembled inbound message fragments
120  * @bc_rcvr: marks that this is a broadcast receiver link
121  * @stats: collects statistics regarding link activity
122  */
123 struct tipc_link {
124 	u32 addr;
125 	char name[TIPC_MAX_LINK_NAME];
126 	struct net *net;
127 
128 	/* Management and link supervision data */
129 	u16 peer_session;
130 	u16 session;
131 	u16 snd_nxt_state;
132 	u16 rcv_nxt_state;
133 	u32 peer_bearer_id;
134 	u32 bearer_id;
135 	u32 tolerance;
136 	u32 abort_limit;
137 	u32 state;
138 	u16 peer_caps;
139 	bool in_session;
140 	bool active;
141 	u32 silent_intv_cnt;
142 	char if_name[TIPC_MAX_IF_NAME];
143 	u32 priority;
144 	char net_plane;
145 	struct tipc_mon_state mon_state;
146 	u16 rst_cnt;
147 
148 	/* Failover/synch */
149 	u16 drop_point;
150 	struct sk_buff *failover_reasm_skb;
151 	struct sk_buff_head failover_deferdq;
152 
153 	/* Max packet negotiation */
154 	u16 mtu;
155 	u16 advertised_mtu;
156 
157 	/* Sending */
158 	struct sk_buff_head transmq;
159 	struct sk_buff_head backlogq;
160 	struct {
161 		u16 len;
162 		u16 limit;
163 		struct sk_buff *target_bskb;
164 	} backlog[5];
165 	u16 snd_nxt;
166 	u16 window;
167 
168 	/* Reception */
169 	u16 rcv_nxt;
170 	u32 rcv_unacked;
171 	struct sk_buff_head deferdq;
172 	struct sk_buff_head *inputq;
173 	struct sk_buff_head *namedq;
174 
175 	/* Congestion handling */
176 	struct sk_buff_head wakeupq;
177 
178 	/* Fragmentation/reassembly */
179 	struct sk_buff *reasm_buf;
180 	struct sk_buff *reasm_tnlmsg;
181 
182 	/* Broadcast */
183 	u16 ackers;
184 	u16 acked;
185 	struct tipc_link *bc_rcvlink;
186 	struct tipc_link *bc_sndlink;
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_LIM  (jiffies + msecs_to_jiffies(10))
209 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
210 
211 /*
212  * Interval between NACKs when packets arrive out of order
213  */
214 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
215 
216 /* Link FSM states:
217  */
218 enum {
219 	LINK_ESTABLISHED     = 0xe,
220 	LINK_ESTABLISHING    = 0xe  << 4,
221 	LINK_RESET           = 0x1  << 8,
222 	LINK_RESETTING       = 0x2  << 12,
223 	LINK_PEER_RESET      = 0xd  << 16,
224 	LINK_FAILINGOVER     = 0xf  << 20,
225 	LINK_SYNCHING        = 0xc  << 24
226 };
227 
228 /* Link FSM state checking routines
229  */
230 static int link_is_up(struct tipc_link *l)
231 {
232 	return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
233 }
234 
235 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
236 			       struct sk_buff_head *xmitq);
237 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
238 				      bool probe_reply, u16 rcvgap,
239 				      int tolerance, int priority,
240 				      struct sk_buff_head *xmitq);
241 static void link_print(struct tipc_link *l, const char *str);
242 static int tipc_link_build_nack_msg(struct tipc_link *l,
243 				    struct sk_buff_head *xmitq);
244 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
245 					struct sk_buff_head *xmitq);
246 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
247 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data);
248 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
249 				     struct tipc_gap_ack_blks *ga,
250 				     struct sk_buff_head *xmitq);
251 
252 /*
253  *  Simple non-static link routines (i.e. referenced outside this file)
254  */
255 bool tipc_link_is_up(struct tipc_link *l)
256 {
257 	return link_is_up(l);
258 }
259 
260 bool tipc_link_peer_is_down(struct tipc_link *l)
261 {
262 	return l->state == LINK_PEER_RESET;
263 }
264 
265 bool tipc_link_is_reset(struct tipc_link *l)
266 {
267 	return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
268 }
269 
270 bool tipc_link_is_establishing(struct tipc_link *l)
271 {
272 	return l->state == LINK_ESTABLISHING;
273 }
274 
275 bool tipc_link_is_synching(struct tipc_link *l)
276 {
277 	return l->state == LINK_SYNCHING;
278 }
279 
280 bool tipc_link_is_failingover(struct tipc_link *l)
281 {
282 	return l->state == LINK_FAILINGOVER;
283 }
284 
285 bool tipc_link_is_blocked(struct tipc_link *l)
286 {
287 	return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
288 }
289 
290 static bool link_is_bc_sndlink(struct tipc_link *l)
291 {
292 	return !l->bc_sndlink;
293 }
294 
295 static bool link_is_bc_rcvlink(struct tipc_link *l)
296 {
297 	return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
298 }
299 
300 void tipc_link_set_active(struct tipc_link *l, bool active)
301 {
302 	l->active = active;
303 }
304 
305 u32 tipc_link_id(struct tipc_link *l)
306 {
307 	return l->peer_bearer_id << 16 | l->bearer_id;
308 }
309 
310 int tipc_link_window(struct tipc_link *l)
311 {
312 	return l->window;
313 }
314 
315 int tipc_link_prio(struct tipc_link *l)
316 {
317 	return l->priority;
318 }
319 
320 unsigned long tipc_link_tolerance(struct tipc_link *l)
321 {
322 	return l->tolerance;
323 }
324 
325 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
326 {
327 	return l->inputq;
328 }
329 
330 char tipc_link_plane(struct tipc_link *l)
331 {
332 	return l->net_plane;
333 }
334 
335 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
336 {
337 	l->peer_caps = capabilities;
338 }
339 
340 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
341 			   struct tipc_link *uc_l,
342 			   struct sk_buff_head *xmitq)
343 {
344 	struct tipc_link *rcv_l = uc_l->bc_rcvlink;
345 
346 	snd_l->ackers++;
347 	rcv_l->acked = snd_l->snd_nxt - 1;
348 	snd_l->state = LINK_ESTABLISHED;
349 	tipc_link_build_bc_init_msg(uc_l, xmitq);
350 }
351 
352 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
353 			      struct tipc_link *rcv_l,
354 			      struct sk_buff_head *xmitq)
355 {
356 	u16 ack = snd_l->snd_nxt - 1;
357 
358 	snd_l->ackers--;
359 	rcv_l->bc_peer_is_up = true;
360 	rcv_l->state = LINK_ESTABLISHED;
361 	tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
362 	trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
363 	tipc_link_reset(rcv_l);
364 	rcv_l->state = LINK_RESET;
365 	if (!snd_l->ackers) {
366 		trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
367 		tipc_link_reset(snd_l);
368 		snd_l->state = LINK_RESET;
369 		__skb_queue_purge(xmitq);
370 	}
371 }
372 
373 int tipc_link_bc_peers(struct tipc_link *l)
374 {
375 	return l->ackers;
376 }
377 
378 static u16 link_bc_rcv_gap(struct tipc_link *l)
379 {
380 	struct sk_buff *skb = skb_peek(&l->deferdq);
381 	u16 gap = 0;
382 
383 	if (more(l->snd_nxt, l->rcv_nxt))
384 		gap = l->snd_nxt - l->rcv_nxt;
385 	if (skb)
386 		gap = buf_seqno(skb) - l->rcv_nxt;
387 	return gap;
388 }
389 
390 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
391 {
392 	l->mtu = mtu;
393 }
394 
395 int tipc_link_mtu(struct tipc_link *l)
396 {
397 	return l->mtu;
398 }
399 
400 u16 tipc_link_rcv_nxt(struct tipc_link *l)
401 {
402 	return l->rcv_nxt;
403 }
404 
405 u16 tipc_link_acked(struct tipc_link *l)
406 {
407 	return l->acked;
408 }
409 
410 char *tipc_link_name(struct tipc_link *l)
411 {
412 	return l->name;
413 }
414 
415 u32 tipc_link_state(struct tipc_link *l)
416 {
417 	return l->state;
418 }
419 
420 /**
421  * tipc_link_create - create a new link
422  * @n: pointer to associated node
423  * @if_name: associated interface name
424  * @bearer_id: id (index) of associated bearer
425  * @tolerance: link tolerance to be used by link
426  * @net_plane: network plane (A,B,c..) this link belongs to
427  * @mtu: mtu to be advertised by link
428  * @priority: priority to be used by link
429  * @window: send window to be used by link
430  * @session: session to be used by link
431  * @ownnode: identity of own node
432  * @peer: node id of peer node
433  * @peer_caps: bitmap describing peer node capabilities
434  * @bc_sndlink: the namespace global link used for broadcast sending
435  * @bc_rcvlink: the peer specific link used for broadcast reception
436  * @inputq: queue to put messages ready for delivery
437  * @namedq: queue to put binding table update messages ready for delivery
438  * @link: return value, pointer to put the created link
439  *
440  * Returns true if link was created, otherwise false
441  */
442 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
443 		      int tolerance, char net_plane, u32 mtu, int priority,
444 		      int window, u32 session, u32 self,
445 		      u32 peer, u8 *peer_id, u16 peer_caps,
446 		      struct tipc_link *bc_sndlink,
447 		      struct tipc_link *bc_rcvlink,
448 		      struct sk_buff_head *inputq,
449 		      struct sk_buff_head *namedq,
450 		      struct tipc_link **link)
451 {
452 	char peer_str[NODE_ID_STR_LEN] = {0,};
453 	char self_str[NODE_ID_STR_LEN] = {0,};
454 	struct tipc_link *l;
455 
456 	l = kzalloc(sizeof(*l), GFP_ATOMIC);
457 	if (!l)
458 		return false;
459 	*link = l;
460 	l->session = session;
461 
462 	/* Set link name for unicast links only */
463 	if (peer_id) {
464 		tipc_nodeid2string(self_str, tipc_own_id(net));
465 		if (strlen(self_str) > 16)
466 			sprintf(self_str, "%x", self);
467 		tipc_nodeid2string(peer_str, peer_id);
468 		if (strlen(peer_str) > 16)
469 			sprintf(peer_str, "%x", peer);
470 	}
471 	/* Peer i/f name will be completed by reset/activate message */
472 	snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
473 		 self_str, if_name, peer_str);
474 
475 	strcpy(l->if_name, if_name);
476 	l->addr = peer;
477 	l->peer_caps = peer_caps;
478 	l->net = net;
479 	l->in_session = false;
480 	l->bearer_id = bearer_id;
481 	l->tolerance = tolerance;
482 	if (bc_rcvlink)
483 		bc_rcvlink->tolerance = tolerance;
484 	l->net_plane = net_plane;
485 	l->advertised_mtu = mtu;
486 	l->mtu = mtu;
487 	l->priority = priority;
488 	tipc_link_set_queue_limits(l, window);
489 	l->ackers = 1;
490 	l->bc_sndlink = bc_sndlink;
491 	l->bc_rcvlink = bc_rcvlink;
492 	l->inputq = inputq;
493 	l->namedq = namedq;
494 	l->state = LINK_RESETTING;
495 	__skb_queue_head_init(&l->transmq);
496 	__skb_queue_head_init(&l->backlogq);
497 	__skb_queue_head_init(&l->deferdq);
498 	__skb_queue_head_init(&l->failover_deferdq);
499 	skb_queue_head_init(&l->wakeupq);
500 	skb_queue_head_init(l->inputq);
501 	return true;
502 }
503 
504 /**
505  * tipc_link_bc_create - create new link to be used for broadcast
506  * @n: pointer to associated node
507  * @mtu: mtu to be used initially if no peers
508  * @window: send window to be used
509  * @inputq: queue to put messages ready for delivery
510  * @namedq: queue to put binding table update messages ready for delivery
511  * @link: return value, pointer to put the created link
512  *
513  * Returns true if link was created, otherwise false
514  */
515 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
516 			 int mtu, int window, u16 peer_caps,
517 			 struct sk_buff_head *inputq,
518 			 struct sk_buff_head *namedq,
519 			 struct tipc_link *bc_sndlink,
520 			 struct tipc_link **link)
521 {
522 	struct tipc_link *l;
523 
524 	if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
525 			      0, ownnode, peer, NULL, peer_caps, bc_sndlink,
526 			      NULL, inputq, namedq, link))
527 		return false;
528 
529 	l = *link;
530 	strcpy(l->name, tipc_bclink_name);
531 	trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
532 	tipc_link_reset(l);
533 	l->state = LINK_RESET;
534 	l->ackers = 0;
535 	l->bc_rcvlink = l;
536 
537 	/* Broadcast send link is always up */
538 	if (link_is_bc_sndlink(l))
539 		l->state = LINK_ESTABLISHED;
540 
541 	/* Disable replicast if even a single peer doesn't support it */
542 	if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
543 		tipc_bcast_disable_rcast(net);
544 
545 	return true;
546 }
547 
548 /**
549  * tipc_link_fsm_evt - link finite state machine
550  * @l: pointer to link
551  * @evt: state machine event to be processed
552  */
553 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
554 {
555 	int rc = 0;
556 	int old_state = l->state;
557 
558 	switch (l->state) {
559 	case LINK_RESETTING:
560 		switch (evt) {
561 		case LINK_PEER_RESET_EVT:
562 			l->state = LINK_PEER_RESET;
563 			break;
564 		case LINK_RESET_EVT:
565 			l->state = LINK_RESET;
566 			break;
567 		case LINK_FAILURE_EVT:
568 		case LINK_FAILOVER_BEGIN_EVT:
569 		case LINK_ESTABLISH_EVT:
570 		case LINK_FAILOVER_END_EVT:
571 		case LINK_SYNCH_BEGIN_EVT:
572 		case LINK_SYNCH_END_EVT:
573 		default:
574 			goto illegal_evt;
575 		}
576 		break;
577 	case LINK_RESET:
578 		switch (evt) {
579 		case LINK_PEER_RESET_EVT:
580 			l->state = LINK_ESTABLISHING;
581 			break;
582 		case LINK_FAILOVER_BEGIN_EVT:
583 			l->state = LINK_FAILINGOVER;
584 		case LINK_FAILURE_EVT:
585 		case LINK_RESET_EVT:
586 		case LINK_ESTABLISH_EVT:
587 		case LINK_FAILOVER_END_EVT:
588 			break;
589 		case LINK_SYNCH_BEGIN_EVT:
590 		case LINK_SYNCH_END_EVT:
591 		default:
592 			goto illegal_evt;
593 		}
594 		break;
595 	case LINK_PEER_RESET:
596 		switch (evt) {
597 		case LINK_RESET_EVT:
598 			l->state = LINK_ESTABLISHING;
599 			break;
600 		case LINK_PEER_RESET_EVT:
601 		case LINK_ESTABLISH_EVT:
602 		case LINK_FAILURE_EVT:
603 			break;
604 		case LINK_SYNCH_BEGIN_EVT:
605 		case LINK_SYNCH_END_EVT:
606 		case LINK_FAILOVER_BEGIN_EVT:
607 		case LINK_FAILOVER_END_EVT:
608 		default:
609 			goto illegal_evt;
610 		}
611 		break;
612 	case LINK_FAILINGOVER:
613 		switch (evt) {
614 		case LINK_FAILOVER_END_EVT:
615 			l->state = LINK_RESET;
616 			break;
617 		case LINK_PEER_RESET_EVT:
618 		case LINK_RESET_EVT:
619 		case LINK_ESTABLISH_EVT:
620 		case LINK_FAILURE_EVT:
621 			break;
622 		case LINK_FAILOVER_BEGIN_EVT:
623 		case LINK_SYNCH_BEGIN_EVT:
624 		case LINK_SYNCH_END_EVT:
625 		default:
626 			goto illegal_evt;
627 		}
628 		break;
629 	case LINK_ESTABLISHING:
630 		switch (evt) {
631 		case LINK_ESTABLISH_EVT:
632 			l->state = LINK_ESTABLISHED;
633 			break;
634 		case LINK_FAILOVER_BEGIN_EVT:
635 			l->state = LINK_FAILINGOVER;
636 			break;
637 		case LINK_RESET_EVT:
638 			l->state = LINK_RESET;
639 			break;
640 		case LINK_FAILURE_EVT:
641 		case LINK_PEER_RESET_EVT:
642 		case LINK_SYNCH_BEGIN_EVT:
643 		case LINK_FAILOVER_END_EVT:
644 			break;
645 		case LINK_SYNCH_END_EVT:
646 		default:
647 			goto illegal_evt;
648 		}
649 		break;
650 	case LINK_ESTABLISHED:
651 		switch (evt) {
652 		case LINK_PEER_RESET_EVT:
653 			l->state = LINK_PEER_RESET;
654 			rc |= TIPC_LINK_DOWN_EVT;
655 			break;
656 		case LINK_FAILURE_EVT:
657 			l->state = LINK_RESETTING;
658 			rc |= TIPC_LINK_DOWN_EVT;
659 			break;
660 		case LINK_RESET_EVT:
661 			l->state = LINK_RESET;
662 			break;
663 		case LINK_ESTABLISH_EVT:
664 		case LINK_SYNCH_END_EVT:
665 			break;
666 		case LINK_SYNCH_BEGIN_EVT:
667 			l->state = LINK_SYNCHING;
668 			break;
669 		case LINK_FAILOVER_BEGIN_EVT:
670 		case LINK_FAILOVER_END_EVT:
671 		default:
672 			goto illegal_evt;
673 		}
674 		break;
675 	case LINK_SYNCHING:
676 		switch (evt) {
677 		case LINK_PEER_RESET_EVT:
678 			l->state = LINK_PEER_RESET;
679 			rc |= TIPC_LINK_DOWN_EVT;
680 			break;
681 		case LINK_FAILURE_EVT:
682 			l->state = LINK_RESETTING;
683 			rc |= TIPC_LINK_DOWN_EVT;
684 			break;
685 		case LINK_RESET_EVT:
686 			l->state = LINK_RESET;
687 			break;
688 		case LINK_ESTABLISH_EVT:
689 		case LINK_SYNCH_BEGIN_EVT:
690 			break;
691 		case LINK_SYNCH_END_EVT:
692 			l->state = LINK_ESTABLISHED;
693 			break;
694 		case LINK_FAILOVER_BEGIN_EVT:
695 		case LINK_FAILOVER_END_EVT:
696 		default:
697 			goto illegal_evt;
698 		}
699 		break;
700 	default:
701 		pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
702 	}
703 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
704 	return rc;
705 illegal_evt:
706 	pr_err("Illegal FSM event %x in state %x on link %s\n",
707 	       evt, l->state, l->name);
708 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
709 	return rc;
710 }
711 
712 /* link_profile_stats - update statistical profiling of traffic
713  */
714 static void link_profile_stats(struct tipc_link *l)
715 {
716 	struct sk_buff *skb;
717 	struct tipc_msg *msg;
718 	int length;
719 
720 	/* Update counters used in statistical profiling of send traffic */
721 	l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
722 	l->stats.queue_sz_counts++;
723 
724 	skb = skb_peek(&l->transmq);
725 	if (!skb)
726 		return;
727 	msg = buf_msg(skb);
728 	length = msg_size(msg);
729 
730 	if (msg_user(msg) == MSG_FRAGMENTER) {
731 		if (msg_type(msg) != FIRST_FRAGMENT)
732 			return;
733 		length = msg_size(msg_inner_hdr(msg));
734 	}
735 	l->stats.msg_lengths_total += length;
736 	l->stats.msg_length_counts++;
737 	if (length <= 64)
738 		l->stats.msg_length_profile[0]++;
739 	else if (length <= 256)
740 		l->stats.msg_length_profile[1]++;
741 	else if (length <= 1024)
742 		l->stats.msg_length_profile[2]++;
743 	else if (length <= 4096)
744 		l->stats.msg_length_profile[3]++;
745 	else if (length <= 16384)
746 		l->stats.msg_length_profile[4]++;
747 	else if (length <= 32768)
748 		l->stats.msg_length_profile[5]++;
749 	else
750 		l->stats.msg_length_profile[6]++;
751 }
752 
753 /**
754  * tipc_link_too_silent - check if link is "too silent"
755  * @l: tipc link to be checked
756  *
757  * Returns true if the link 'silent_intv_cnt' is about to reach the
758  * 'abort_limit' value, otherwise false
759  */
760 bool tipc_link_too_silent(struct tipc_link *l)
761 {
762 	return (l->silent_intv_cnt + 2 > l->abort_limit);
763 }
764 
765 /* tipc_link_timeout - perform periodic task as instructed from node timeout
766  */
767 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
768 {
769 	int mtyp = 0;
770 	int rc = 0;
771 	bool state = false;
772 	bool probe = false;
773 	bool setup = false;
774 	u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
775 	u16 bc_acked = l->bc_rcvlink->acked;
776 	struct tipc_mon_state *mstate = &l->mon_state;
777 
778 	trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
779 	trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
780 	switch (l->state) {
781 	case LINK_ESTABLISHED:
782 	case LINK_SYNCHING:
783 		mtyp = STATE_MSG;
784 		link_profile_stats(l);
785 		tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
786 		if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
787 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
788 		state = bc_acked != bc_snt;
789 		state |= l->bc_rcvlink->rcv_unacked;
790 		state |= l->rcv_unacked;
791 		state |= !skb_queue_empty(&l->transmq);
792 		state |= !skb_queue_empty(&l->deferdq);
793 		probe = mstate->probing;
794 		probe |= l->silent_intv_cnt;
795 		if (probe || mstate->monitoring)
796 			l->silent_intv_cnt++;
797 		break;
798 	case LINK_RESET:
799 		setup = l->rst_cnt++ <= 4;
800 		setup |= !(l->rst_cnt % 16);
801 		mtyp = RESET_MSG;
802 		break;
803 	case LINK_ESTABLISHING:
804 		setup = true;
805 		mtyp = ACTIVATE_MSG;
806 		break;
807 	case LINK_PEER_RESET:
808 	case LINK_RESETTING:
809 	case LINK_FAILINGOVER:
810 		break;
811 	default:
812 		break;
813 	}
814 
815 	if (state || probe || setup)
816 		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
817 
818 	return rc;
819 }
820 
821 /**
822  * link_schedule_user - schedule a message sender for wakeup after congestion
823  * @l: congested link
824  * @hdr: header of message that is being sent
825  * Create pseudo msg to send back to user when congestion abates
826  */
827 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
828 {
829 	u32 dnode = tipc_own_addr(l->net);
830 	u32 dport = msg_origport(hdr);
831 	struct sk_buff *skb;
832 
833 	/* Create and schedule wakeup pseudo message */
834 	skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
835 			      dnode, l->addr, dport, 0, 0);
836 	if (!skb)
837 		return -ENOBUFS;
838 	msg_set_dest_droppable(buf_msg(skb), true);
839 	TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
840 	skb_queue_tail(&l->wakeupq, skb);
841 	l->stats.link_congs++;
842 	trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
843 	return -ELINKCONG;
844 }
845 
846 /**
847  * link_prepare_wakeup - prepare users for wakeup after congestion
848  * @l: congested link
849  * Wake up a number of waiting users, as permitted by available space
850  * in the send queue
851  */
852 static void link_prepare_wakeup(struct tipc_link *l)
853 {
854 	struct sk_buff_head *wakeupq = &l->wakeupq;
855 	struct sk_buff_head *inputq = l->inputq;
856 	struct sk_buff *skb, *tmp;
857 	struct sk_buff_head tmpq;
858 	int avail[5] = {0,};
859 	int imp = 0;
860 
861 	__skb_queue_head_init(&tmpq);
862 
863 	for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
864 		avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
865 
866 	skb_queue_walk_safe(wakeupq, skb, tmp) {
867 		imp = TIPC_SKB_CB(skb)->chain_imp;
868 		if (avail[imp] <= 0)
869 			continue;
870 		avail[imp]--;
871 		__skb_unlink(skb, wakeupq);
872 		__skb_queue_tail(&tmpq, skb);
873 	}
874 
875 	spin_lock_bh(&inputq->lock);
876 	skb_queue_splice_tail(&tmpq, inputq);
877 	spin_unlock_bh(&inputq->lock);
878 
879 }
880 
881 void tipc_link_reset(struct tipc_link *l)
882 {
883 	struct sk_buff_head list;
884 	u32 imp;
885 
886 	__skb_queue_head_init(&list);
887 
888 	l->in_session = false;
889 	/* Force re-synch of peer session number before establishing */
890 	l->peer_session--;
891 	l->session++;
892 	l->mtu = l->advertised_mtu;
893 
894 	spin_lock_bh(&l->wakeupq.lock);
895 	skb_queue_splice_init(&l->wakeupq, &list);
896 	spin_unlock_bh(&l->wakeupq.lock);
897 
898 	spin_lock_bh(&l->inputq->lock);
899 	skb_queue_splice_init(&list, l->inputq);
900 	spin_unlock_bh(&l->inputq->lock);
901 
902 	__skb_queue_purge(&l->transmq);
903 	__skb_queue_purge(&l->deferdq);
904 	__skb_queue_purge(&l->backlogq);
905 	__skb_queue_purge(&l->failover_deferdq);
906 	for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
907 		l->backlog[imp].len = 0;
908 		l->backlog[imp].target_bskb = NULL;
909 	}
910 	kfree_skb(l->reasm_buf);
911 	kfree_skb(l->reasm_tnlmsg);
912 	kfree_skb(l->failover_reasm_skb);
913 	l->reasm_buf = NULL;
914 	l->reasm_tnlmsg = NULL;
915 	l->failover_reasm_skb = NULL;
916 	l->rcv_unacked = 0;
917 	l->snd_nxt = 1;
918 	l->rcv_nxt = 1;
919 	l->snd_nxt_state = 1;
920 	l->rcv_nxt_state = 1;
921 	l->acked = 0;
922 	l->silent_intv_cnt = 0;
923 	l->rst_cnt = 0;
924 	l->bc_peer_is_up = false;
925 	memset(&l->mon_state, 0, sizeof(l->mon_state));
926 	tipc_link_reset_stats(l);
927 }
928 
929 /**
930  * tipc_link_xmit(): enqueue buffer list according to queue situation
931  * @link: link to use
932  * @list: chain of buffers containing message
933  * @xmitq: returned list of packets to be sent by caller
934  *
935  * Consumes the buffer chain.
936  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
937  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
938  */
939 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
940 		   struct sk_buff_head *xmitq)
941 {
942 	struct tipc_msg *hdr = buf_msg(skb_peek(list));
943 	struct sk_buff_head *backlogq = &l->backlogq;
944 	struct sk_buff_head *transmq = &l->transmq;
945 	struct sk_buff *skb, *_skb;
946 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
947 	u16 ack = l->rcv_nxt - 1;
948 	u16 seqno = l->snd_nxt;
949 	int pkt_cnt = skb_queue_len(list);
950 	int imp = msg_importance(hdr);
951 	unsigned int maxwin = l->window;
952 	unsigned int mtu = l->mtu;
953 	bool new_bundle;
954 	int rc = 0;
955 
956 	if (unlikely(msg_size(hdr) > mtu)) {
957 		pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
958 			skb_queue_len(list), msg_user(hdr),
959 			msg_type(hdr), msg_size(hdr), mtu);
960 		__skb_queue_purge(list);
961 		return -EMSGSIZE;
962 	}
963 
964 	/* Allow oversubscription of one data msg per source at congestion */
965 	if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
966 		if (imp == TIPC_SYSTEM_IMPORTANCE) {
967 			pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
968 			return -ENOBUFS;
969 		}
970 		rc = link_schedule_user(l, hdr);
971 	}
972 
973 	if (pkt_cnt > 1) {
974 		l->stats.sent_fragmented++;
975 		l->stats.sent_fragments += pkt_cnt;
976 	}
977 
978 	/* Prepare each packet for sending, and add to relevant queue: */
979 	while ((skb = __skb_dequeue(list))) {
980 		if (likely(skb_queue_len(transmq) < maxwin)) {
981 			hdr = buf_msg(skb);
982 			msg_set_seqno(hdr, seqno);
983 			msg_set_ack(hdr, ack);
984 			msg_set_bcast_ack(hdr, bc_ack);
985 			_skb = skb_clone(skb, GFP_ATOMIC);
986 			if (!_skb) {
987 				kfree_skb(skb);
988 				__skb_queue_purge(list);
989 				return -ENOBUFS;
990 			}
991 			__skb_queue_tail(transmq, skb);
992 			/* next retransmit attempt */
993 			if (link_is_bc_sndlink(l))
994 				TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
995 			__skb_queue_tail(xmitq, _skb);
996 			TIPC_SKB_CB(skb)->ackers = l->ackers;
997 			l->rcv_unacked = 0;
998 			l->stats.sent_pkts++;
999 			seqno++;
1000 			continue;
1001 		}
1002 		if (tipc_msg_try_bundle(l->backlog[imp].target_bskb, &skb,
1003 					mtu - INT_H_SIZE, l->addr,
1004 					&new_bundle)) {
1005 			if (skb) {
1006 				/* Keep a ref. to the skb for next try */
1007 				l->backlog[imp].target_bskb = skb;
1008 				l->backlog[imp].len++;
1009 				__skb_queue_tail(backlogq, skb);
1010 			} else {
1011 				if (new_bundle) {
1012 					l->stats.sent_bundles++;
1013 					l->stats.sent_bundled++;
1014 				}
1015 				l->stats.sent_bundled++;
1016 			}
1017 			continue;
1018 		}
1019 		l->backlog[imp].target_bskb = NULL;
1020 		l->backlog[imp].len += (1 + skb_queue_len(list));
1021 		__skb_queue_tail(backlogq, skb);
1022 		skb_queue_splice_tail_init(list, backlogq);
1023 	}
1024 	l->snd_nxt = seqno;
1025 	return rc;
1026 }
1027 
1028 static void tipc_link_advance_backlog(struct tipc_link *l,
1029 				      struct sk_buff_head *xmitq)
1030 {
1031 	struct sk_buff *skb, *_skb;
1032 	struct tipc_msg *hdr;
1033 	u16 seqno = l->snd_nxt;
1034 	u16 ack = l->rcv_nxt - 1;
1035 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1036 	u32 imp;
1037 
1038 	while (skb_queue_len(&l->transmq) < l->window) {
1039 		skb = skb_peek(&l->backlogq);
1040 		if (!skb)
1041 			break;
1042 		_skb = skb_clone(skb, GFP_ATOMIC);
1043 		if (!_skb)
1044 			break;
1045 		__skb_dequeue(&l->backlogq);
1046 		hdr = buf_msg(skb);
1047 		imp = msg_importance(hdr);
1048 		l->backlog[imp].len--;
1049 		if (unlikely(skb == l->backlog[imp].target_bskb))
1050 			l->backlog[imp].target_bskb = NULL;
1051 		__skb_queue_tail(&l->transmq, skb);
1052 		/* next retransmit attempt */
1053 		if (link_is_bc_sndlink(l))
1054 			TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1055 
1056 		__skb_queue_tail(xmitq, _skb);
1057 		TIPC_SKB_CB(skb)->ackers = l->ackers;
1058 		msg_set_seqno(hdr, seqno);
1059 		msg_set_ack(hdr, ack);
1060 		msg_set_bcast_ack(hdr, bc_ack);
1061 		l->rcv_unacked = 0;
1062 		l->stats.sent_pkts++;
1063 		seqno++;
1064 	}
1065 	l->snd_nxt = seqno;
1066 }
1067 
1068 /**
1069  * link_retransmit_failure() - Detect repeated retransmit failures
1070  * @l: tipc link sender
1071  * @r: tipc link receiver (= l in case of unicast)
1072  * @rc: returned code
1073  *
1074  * Return: true if the repeated retransmit failures happens, otherwise
1075  * false
1076  */
1077 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1078 				    int *rc)
1079 {
1080 	struct sk_buff *skb = skb_peek(&l->transmq);
1081 	struct tipc_msg *hdr;
1082 
1083 	if (!skb)
1084 		return false;
1085 
1086 	if (!TIPC_SKB_CB(skb)->retr_cnt)
1087 		return false;
1088 
1089 	if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1090 			msecs_to_jiffies(r->tolerance * 10)))
1091 		return false;
1092 
1093 	hdr = buf_msg(skb);
1094 	if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1095 		return false;
1096 
1097 	pr_warn("Retransmission failure on link <%s>\n", l->name);
1098 	link_print(l, "State of link ");
1099 	pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1100 		msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1101 	pr_info("sqno %u, prev: %x, dest: %x\n",
1102 		msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1103 	pr_info("retr_stamp %d, retr_cnt %d\n",
1104 		jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1105 		TIPC_SKB_CB(skb)->retr_cnt);
1106 
1107 	trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1108 	trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1109 	trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1110 
1111 	if (link_is_bc_sndlink(l)) {
1112 		r->state = LINK_RESET;
1113 		*rc = TIPC_LINK_DOWN_EVT;
1114 	} else {
1115 		*rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1116 	}
1117 
1118 	return true;
1119 }
1120 
1121 /* tipc_link_bc_retrans() - retransmit zero or more packets
1122  * @l: the link to transmit on
1123  * @r: the receiving link ordering the retransmit. Same as l if unicast
1124  * @from: retransmit from (inclusive) this sequence number
1125  * @to: retransmit to (inclusive) this sequence number
1126  * xmitq: queue for accumulating the retransmitted packets
1127  */
1128 static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r,
1129 				u16 from, u16 to, struct sk_buff_head *xmitq)
1130 {
1131 	struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1132 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1133 	u16 ack = l->rcv_nxt - 1;
1134 	struct tipc_msg *hdr;
1135 	int rc = 0;
1136 
1137 	if (!skb)
1138 		return 0;
1139 	if (less(to, from))
1140 		return 0;
1141 
1142 	trace_tipc_link_retrans(r, from, to, &l->transmq);
1143 
1144 	if (link_retransmit_failure(l, r, &rc))
1145 		return rc;
1146 
1147 	skb_queue_walk(&l->transmq, skb) {
1148 		hdr = buf_msg(skb);
1149 		if (less(msg_seqno(hdr), from))
1150 			continue;
1151 		if (more(msg_seqno(hdr), to))
1152 			break;
1153 
1154 		if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1155 			continue;
1156 		TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1157 		_skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE, GFP_ATOMIC);
1158 		if (!_skb)
1159 			return 0;
1160 		hdr = buf_msg(_skb);
1161 		msg_set_ack(hdr, ack);
1162 		msg_set_bcast_ack(hdr, bc_ack);
1163 		_skb->priority = TC_PRIO_CONTROL;
1164 		__skb_queue_tail(xmitq, _skb);
1165 		l->stats.retransmitted++;
1166 
1167 		/* Increase actual retrans counter & mark first time */
1168 		if (!TIPC_SKB_CB(skb)->retr_cnt++)
1169 			TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1170 	}
1171 	return 0;
1172 }
1173 
1174 /* tipc_data_input - deliver data and name distr msgs to upper layer
1175  *
1176  * Consumes buffer if message is of right type
1177  * Node lock must be held
1178  */
1179 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1180 			    struct sk_buff_head *inputq)
1181 {
1182 	struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1183 	struct tipc_msg *hdr = buf_msg(skb);
1184 
1185 	switch (msg_user(hdr)) {
1186 	case TIPC_LOW_IMPORTANCE:
1187 	case TIPC_MEDIUM_IMPORTANCE:
1188 	case TIPC_HIGH_IMPORTANCE:
1189 	case TIPC_CRITICAL_IMPORTANCE:
1190 		if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1191 			skb_queue_tail(mc_inputq, skb);
1192 			return true;
1193 		}
1194 		/* fall through */
1195 	case CONN_MANAGER:
1196 		skb_queue_tail(inputq, skb);
1197 		return true;
1198 	case GROUP_PROTOCOL:
1199 		skb_queue_tail(mc_inputq, skb);
1200 		return true;
1201 	case NAME_DISTRIBUTOR:
1202 		l->bc_rcvlink->state = LINK_ESTABLISHED;
1203 		skb_queue_tail(l->namedq, skb);
1204 		return true;
1205 	case MSG_BUNDLER:
1206 	case TUNNEL_PROTOCOL:
1207 	case MSG_FRAGMENTER:
1208 	case BCAST_PROTOCOL:
1209 		return false;
1210 	default:
1211 		pr_warn("Dropping received illegal msg type\n");
1212 		kfree_skb(skb);
1213 		return true;
1214 	};
1215 }
1216 
1217 /* tipc_link_input - process packet that has passed link protocol check
1218  *
1219  * Consumes buffer
1220  */
1221 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1222 			   struct sk_buff_head *inputq,
1223 			   struct sk_buff **reasm_skb)
1224 {
1225 	struct tipc_msg *hdr = buf_msg(skb);
1226 	struct sk_buff *iskb;
1227 	struct sk_buff_head tmpq;
1228 	int usr = msg_user(hdr);
1229 	int pos = 0;
1230 
1231 	if (usr == MSG_BUNDLER) {
1232 		skb_queue_head_init(&tmpq);
1233 		l->stats.recv_bundles++;
1234 		l->stats.recv_bundled += msg_msgcnt(hdr);
1235 		while (tipc_msg_extract(skb, &iskb, &pos))
1236 			tipc_data_input(l, iskb, &tmpq);
1237 		tipc_skb_queue_splice_tail(&tmpq, inputq);
1238 		return 0;
1239 	} else if (usr == MSG_FRAGMENTER) {
1240 		l->stats.recv_fragments++;
1241 		if (tipc_buf_append(reasm_skb, &skb)) {
1242 			l->stats.recv_fragmented++;
1243 			tipc_data_input(l, skb, inputq);
1244 		} else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1245 			pr_warn_ratelimited("Unable to build fragment list\n");
1246 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1247 		}
1248 		return 0;
1249 	} else if (usr == BCAST_PROTOCOL) {
1250 		tipc_bcast_lock(l->net);
1251 		tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1252 		tipc_bcast_unlock(l->net);
1253 	}
1254 
1255 	kfree_skb(skb);
1256 	return 0;
1257 }
1258 
1259 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1260  *			 inner message along with the ones in the old link's
1261  *			 deferdq
1262  * @l: tunnel link
1263  * @skb: TUNNEL_PROTOCOL message
1264  * @inputq: queue to put messages ready for delivery
1265  */
1266 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1267 			     struct sk_buff_head *inputq)
1268 {
1269 	struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1270 	struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1271 	struct sk_buff_head *fdefq = &l->failover_deferdq;
1272 	struct tipc_msg *hdr = buf_msg(skb);
1273 	struct sk_buff *iskb;
1274 	int ipos = 0;
1275 	int rc = 0;
1276 	u16 seqno;
1277 
1278 	if (msg_type(hdr) == SYNCH_MSG) {
1279 		kfree_skb(skb);
1280 		return 0;
1281 	}
1282 
1283 	/* Not a fragment? */
1284 	if (likely(!msg_nof_fragms(hdr))) {
1285 		if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1286 			pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1287 					    skb_queue_len(fdefq));
1288 			return 0;
1289 		}
1290 		kfree_skb(skb);
1291 	} else {
1292 		/* Set fragment type for buf_append */
1293 		if (msg_fragm_no(hdr) == 1)
1294 			msg_set_type(hdr, FIRST_FRAGMENT);
1295 		else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1296 			msg_set_type(hdr, FRAGMENT);
1297 		else
1298 			msg_set_type(hdr, LAST_FRAGMENT);
1299 
1300 		if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1301 			/* Successful but non-complete reassembly? */
1302 			if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1303 				return 0;
1304 			pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1305 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1306 		}
1307 		iskb = skb;
1308 	}
1309 
1310 	do {
1311 		seqno = buf_seqno(iskb);
1312 		if (unlikely(less(seqno, l->drop_point))) {
1313 			kfree_skb(iskb);
1314 			continue;
1315 		}
1316 		if (unlikely(seqno != l->drop_point)) {
1317 			__tipc_skb_queue_sorted(fdefq, seqno, iskb);
1318 			continue;
1319 		}
1320 
1321 		l->drop_point++;
1322 		if (!tipc_data_input(l, iskb, inputq))
1323 			rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1324 		if (unlikely(rc))
1325 			break;
1326 	} while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1327 
1328 	return rc;
1329 }
1330 
1331 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1332 {
1333 	bool released = false;
1334 	struct sk_buff *skb, *tmp;
1335 
1336 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1337 		if (more(buf_seqno(skb), acked))
1338 			break;
1339 		__skb_unlink(skb, &l->transmq);
1340 		kfree_skb(skb);
1341 		released = true;
1342 	}
1343 	return released;
1344 }
1345 
1346 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1347  * @l: tipc link that data have come with gaps in sequence if any
1348  * @data: data buffer to store the Gap ACK blocks after built
1349  *
1350  * returns the actual allocated memory size
1351  */
1352 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data)
1353 {
1354 	struct sk_buff *skb = skb_peek(&l->deferdq);
1355 	struct tipc_gap_ack_blks *ga = data;
1356 	u16 len, expect, seqno = 0;
1357 	u8 n = 0;
1358 
1359 	if (!skb)
1360 		goto exit;
1361 
1362 	expect = buf_seqno(skb);
1363 	skb_queue_walk(&l->deferdq, skb) {
1364 		seqno = buf_seqno(skb);
1365 		if (unlikely(more(seqno, expect))) {
1366 			ga->gacks[n].ack = htons(expect - 1);
1367 			ga->gacks[n].gap = htons(seqno - expect);
1368 			if (++n >= MAX_GAP_ACK_BLKS) {
1369 				pr_info_ratelimited("Too few Gap ACK blocks!\n");
1370 				goto exit;
1371 			}
1372 		} else if (unlikely(less(seqno, expect))) {
1373 			pr_warn("Unexpected skb in deferdq!\n");
1374 			continue;
1375 		}
1376 		expect = seqno + 1;
1377 	}
1378 
1379 	/* last block */
1380 	ga->gacks[n].ack = htons(seqno);
1381 	ga->gacks[n].gap = 0;
1382 	n++;
1383 
1384 exit:
1385 	len = tipc_gap_ack_blks_sz(n);
1386 	ga->len = htons(len);
1387 	ga->gack_cnt = n;
1388 	return len;
1389 }
1390 
1391 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1392  *			       acked packets, also doing retransmissions if
1393  *			       gaps found
1394  * @l: tipc link with transmq queue to be advanced
1395  * @acked: seqno of last packet acked by peer without any gaps before
1396  * @gap: # of gap packets
1397  * @ga: buffer pointer to Gap ACK blocks from peer
1398  * @xmitq: queue for accumulating the retransmitted packets if any
1399  *
1400  * In case of a repeated retransmit failures, the call will return shortly
1401  * with a returned code (e.g. TIPC_LINK_DOWN_EVT)
1402  */
1403 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
1404 				     struct tipc_gap_ack_blks *ga,
1405 				     struct sk_buff_head *xmitq)
1406 {
1407 	struct sk_buff *skb, *_skb, *tmp;
1408 	struct tipc_msg *hdr;
1409 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1410 	u16 ack = l->rcv_nxt - 1;
1411 	bool passed = false;
1412 	u16 seqno, n = 0;
1413 	int rc = 0;
1414 
1415 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1416 		seqno = buf_seqno(skb);
1417 
1418 next_gap_ack:
1419 		if (less_eq(seqno, acked)) {
1420 			/* release skb */
1421 			__skb_unlink(skb, &l->transmq);
1422 			kfree_skb(skb);
1423 		} else if (less_eq(seqno, acked + gap)) {
1424 			/* First, check if repeated retrans failures occurs? */
1425 			if (!passed && link_retransmit_failure(l, l, &rc))
1426 				return rc;
1427 			passed = true;
1428 
1429 			/* retransmit skb if unrestricted*/
1430 			if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1431 				continue;
1432 			TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
1433 			_skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE,
1434 					   GFP_ATOMIC);
1435 			if (!_skb)
1436 				continue;
1437 			hdr = buf_msg(_skb);
1438 			msg_set_ack(hdr, ack);
1439 			msg_set_bcast_ack(hdr, bc_ack);
1440 			_skb->priority = TC_PRIO_CONTROL;
1441 			__skb_queue_tail(xmitq, _skb);
1442 			l->stats.retransmitted++;
1443 
1444 			/* Increase actual retrans counter & mark first time */
1445 			if (!TIPC_SKB_CB(skb)->retr_cnt++)
1446 				TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1447 		} else {
1448 			/* retry with Gap ACK blocks if any */
1449 			if (!ga || n >= ga->gack_cnt)
1450 				break;
1451 			acked = ntohs(ga->gacks[n].ack);
1452 			gap = ntohs(ga->gacks[n].gap);
1453 			n++;
1454 			goto next_gap_ack;
1455 		}
1456 	}
1457 
1458 	return 0;
1459 }
1460 
1461 /* tipc_link_build_state_msg: prepare link state message for transmission
1462  *
1463  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1464  * risk of ack storms towards the sender
1465  */
1466 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1467 {
1468 	if (!l)
1469 		return 0;
1470 
1471 	/* Broadcast ACK must be sent via a unicast link => defer to caller */
1472 	if (link_is_bc_rcvlink(l)) {
1473 		if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1474 			return 0;
1475 		l->rcv_unacked = 0;
1476 
1477 		/* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1478 		l->snd_nxt = l->rcv_nxt;
1479 		return TIPC_LINK_SND_STATE;
1480 	}
1481 
1482 	/* Unicast ACK */
1483 	l->rcv_unacked = 0;
1484 	l->stats.sent_acks++;
1485 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1486 	return 0;
1487 }
1488 
1489 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1490  */
1491 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1492 {
1493 	int mtyp = RESET_MSG;
1494 	struct sk_buff *skb;
1495 
1496 	if (l->state == LINK_ESTABLISHING)
1497 		mtyp = ACTIVATE_MSG;
1498 
1499 	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1500 
1501 	/* Inform peer that this endpoint is going down if applicable */
1502 	skb = skb_peek_tail(xmitq);
1503 	if (skb && (l->state == LINK_RESET))
1504 		msg_set_peer_stopping(buf_msg(skb), 1);
1505 }
1506 
1507 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1508  * Note that sending of broadcast NACK is coordinated among nodes, to
1509  * reduce the risk of NACK storms towards the sender
1510  */
1511 static int tipc_link_build_nack_msg(struct tipc_link *l,
1512 				    struct sk_buff_head *xmitq)
1513 {
1514 	u32 def_cnt = ++l->stats.deferred_recv;
1515 	u32 defq_len = skb_queue_len(&l->deferdq);
1516 	int match1, match2;
1517 
1518 	if (link_is_bc_rcvlink(l)) {
1519 		match1 = def_cnt & 0xf;
1520 		match2 = tipc_own_addr(l->net) & 0xf;
1521 		if (match1 == match2)
1522 			return TIPC_LINK_SND_STATE;
1523 		return 0;
1524 	}
1525 
1526 	if (defq_len >= 3 && !((defq_len - 3) % 16))
1527 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1528 	return 0;
1529 }
1530 
1531 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1532  * @l: the link that should handle the message
1533  * @skb: TIPC packet
1534  * @xmitq: queue to place packets to be sent after this call
1535  */
1536 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1537 		  struct sk_buff_head *xmitq)
1538 {
1539 	struct sk_buff_head *defq = &l->deferdq;
1540 	struct tipc_msg *hdr = buf_msg(skb);
1541 	u16 seqno, rcv_nxt, win_lim;
1542 	int rc = 0;
1543 
1544 	/* Verify and update link state */
1545 	if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1546 		return tipc_link_proto_rcv(l, skb, xmitq);
1547 
1548 	/* Don't send probe at next timeout expiration */
1549 	l->silent_intv_cnt = 0;
1550 
1551 	do {
1552 		hdr = buf_msg(skb);
1553 		seqno = msg_seqno(hdr);
1554 		rcv_nxt = l->rcv_nxt;
1555 		win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1556 
1557 		if (unlikely(!link_is_up(l))) {
1558 			if (l->state == LINK_ESTABLISHING)
1559 				rc = TIPC_LINK_UP_EVT;
1560 			goto drop;
1561 		}
1562 
1563 		/* Drop if outside receive window */
1564 		if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1565 			l->stats.duplicates++;
1566 			goto drop;
1567 		}
1568 
1569 		/* Forward queues and wake up waiting users */
1570 		if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1571 			tipc_link_advance_backlog(l, xmitq);
1572 			if (unlikely(!skb_queue_empty(&l->wakeupq)))
1573 				link_prepare_wakeup(l);
1574 		}
1575 
1576 		/* Defer delivery if sequence gap */
1577 		if (unlikely(seqno != rcv_nxt)) {
1578 			__tipc_skb_queue_sorted(defq, seqno, skb);
1579 			rc |= tipc_link_build_nack_msg(l, xmitq);
1580 			break;
1581 		}
1582 
1583 		/* Deliver packet */
1584 		l->rcv_nxt++;
1585 		l->stats.recv_pkts++;
1586 
1587 		if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1588 			rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1589 		else if (!tipc_data_input(l, skb, l->inputq))
1590 			rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1591 		if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1592 			rc |= tipc_link_build_state_msg(l, xmitq);
1593 		if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1594 			break;
1595 	} while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1596 
1597 	return rc;
1598 drop:
1599 	kfree_skb(skb);
1600 	return rc;
1601 }
1602 
1603 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1604 				      bool probe_reply, u16 rcvgap,
1605 				      int tolerance, int priority,
1606 				      struct sk_buff_head *xmitq)
1607 {
1608 	struct tipc_link *bcl = l->bc_rcvlink;
1609 	struct sk_buff *skb;
1610 	struct tipc_msg *hdr;
1611 	struct sk_buff_head *dfq = &l->deferdq;
1612 	bool node_up = link_is_up(bcl);
1613 	struct tipc_mon_state *mstate = &l->mon_state;
1614 	int dlen = 0;
1615 	void *data;
1616 	u16 glen = 0;
1617 
1618 	/* Don't send protocol message during reset or link failover */
1619 	if (tipc_link_is_blocked(l))
1620 		return;
1621 
1622 	if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1623 		return;
1624 
1625 	if (!skb_queue_empty(dfq))
1626 		rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1627 
1628 	skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1629 			      tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1630 			      l->addr, tipc_own_addr(l->net), 0, 0, 0);
1631 	if (!skb)
1632 		return;
1633 
1634 	hdr = buf_msg(skb);
1635 	data = msg_data(hdr);
1636 	msg_set_session(hdr, l->session);
1637 	msg_set_bearer_id(hdr, l->bearer_id);
1638 	msg_set_net_plane(hdr, l->net_plane);
1639 	msg_set_next_sent(hdr, l->snd_nxt);
1640 	msg_set_ack(hdr, l->rcv_nxt - 1);
1641 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1642 	msg_set_bc_ack_invalid(hdr, !node_up);
1643 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1644 	msg_set_link_tolerance(hdr, tolerance);
1645 	msg_set_linkprio(hdr, priority);
1646 	msg_set_redundant_link(hdr, node_up);
1647 	msg_set_seq_gap(hdr, 0);
1648 	msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1649 
1650 	if (mtyp == STATE_MSG) {
1651 		if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1652 			msg_set_seqno(hdr, l->snd_nxt_state++);
1653 		msg_set_seq_gap(hdr, rcvgap);
1654 		msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1655 		msg_set_probe(hdr, probe);
1656 		msg_set_is_keepalive(hdr, probe || probe_reply);
1657 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1658 			glen = tipc_build_gap_ack_blks(l, data);
1659 		tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1660 		msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1661 		skb_trim(skb, INT_H_SIZE + glen + dlen);
1662 		l->stats.sent_states++;
1663 		l->rcv_unacked = 0;
1664 	} else {
1665 		/* RESET_MSG or ACTIVATE_MSG */
1666 		if (mtyp == ACTIVATE_MSG) {
1667 			msg_set_dest_session_valid(hdr, 1);
1668 			msg_set_dest_session(hdr, l->peer_session);
1669 		}
1670 		msg_set_max_pkt(hdr, l->advertised_mtu);
1671 		strcpy(data, l->if_name);
1672 		msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1673 		skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1674 	}
1675 	if (probe)
1676 		l->stats.sent_probes++;
1677 	if (rcvgap)
1678 		l->stats.sent_nacks++;
1679 	skb->priority = TC_PRIO_CONTROL;
1680 	__skb_queue_tail(xmitq, skb);
1681 	trace_tipc_proto_build(skb, false, l->name);
1682 }
1683 
1684 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1685 				    struct sk_buff_head *xmitq)
1686 {
1687 	u32 onode = tipc_own_addr(l->net);
1688 	struct tipc_msg *hdr, *ihdr;
1689 	struct sk_buff_head tnlq;
1690 	struct sk_buff *skb;
1691 	u32 dnode = l->addr;
1692 
1693 	__skb_queue_head_init(&tnlq);
1694 	skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1695 			      INT_H_SIZE, BASIC_H_SIZE,
1696 			      dnode, onode, 0, 0, 0);
1697 	if (!skb) {
1698 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1699 		return;
1700 	}
1701 
1702 	hdr = buf_msg(skb);
1703 	msg_set_msgcnt(hdr, 1);
1704 	msg_set_bearer_id(hdr, l->peer_bearer_id);
1705 
1706 	ihdr = (struct tipc_msg *)msg_data(hdr);
1707 	tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1708 		      BASIC_H_SIZE, dnode);
1709 	msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1710 	__skb_queue_tail(&tnlq, skb);
1711 	tipc_link_xmit(l, &tnlq, xmitq);
1712 }
1713 
1714 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1715  * with contents of the link's transmit and backlog queues.
1716  */
1717 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1718 			   int mtyp, struct sk_buff_head *xmitq)
1719 {
1720 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1721 	struct sk_buff *skb, *tnlskb;
1722 	struct tipc_msg *hdr, tnlhdr;
1723 	struct sk_buff_head *queue = &l->transmq;
1724 	struct sk_buff_head tmpxq, tnlq, frags;
1725 	u16 pktlen, pktcnt, seqno = l->snd_nxt;
1726 	bool pktcnt_need_update = false;
1727 	u16 syncpt;
1728 	int rc;
1729 
1730 	if (!tnl)
1731 		return;
1732 
1733 	__skb_queue_head_init(&tnlq);
1734 	/* Link Synching:
1735 	 * From now on, send only one single ("dummy") SYNCH message
1736 	 * to peer. The SYNCH message does not contain any data, just
1737 	 * a header conveying the synch point to the peer.
1738 	 */
1739 	if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1740 		tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1741 					 INT_H_SIZE, 0, l->addr,
1742 					 tipc_own_addr(l->net),
1743 					 0, 0, 0);
1744 		if (!tnlskb) {
1745 			pr_warn("%sunable to create dummy SYNCH_MSG\n",
1746 				link_co_err);
1747 			return;
1748 		}
1749 
1750 		hdr = buf_msg(tnlskb);
1751 		syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
1752 		msg_set_syncpt(hdr, syncpt);
1753 		msg_set_bearer_id(hdr, l->peer_bearer_id);
1754 		__skb_queue_tail(&tnlq, tnlskb);
1755 		tipc_link_xmit(tnl, &tnlq, xmitq);
1756 		return;
1757 	}
1758 
1759 	__skb_queue_head_init(&tmpxq);
1760 	__skb_queue_head_init(&frags);
1761 	/* At least one packet required for safe algorithm => add dummy */
1762 	skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1763 			      BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1764 			      0, 0, TIPC_ERR_NO_PORT);
1765 	if (!skb) {
1766 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1767 		return;
1768 	}
1769 	__skb_queue_tail(&tnlq, skb);
1770 	tipc_link_xmit(l, &tnlq, &tmpxq);
1771 	__skb_queue_purge(&tmpxq);
1772 
1773 	/* Initialize reusable tunnel packet header */
1774 	tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1775 		      mtyp, INT_H_SIZE, l->addr);
1776 	if (mtyp == SYNCH_MSG)
1777 		pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
1778 	else
1779 		pktcnt = skb_queue_len(&l->transmq);
1780 	pktcnt += skb_queue_len(&l->backlogq);
1781 	msg_set_msgcnt(&tnlhdr, pktcnt);
1782 	msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1783 tnl:
1784 	/* Wrap each packet into a tunnel packet */
1785 	skb_queue_walk(queue, skb) {
1786 		hdr = buf_msg(skb);
1787 		if (queue == &l->backlogq)
1788 			msg_set_seqno(hdr, seqno++);
1789 		pktlen = msg_size(hdr);
1790 
1791 		/* Tunnel link MTU is not large enough? This could be
1792 		 * due to:
1793 		 * 1) Link MTU has just changed or set differently;
1794 		 * 2) Or FAILOVER on the top of a SYNCH message
1795 		 *
1796 		 * The 2nd case should not happen if peer supports
1797 		 * TIPC_TUNNEL_ENHANCED
1798 		 */
1799 		if (pktlen > tnl->mtu - INT_H_SIZE) {
1800 			if (mtyp == FAILOVER_MSG &&
1801 			    (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1802 				rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
1803 						       &frags);
1804 				if (rc) {
1805 					pr_warn("%sunable to frag msg: rc %d\n",
1806 						link_co_err, rc);
1807 					return;
1808 				}
1809 				pktcnt += skb_queue_len(&frags) - 1;
1810 				pktcnt_need_update = true;
1811 				skb_queue_splice_tail_init(&frags, &tnlq);
1812 				continue;
1813 			}
1814 			/* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
1815 			 * => Just warn it and return!
1816 			 */
1817 			pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
1818 					    link_co_err, msg_user(hdr),
1819 					    msg_type(hdr), msg_size(hdr));
1820 			return;
1821 		}
1822 
1823 		msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1824 		tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1825 		if (!tnlskb) {
1826 			pr_warn("%sunable to send packet\n", link_co_err);
1827 			return;
1828 		}
1829 		skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1830 		skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1831 		__skb_queue_tail(&tnlq, tnlskb);
1832 	}
1833 	if (queue != &l->backlogq) {
1834 		queue = &l->backlogq;
1835 		goto tnl;
1836 	}
1837 
1838 	if (pktcnt_need_update)
1839 		skb_queue_walk(&tnlq, skb) {
1840 			hdr = buf_msg(skb);
1841 			msg_set_msgcnt(hdr, pktcnt);
1842 		}
1843 
1844 	tipc_link_xmit(tnl, &tnlq, xmitq);
1845 
1846 	if (mtyp == FAILOVER_MSG) {
1847 		tnl->drop_point = l->rcv_nxt;
1848 		tnl->failover_reasm_skb = l->reasm_buf;
1849 		l->reasm_buf = NULL;
1850 
1851 		/* Failover the link's deferdq */
1852 		if (unlikely(!skb_queue_empty(fdefq))) {
1853 			pr_warn("Link failover deferdq not empty: %d!\n",
1854 				skb_queue_len(fdefq));
1855 			__skb_queue_purge(fdefq);
1856 		}
1857 		skb_queue_splice_init(&l->deferdq, fdefq);
1858 	}
1859 }
1860 
1861 /**
1862  * tipc_link_failover_prepare() - prepare tnl for link failover
1863  *
1864  * This is a special version of the precursor - tipc_link_tnl_prepare(),
1865  * see the tipc_node_link_failover() for details
1866  *
1867  * @l: failover link
1868  * @tnl: tunnel link
1869  * @xmitq: queue for messages to be xmited
1870  */
1871 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
1872 				struct sk_buff_head *xmitq)
1873 {
1874 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1875 
1876 	tipc_link_create_dummy_tnl_msg(tnl, xmitq);
1877 
1878 	/* This failover link endpoint was never established before,
1879 	 * so it has not received anything from peer.
1880 	 * Otherwise, it must be a normal failover situation or the
1881 	 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
1882 	 * would have to start over from scratch instead.
1883 	 */
1884 	tnl->drop_point = 1;
1885 	tnl->failover_reasm_skb = NULL;
1886 
1887 	/* Initiate the link's failover deferdq */
1888 	if (unlikely(!skb_queue_empty(fdefq))) {
1889 		pr_warn("Link failover deferdq not empty: %d!\n",
1890 			skb_queue_len(fdefq));
1891 		__skb_queue_purge(fdefq);
1892 	}
1893 }
1894 
1895 /* tipc_link_validate_msg(): validate message against current link state
1896  * Returns true if message should be accepted, otherwise false
1897  */
1898 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
1899 {
1900 	u16 curr_session = l->peer_session;
1901 	u16 session = msg_session(hdr);
1902 	int mtyp = msg_type(hdr);
1903 
1904 	if (msg_user(hdr) != LINK_PROTOCOL)
1905 		return true;
1906 
1907 	switch (mtyp) {
1908 	case RESET_MSG:
1909 		if (!l->in_session)
1910 			return true;
1911 		/* Accept only RESET with new session number */
1912 		return more(session, curr_session);
1913 	case ACTIVATE_MSG:
1914 		if (!l->in_session)
1915 			return true;
1916 		/* Accept only ACTIVATE with new or current session number */
1917 		return !less(session, curr_session);
1918 	case STATE_MSG:
1919 		/* Accept only STATE with current session number */
1920 		if (!l->in_session)
1921 			return false;
1922 		if (session != curr_session)
1923 			return false;
1924 		/* Extra sanity check */
1925 		if (!link_is_up(l) && msg_ack(hdr))
1926 			return false;
1927 		if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
1928 			return true;
1929 		/* Accept only STATE with new sequence number */
1930 		return !less(msg_seqno(hdr), l->rcv_nxt_state);
1931 	default:
1932 		return false;
1933 	}
1934 }
1935 
1936 /* tipc_link_proto_rcv(): receive link level protocol message :
1937  * Note that network plane id propagates through the network, and may
1938  * change at any time. The node with lowest numerical id determines
1939  * network plane
1940  */
1941 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1942 			       struct sk_buff_head *xmitq)
1943 {
1944 	struct tipc_msg *hdr = buf_msg(skb);
1945 	struct tipc_gap_ack_blks *ga = NULL;
1946 	u16 rcvgap = 0;
1947 	u16 ack = msg_ack(hdr);
1948 	u16 gap = msg_seq_gap(hdr);
1949 	u16 peers_snd_nxt =  msg_next_sent(hdr);
1950 	u16 peers_tol = msg_link_tolerance(hdr);
1951 	u16 peers_prio = msg_linkprio(hdr);
1952 	u16 rcv_nxt = l->rcv_nxt;
1953 	u16 dlen = msg_data_sz(hdr);
1954 	int mtyp = msg_type(hdr);
1955 	bool reply = msg_probe(hdr);
1956 	u16 glen = 0;
1957 	void *data;
1958 	char *if_name;
1959 	int rc = 0;
1960 
1961 	trace_tipc_proto_rcv(skb, false, l->name);
1962 	if (tipc_link_is_blocked(l) || !xmitq)
1963 		goto exit;
1964 
1965 	if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1966 		l->net_plane = msg_net_plane(hdr);
1967 
1968 	skb_linearize(skb);
1969 	hdr = buf_msg(skb);
1970 	data = msg_data(hdr);
1971 
1972 	if (!tipc_link_validate_msg(l, hdr)) {
1973 		trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
1974 		trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
1975 		goto exit;
1976 	}
1977 
1978 	switch (mtyp) {
1979 	case RESET_MSG:
1980 	case ACTIVATE_MSG:
1981 		/* Complete own link name with peer's interface name */
1982 		if_name =  strrchr(l->name, ':') + 1;
1983 		if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1984 			break;
1985 		if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1986 			break;
1987 		strncpy(if_name, data, TIPC_MAX_IF_NAME);
1988 
1989 		/* Update own tolerance if peer indicates a non-zero value */
1990 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1991 			l->tolerance = peers_tol;
1992 			l->bc_rcvlink->tolerance = peers_tol;
1993 		}
1994 		/* Update own priority if peer's priority is higher */
1995 		if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1996 			l->priority = peers_prio;
1997 
1998 		/* If peer is going down we want full re-establish cycle */
1999 		if (msg_peer_stopping(hdr)) {
2000 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2001 			break;
2002 		}
2003 
2004 		/* If this endpoint was re-created while peer was ESTABLISHING
2005 		 * it doesn't know current session number. Force re-synch.
2006 		 */
2007 		if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2008 		    l->session != msg_dest_session(hdr)) {
2009 			if (less(l->session, msg_dest_session(hdr)))
2010 				l->session = msg_dest_session(hdr) + 1;
2011 			break;
2012 		}
2013 
2014 		/* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2015 		if (mtyp == RESET_MSG || !link_is_up(l))
2016 			rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2017 
2018 		/* ACTIVATE_MSG takes up link if it was already locally reset */
2019 		if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2020 			rc = TIPC_LINK_UP_EVT;
2021 
2022 		l->peer_session = msg_session(hdr);
2023 		l->in_session = true;
2024 		l->peer_bearer_id = msg_bearer_id(hdr);
2025 		if (l->mtu > msg_max_pkt(hdr))
2026 			l->mtu = msg_max_pkt(hdr);
2027 		break;
2028 
2029 	case STATE_MSG:
2030 		l->rcv_nxt_state = msg_seqno(hdr) + 1;
2031 
2032 		/* Update own tolerance if peer indicates a non-zero value */
2033 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2034 			l->tolerance = peers_tol;
2035 			l->bc_rcvlink->tolerance = peers_tol;
2036 		}
2037 		/* Update own prio if peer indicates a different value */
2038 		if ((peers_prio != l->priority) &&
2039 		    in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2040 			l->priority = peers_prio;
2041 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2042 		}
2043 
2044 		l->silent_intv_cnt = 0;
2045 		l->stats.recv_states++;
2046 		if (msg_probe(hdr))
2047 			l->stats.recv_probes++;
2048 
2049 		if (!link_is_up(l)) {
2050 			if (l->state == LINK_ESTABLISHING)
2051 				rc = TIPC_LINK_UP_EVT;
2052 			break;
2053 		}
2054 
2055 		/* Receive Gap ACK blocks from peer if any */
2056 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
2057 			ga = (struct tipc_gap_ack_blks *)data;
2058 			glen = ntohs(ga->len);
2059 			/* sanity check: if failed, ignore Gap ACK blocks */
2060 			if (glen != tipc_gap_ack_blks_sz(ga->gack_cnt))
2061 				ga = NULL;
2062 		}
2063 
2064 		tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2065 			     &l->mon_state, l->bearer_id);
2066 
2067 		/* Send NACK if peer has sent pkts we haven't received yet */
2068 		if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
2069 			rcvgap = peers_snd_nxt - l->rcv_nxt;
2070 		if (rcvgap || reply)
2071 			tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2072 						  rcvgap, 0, 0, xmitq);
2073 
2074 		rc |= tipc_link_advance_transmq(l, ack, gap, ga, xmitq);
2075 
2076 		/* If NACK, retransmit will now start at right position */
2077 		if (gap)
2078 			l->stats.recv_nacks++;
2079 
2080 		tipc_link_advance_backlog(l, xmitq);
2081 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
2082 			link_prepare_wakeup(l);
2083 	}
2084 exit:
2085 	kfree_skb(skb);
2086 	return rc;
2087 }
2088 
2089 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2090  */
2091 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2092 					 u16 peers_snd_nxt,
2093 					 struct sk_buff_head *xmitq)
2094 {
2095 	struct sk_buff *skb;
2096 	struct tipc_msg *hdr;
2097 	struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2098 	u16 ack = l->rcv_nxt - 1;
2099 	u16 gap_to = peers_snd_nxt - 1;
2100 
2101 	skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2102 			      0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2103 	if (!skb)
2104 		return false;
2105 	hdr = buf_msg(skb);
2106 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2107 	msg_set_bcast_ack(hdr, ack);
2108 	msg_set_bcgap_after(hdr, ack);
2109 	if (dfrd_skb)
2110 		gap_to = buf_seqno(dfrd_skb) - 1;
2111 	msg_set_bcgap_to(hdr, gap_to);
2112 	msg_set_non_seq(hdr, bcast);
2113 	__skb_queue_tail(xmitq, skb);
2114 	return true;
2115 }
2116 
2117 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2118  *
2119  * Give a newly added peer node the sequence number where it should
2120  * start receiving and acking broadcast packets.
2121  */
2122 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2123 					struct sk_buff_head *xmitq)
2124 {
2125 	struct sk_buff_head list;
2126 
2127 	__skb_queue_head_init(&list);
2128 	if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2129 		return;
2130 	msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2131 	tipc_link_xmit(l, &list, xmitq);
2132 }
2133 
2134 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2135  */
2136 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2137 {
2138 	int mtyp = msg_type(hdr);
2139 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2140 
2141 	if (link_is_up(l))
2142 		return;
2143 
2144 	if (msg_user(hdr) == BCAST_PROTOCOL) {
2145 		l->rcv_nxt = peers_snd_nxt;
2146 		l->state = LINK_ESTABLISHED;
2147 		return;
2148 	}
2149 
2150 	if (l->peer_caps & TIPC_BCAST_SYNCH)
2151 		return;
2152 
2153 	if (msg_peer_node_is_up(hdr))
2154 		return;
2155 
2156 	/* Compatibility: accept older, less safe initial synch data */
2157 	if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2158 		l->rcv_nxt = peers_snd_nxt;
2159 }
2160 
2161 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2162  */
2163 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2164 			  struct sk_buff_head *xmitq)
2165 {
2166 	struct tipc_link *snd_l = l->bc_sndlink;
2167 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2168 	u16 from = msg_bcast_ack(hdr) + 1;
2169 	u16 to = from + msg_bc_gap(hdr) - 1;
2170 	int rc = 0;
2171 
2172 	if (!link_is_up(l))
2173 		return rc;
2174 
2175 	if (!msg_peer_node_is_up(hdr))
2176 		return rc;
2177 
2178 	/* Open when peer ackowledges our bcast init msg (pkt #1) */
2179 	if (msg_ack(hdr))
2180 		l->bc_peer_is_up = true;
2181 
2182 	if (!l->bc_peer_is_up)
2183 		return rc;
2184 
2185 	l->stats.recv_nacks++;
2186 
2187 	/* Ignore if peers_snd_nxt goes beyond receive window */
2188 	if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2189 		return rc;
2190 
2191 	rc = tipc_link_bc_retrans(snd_l, l, from, to, xmitq);
2192 
2193 	l->snd_nxt = peers_snd_nxt;
2194 	if (link_bc_rcv_gap(l))
2195 		rc |= TIPC_LINK_SND_STATE;
2196 
2197 	/* Return now if sender supports nack via STATE messages */
2198 	if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2199 		return rc;
2200 
2201 	/* Otherwise, be backwards compatible */
2202 
2203 	if (!more(peers_snd_nxt, l->rcv_nxt)) {
2204 		l->nack_state = BC_NACK_SND_CONDITIONAL;
2205 		return 0;
2206 	}
2207 
2208 	/* Don't NACK if one was recently sent or peeked */
2209 	if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2210 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2211 		return 0;
2212 	}
2213 
2214 	/* Conditionally delay NACK sending until next synch rcv */
2215 	if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2216 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2217 		if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2218 			return 0;
2219 	}
2220 
2221 	/* Send NACK now but suppress next one */
2222 	tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2223 	l->nack_state = BC_NACK_SND_SUPPRESS;
2224 	return 0;
2225 }
2226 
2227 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
2228 			  struct sk_buff_head *xmitq)
2229 {
2230 	struct sk_buff *skb, *tmp;
2231 	struct tipc_link *snd_l = l->bc_sndlink;
2232 
2233 	if (!link_is_up(l) || !l->bc_peer_is_up)
2234 		return;
2235 
2236 	if (!more(acked, l->acked))
2237 		return;
2238 
2239 	trace_tipc_link_bc_ack(l, l->acked, acked, &snd_l->transmq);
2240 	/* Skip over packets peer has already acked */
2241 	skb_queue_walk(&snd_l->transmq, skb) {
2242 		if (more(buf_seqno(skb), l->acked))
2243 			break;
2244 	}
2245 
2246 	/* Update/release the packets peer is acking now */
2247 	skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
2248 		if (more(buf_seqno(skb), acked))
2249 			break;
2250 		if (!--TIPC_SKB_CB(skb)->ackers) {
2251 			__skb_unlink(skb, &snd_l->transmq);
2252 			kfree_skb(skb);
2253 		}
2254 	}
2255 	l->acked = acked;
2256 	tipc_link_advance_backlog(snd_l, xmitq);
2257 	if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
2258 		link_prepare_wakeup(snd_l);
2259 }
2260 
2261 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2262  * This function is here for backwards compatibility, since
2263  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2264  */
2265 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2266 			  struct sk_buff_head *xmitq)
2267 {
2268 	struct tipc_msg *hdr = buf_msg(skb);
2269 	u32 dnode = msg_destnode(hdr);
2270 	int mtyp = msg_type(hdr);
2271 	u16 acked = msg_bcast_ack(hdr);
2272 	u16 from = acked + 1;
2273 	u16 to = msg_bcgap_to(hdr);
2274 	u16 peers_snd_nxt = to + 1;
2275 	int rc = 0;
2276 
2277 	kfree_skb(skb);
2278 
2279 	if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2280 		return 0;
2281 
2282 	if (mtyp != STATE_MSG)
2283 		return 0;
2284 
2285 	if (dnode == tipc_own_addr(l->net)) {
2286 		tipc_link_bc_ack_rcv(l, acked, xmitq);
2287 		rc = tipc_link_bc_retrans(l->bc_sndlink, l, from, to, xmitq);
2288 		l->stats.recv_nacks++;
2289 		return rc;
2290 	}
2291 
2292 	/* Msg for other node => suppress own NACK at next sync if applicable */
2293 	if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2294 		l->nack_state = BC_NACK_SND_SUPPRESS;
2295 
2296 	return 0;
2297 }
2298 
2299 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
2300 {
2301 	int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2302 
2303 	l->window = win;
2304 	l->backlog[TIPC_LOW_IMPORTANCE].limit      = max_t(u16, 50, win);
2305 	l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = max_t(u16, 100, win * 2);
2306 	l->backlog[TIPC_HIGH_IMPORTANCE].limit     = max_t(u16, 150, win * 3);
2307 	l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
2308 	l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
2309 }
2310 
2311 /**
2312  * link_reset_stats - reset link statistics
2313  * @l: pointer to link
2314  */
2315 void tipc_link_reset_stats(struct tipc_link *l)
2316 {
2317 	memset(&l->stats, 0, sizeof(l->stats));
2318 }
2319 
2320 static void link_print(struct tipc_link *l, const char *str)
2321 {
2322 	struct sk_buff *hskb = skb_peek(&l->transmq);
2323 	u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2324 	u16 tail = l->snd_nxt - 1;
2325 
2326 	pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2327 	pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2328 		skb_queue_len(&l->transmq), head, tail,
2329 		skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2330 }
2331 
2332 /* Parse and validate nested (link) properties valid for media, bearer and link
2333  */
2334 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2335 {
2336 	int err;
2337 
2338 	err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2339 					  tipc_nl_prop_policy, NULL);
2340 	if (err)
2341 		return err;
2342 
2343 	if (props[TIPC_NLA_PROP_PRIO]) {
2344 		u32 prio;
2345 
2346 		prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2347 		if (prio > TIPC_MAX_LINK_PRI)
2348 			return -EINVAL;
2349 	}
2350 
2351 	if (props[TIPC_NLA_PROP_TOL]) {
2352 		u32 tol;
2353 
2354 		tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2355 		if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2356 			return -EINVAL;
2357 	}
2358 
2359 	if (props[TIPC_NLA_PROP_WIN]) {
2360 		u32 win;
2361 
2362 		win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2363 		if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
2364 			return -EINVAL;
2365 	}
2366 
2367 	return 0;
2368 }
2369 
2370 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2371 {
2372 	int i;
2373 	struct nlattr *stats;
2374 
2375 	struct nla_map {
2376 		u32 key;
2377 		u32 val;
2378 	};
2379 
2380 	struct nla_map map[] = {
2381 		{TIPC_NLA_STATS_RX_INFO, 0},
2382 		{TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2383 		{TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2384 		{TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2385 		{TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2386 		{TIPC_NLA_STATS_TX_INFO, 0},
2387 		{TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2388 		{TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2389 		{TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2390 		{TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2391 		{TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2392 			s->msg_length_counts : 1},
2393 		{TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2394 		{TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2395 		{TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2396 		{TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2397 		{TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2398 		{TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2399 		{TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2400 		{TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2401 		{TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2402 		{TIPC_NLA_STATS_RX_STATES, s->recv_states},
2403 		{TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2404 		{TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2405 		{TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2406 		{TIPC_NLA_STATS_TX_STATES, s->sent_states},
2407 		{TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2408 		{TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2409 		{TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2410 		{TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2411 		{TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2412 		{TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2413 		{TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2414 		{TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2415 			(s->accu_queue_sz / s->queue_sz_counts) : 0}
2416 	};
2417 
2418 	stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2419 	if (!stats)
2420 		return -EMSGSIZE;
2421 
2422 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2423 		if (nla_put_u32(skb, map[i].key, map[i].val))
2424 			goto msg_full;
2425 
2426 	nla_nest_end(skb, stats);
2427 
2428 	return 0;
2429 msg_full:
2430 	nla_nest_cancel(skb, stats);
2431 
2432 	return -EMSGSIZE;
2433 }
2434 
2435 /* Caller should hold appropriate locks to protect the link */
2436 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2437 		       struct tipc_link *link, int nlflags)
2438 {
2439 	u32 self = tipc_own_addr(net);
2440 	struct nlattr *attrs;
2441 	struct nlattr *prop;
2442 	void *hdr;
2443 	int err;
2444 
2445 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2446 			  nlflags, TIPC_NL_LINK_GET);
2447 	if (!hdr)
2448 		return -EMSGSIZE;
2449 
2450 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2451 	if (!attrs)
2452 		goto msg_full;
2453 
2454 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2455 		goto attr_msg_full;
2456 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2457 		goto attr_msg_full;
2458 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2459 		goto attr_msg_full;
2460 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2461 		goto attr_msg_full;
2462 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2463 		goto attr_msg_full;
2464 
2465 	if (tipc_link_is_up(link))
2466 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2467 			goto attr_msg_full;
2468 	if (link->active)
2469 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2470 			goto attr_msg_full;
2471 
2472 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2473 	if (!prop)
2474 		goto attr_msg_full;
2475 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2476 		goto prop_msg_full;
2477 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2478 		goto prop_msg_full;
2479 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2480 			link->window))
2481 		goto prop_msg_full;
2482 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2483 		goto prop_msg_full;
2484 	nla_nest_end(msg->skb, prop);
2485 
2486 	err = __tipc_nl_add_stats(msg->skb, &link->stats);
2487 	if (err)
2488 		goto attr_msg_full;
2489 
2490 	nla_nest_end(msg->skb, attrs);
2491 	genlmsg_end(msg->skb, hdr);
2492 
2493 	return 0;
2494 
2495 prop_msg_full:
2496 	nla_nest_cancel(msg->skb, prop);
2497 attr_msg_full:
2498 	nla_nest_cancel(msg->skb, attrs);
2499 msg_full:
2500 	genlmsg_cancel(msg->skb, hdr);
2501 
2502 	return -EMSGSIZE;
2503 }
2504 
2505 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2506 				      struct tipc_stats *stats)
2507 {
2508 	int i;
2509 	struct nlattr *nest;
2510 
2511 	struct nla_map {
2512 		__u32 key;
2513 		__u32 val;
2514 	};
2515 
2516 	struct nla_map map[] = {
2517 		{TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2518 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2519 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2520 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2521 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2522 		{TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2523 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2524 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2525 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2526 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2527 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2528 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2529 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2530 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2531 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2532 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2533 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2534 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2535 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2536 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2537 	};
2538 
2539 	nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2540 	if (!nest)
2541 		return -EMSGSIZE;
2542 
2543 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2544 		if (nla_put_u32(skb, map[i].key, map[i].val))
2545 			goto msg_full;
2546 
2547 	nla_nest_end(skb, nest);
2548 
2549 	return 0;
2550 msg_full:
2551 	nla_nest_cancel(skb, nest);
2552 
2553 	return -EMSGSIZE;
2554 }
2555 
2556 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2557 {
2558 	int err;
2559 	void *hdr;
2560 	struct nlattr *attrs;
2561 	struct nlattr *prop;
2562 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2563 	u32 bc_mode = tipc_bcast_get_broadcast_mode(net);
2564 	u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2565 	struct tipc_link *bcl = tn->bcl;
2566 
2567 	if (!bcl)
2568 		return 0;
2569 
2570 	tipc_bcast_lock(net);
2571 
2572 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2573 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
2574 	if (!hdr) {
2575 		tipc_bcast_unlock(net);
2576 		return -EMSGSIZE;
2577 	}
2578 
2579 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2580 	if (!attrs)
2581 		goto msg_full;
2582 
2583 	/* The broadcast link is always up */
2584 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2585 		goto attr_msg_full;
2586 
2587 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2588 		goto attr_msg_full;
2589 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2590 		goto attr_msg_full;
2591 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2592 		goto attr_msg_full;
2593 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2594 		goto attr_msg_full;
2595 
2596 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2597 	if (!prop)
2598 		goto attr_msg_full;
2599 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2600 		goto prop_msg_full;
2601 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2602 		goto prop_msg_full;
2603 	if (bc_mode & BCLINK_MODE_SEL)
2604 		if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2605 				bc_ratio))
2606 			goto prop_msg_full;
2607 	nla_nest_end(msg->skb, prop);
2608 
2609 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2610 	if (err)
2611 		goto attr_msg_full;
2612 
2613 	tipc_bcast_unlock(net);
2614 	nla_nest_end(msg->skb, attrs);
2615 	genlmsg_end(msg->skb, hdr);
2616 
2617 	return 0;
2618 
2619 prop_msg_full:
2620 	nla_nest_cancel(msg->skb, prop);
2621 attr_msg_full:
2622 	nla_nest_cancel(msg->skb, attrs);
2623 msg_full:
2624 	tipc_bcast_unlock(net);
2625 	genlmsg_cancel(msg->skb, hdr);
2626 
2627 	return -EMSGSIZE;
2628 }
2629 
2630 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2631 			     struct sk_buff_head *xmitq)
2632 {
2633 	l->tolerance = tol;
2634 	if (l->bc_rcvlink)
2635 		l->bc_rcvlink->tolerance = tol;
2636 	if (link_is_up(l))
2637 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2638 }
2639 
2640 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2641 			struct sk_buff_head *xmitq)
2642 {
2643 	l->priority = prio;
2644 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2645 }
2646 
2647 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2648 {
2649 	l->abort_limit = limit;
2650 }
2651 
2652 char *tipc_link_name_ext(struct tipc_link *l, char *buf)
2653 {
2654 	if (!l)
2655 		scnprintf(buf, TIPC_MAX_LINK_NAME, "null");
2656 	else if (link_is_bc_sndlink(l))
2657 		scnprintf(buf, TIPC_MAX_LINK_NAME, "broadcast-sender");
2658 	else if (link_is_bc_rcvlink(l))
2659 		scnprintf(buf, TIPC_MAX_LINK_NAME,
2660 			  "broadcast-receiver, peer %x", l->addr);
2661 	else
2662 		memcpy(buf, l->name, TIPC_MAX_LINK_NAME);
2663 
2664 	return buf;
2665 }
2666 
2667 /**
2668  * tipc_link_dump - dump TIPC link data
2669  * @l: tipc link to be dumped
2670  * @dqueues: bitmask to decide if any link queue to be dumped?
2671  *           - TIPC_DUMP_NONE: don't dump link queues
2672  *           - TIPC_DUMP_TRANSMQ: dump link transmq queue
2673  *           - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2674  *           - TIPC_DUMP_DEFERDQ: dump link deferd queue
2675  *           - TIPC_DUMP_INPUTQ: dump link input queue
2676  *           - TIPC_DUMP_WAKEUP: dump link wakeup queue
2677  *           - TIPC_DUMP_ALL: dump all the link queues above
2678  * @buf: returned buffer of dump data in format
2679  */
2680 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2681 {
2682 	int i = 0;
2683 	size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2684 	struct sk_buff_head *list;
2685 	struct sk_buff *hskb, *tskb;
2686 	u32 len;
2687 
2688 	if (!l) {
2689 		i += scnprintf(buf, sz, "link data: (null)\n");
2690 		return i;
2691 	}
2692 
2693 	i += scnprintf(buf, sz, "link data: %x", l->addr);
2694 	i += scnprintf(buf + i, sz - i, " %x", l->state);
2695 	i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2696 	i += scnprintf(buf + i, sz - i, " %u", l->session);
2697 	i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2698 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2699 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2700 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2701 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2702 	i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2703 	i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2704 	i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2705 	i += scnprintf(buf + i, sz - i, " %u", 0);
2706 	i += scnprintf(buf + i, sz - i, " %u", 0);
2707 	i += scnprintf(buf + i, sz - i, " %u", l->acked);
2708 
2709 	list = &l->transmq;
2710 	len = skb_queue_len(list);
2711 	hskb = skb_peek(list);
2712 	tskb = skb_peek_tail(list);
2713 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2714 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2715 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2716 
2717 	list = &l->deferdq;
2718 	len = skb_queue_len(list);
2719 	hskb = skb_peek(list);
2720 	tskb = skb_peek_tail(list);
2721 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2722 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2723 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2724 
2725 	list = &l->backlogq;
2726 	len = skb_queue_len(list);
2727 	hskb = skb_peek(list);
2728 	tskb = skb_peek_tail(list);
2729 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2730 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2731 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2732 
2733 	list = l->inputq;
2734 	len = skb_queue_len(list);
2735 	hskb = skb_peek(list);
2736 	tskb = skb_peek_tail(list);
2737 	i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2738 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2739 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2740 
2741 	if (dqueues & TIPC_DUMP_TRANSMQ) {
2742 		i += scnprintf(buf + i, sz - i, "transmq: ");
2743 		i += tipc_list_dump(&l->transmq, false, buf + i);
2744 	}
2745 	if (dqueues & TIPC_DUMP_BACKLOGQ) {
2746 		i += scnprintf(buf + i, sz - i,
2747 			       "backlogq: <%u %u %u %u %u>, ",
2748 			       l->backlog[TIPC_LOW_IMPORTANCE].len,
2749 			       l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2750 			       l->backlog[TIPC_HIGH_IMPORTANCE].len,
2751 			       l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2752 			       l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2753 		i += tipc_list_dump(&l->backlogq, false, buf + i);
2754 	}
2755 	if (dqueues & TIPC_DUMP_DEFERDQ) {
2756 		i += scnprintf(buf + i, sz - i, "deferdq: ");
2757 		i += tipc_list_dump(&l->deferdq, false, buf + i);
2758 	}
2759 	if (dqueues & TIPC_DUMP_INPUTQ) {
2760 		i += scnprintf(buf + i, sz - i, "inputq: ");
2761 		i += tipc_list_dump(l->inputq, false, buf + i);
2762 	}
2763 	if (dqueues & TIPC_DUMP_WAKEUP) {
2764 		i += scnprintf(buf + i, sz - i, "wakeup: ");
2765 		i += tipc_list_dump(&l->wakeupq, false, buf + i);
2766 	}
2767 
2768 	return i;
2769 }
2770