xref: /openbmc/linux/net/tipc/link.c (revision 91db9311)
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_limit: time when repeated identical retransmits must force link reset
111  * @ackers: # of peers that needs to ack each packet before it can be released
112  * @acked: # last packet acked by a certain peer. Used for broadcast.
113  * @rcv_nxt: next sequence number to expect for inbound messages
114  * @deferred_queue: deferred queue saved OOS b'cast message received from node
115  * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
116  * @inputq: buffer queue for messages to be delivered upwards
117  * @namedq: buffer queue for name table messages to be delivered upwards
118  * @next_out: ptr to first unsent outbound message in queue
119  * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
120  * @long_msg_seq_no: next identifier to use for outbound fragmented messages
121  * @reasm_buf: head of partially reassembled inbound message fragments
122  * @bc_rcvr: marks that this is a broadcast receiver link
123  * @stats: collects statistics regarding link activity
124  */
125 struct tipc_link {
126 	u32 addr;
127 	char name[TIPC_MAX_LINK_NAME];
128 	struct net *net;
129 
130 	/* Management and link supervision data */
131 	u16 peer_session;
132 	u16 session;
133 	u16 snd_nxt_state;
134 	u16 rcv_nxt_state;
135 	u32 peer_bearer_id;
136 	u32 bearer_id;
137 	u32 tolerance;
138 	u32 abort_limit;
139 	u32 state;
140 	u16 peer_caps;
141 	bool in_session;
142 	bool active;
143 	u32 silent_intv_cnt;
144 	char if_name[TIPC_MAX_IF_NAME];
145 	u32 priority;
146 	char net_plane;
147 	struct tipc_mon_state mon_state;
148 	u16 rst_cnt;
149 
150 	/* Failover/synch */
151 	u16 drop_point;
152 	struct sk_buff *failover_reasm_skb;
153 	struct sk_buff_head failover_deferdq;
154 
155 	/* Max packet negotiation */
156 	u16 mtu;
157 	u16 advertised_mtu;
158 
159 	/* Sending */
160 	struct sk_buff_head transmq;
161 	struct sk_buff_head backlogq;
162 	struct {
163 		u16 len;
164 		u16 limit;
165 	} backlog[5];
166 	u16 snd_nxt;
167 	u16 prev_from;
168 	u16 window;
169 	unsigned long stale_limit;
170 
171 	/* Reception */
172 	u16 rcv_nxt;
173 	u32 rcv_unacked;
174 	struct sk_buff_head deferdq;
175 	struct sk_buff_head *inputq;
176 	struct sk_buff_head *namedq;
177 
178 	/* Congestion handling */
179 	struct sk_buff_head wakeupq;
180 
181 	/* Fragmentation/reassembly */
182 	struct sk_buff *reasm_buf;
183 
184 	/* Broadcast */
185 	u16 ackers;
186 	u16 acked;
187 	struct tipc_link *bc_rcvlink;
188 	struct tipc_link *bc_sndlink;
189 	u8 nack_state;
190 	bool bc_peer_is_up;
191 
192 	/* Statistics */
193 	struct tipc_stats stats;
194 };
195 
196 /*
197  * Error message prefixes
198  */
199 static const char *link_co_err = "Link tunneling error, ";
200 static const char *link_rst_msg = "Resetting link ";
201 
202 /* Send states for broadcast NACKs
203  */
204 enum {
205 	BC_NACK_SND_CONDITIONAL,
206 	BC_NACK_SND_UNCONDITIONAL,
207 	BC_NACK_SND_SUPPRESS,
208 };
209 
210 #define TIPC_BC_RETR_LIM  (jiffies + msecs_to_jiffies(10))
211 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
212 
213 /*
214  * Interval between NACKs when packets arrive out of order
215  */
216 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
217 
218 /* Link FSM states:
219  */
220 enum {
221 	LINK_ESTABLISHED     = 0xe,
222 	LINK_ESTABLISHING    = 0xe  << 4,
223 	LINK_RESET           = 0x1  << 8,
224 	LINK_RESETTING       = 0x2  << 12,
225 	LINK_PEER_RESET      = 0xd  << 16,
226 	LINK_FAILINGOVER     = 0xf  << 20,
227 	LINK_SYNCHING        = 0xc  << 24
228 };
229 
230 /* Link FSM state checking routines
231  */
232 static int link_is_up(struct tipc_link *l)
233 {
234 	return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
235 }
236 
237 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
238 			       struct sk_buff_head *xmitq);
239 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
240 				      bool probe_reply, u16 rcvgap,
241 				      int tolerance, int priority,
242 				      struct sk_buff_head *xmitq);
243 static void link_print(struct tipc_link *l, const char *str);
244 static int tipc_link_build_nack_msg(struct tipc_link *l,
245 				    struct sk_buff_head *xmitq);
246 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
247 					struct sk_buff_head *xmitq);
248 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
249 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data);
250 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
251 				     struct tipc_gap_ack_blks *ga,
252 				     struct sk_buff_head *xmitq);
253 
254 /*
255  *  Simple non-static link routines (i.e. referenced outside this file)
256  */
257 bool tipc_link_is_up(struct tipc_link *l)
258 {
259 	return link_is_up(l);
260 }
261 
262 bool tipc_link_peer_is_down(struct tipc_link *l)
263 {
264 	return l->state == LINK_PEER_RESET;
265 }
266 
267 bool tipc_link_is_reset(struct tipc_link *l)
268 {
269 	return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
270 }
271 
272 bool tipc_link_is_establishing(struct tipc_link *l)
273 {
274 	return l->state == LINK_ESTABLISHING;
275 }
276 
277 bool tipc_link_is_synching(struct tipc_link *l)
278 {
279 	return l->state == LINK_SYNCHING;
280 }
281 
282 bool tipc_link_is_failingover(struct tipc_link *l)
283 {
284 	return l->state == LINK_FAILINGOVER;
285 }
286 
287 bool tipc_link_is_blocked(struct tipc_link *l)
288 {
289 	return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
290 }
291 
292 static bool link_is_bc_sndlink(struct tipc_link *l)
293 {
294 	return !l->bc_sndlink;
295 }
296 
297 static bool link_is_bc_rcvlink(struct tipc_link *l)
298 {
299 	return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
300 }
301 
302 void tipc_link_set_active(struct tipc_link *l, bool active)
303 {
304 	l->active = active;
305 }
306 
307 u32 tipc_link_id(struct tipc_link *l)
308 {
309 	return l->peer_bearer_id << 16 | l->bearer_id;
310 }
311 
312 int tipc_link_window(struct tipc_link *l)
313 {
314 	return l->window;
315 }
316 
317 int tipc_link_prio(struct tipc_link *l)
318 {
319 	return l->priority;
320 }
321 
322 unsigned long tipc_link_tolerance(struct tipc_link *l)
323 {
324 	return l->tolerance;
325 }
326 
327 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
328 {
329 	return l->inputq;
330 }
331 
332 char tipc_link_plane(struct tipc_link *l)
333 {
334 	return l->net_plane;
335 }
336 
337 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
338 {
339 	l->peer_caps = capabilities;
340 }
341 
342 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
343 			   struct tipc_link *uc_l,
344 			   struct sk_buff_head *xmitq)
345 {
346 	struct tipc_link *rcv_l = uc_l->bc_rcvlink;
347 
348 	snd_l->ackers++;
349 	rcv_l->acked = snd_l->snd_nxt - 1;
350 	snd_l->state = LINK_ESTABLISHED;
351 	tipc_link_build_bc_init_msg(uc_l, xmitq);
352 }
353 
354 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
355 			      struct tipc_link *rcv_l,
356 			      struct sk_buff_head *xmitq)
357 {
358 	u16 ack = snd_l->snd_nxt - 1;
359 
360 	snd_l->ackers--;
361 	rcv_l->bc_peer_is_up = true;
362 	rcv_l->state = LINK_ESTABLISHED;
363 	tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
364 	trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
365 	tipc_link_reset(rcv_l);
366 	rcv_l->state = LINK_RESET;
367 	if (!snd_l->ackers) {
368 		trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
369 		tipc_link_reset(snd_l);
370 		snd_l->state = LINK_RESET;
371 		__skb_queue_purge(xmitq);
372 	}
373 }
374 
375 int tipc_link_bc_peers(struct tipc_link *l)
376 {
377 	return l->ackers;
378 }
379 
380 static u16 link_bc_rcv_gap(struct tipc_link *l)
381 {
382 	struct sk_buff *skb = skb_peek(&l->deferdq);
383 	u16 gap = 0;
384 
385 	if (more(l->snd_nxt, l->rcv_nxt))
386 		gap = l->snd_nxt - l->rcv_nxt;
387 	if (skb)
388 		gap = buf_seqno(skb) - l->rcv_nxt;
389 	return gap;
390 }
391 
392 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
393 {
394 	l->mtu = mtu;
395 }
396 
397 int tipc_link_mtu(struct tipc_link *l)
398 {
399 	return l->mtu;
400 }
401 
402 u16 tipc_link_rcv_nxt(struct tipc_link *l)
403 {
404 	return l->rcv_nxt;
405 }
406 
407 u16 tipc_link_acked(struct tipc_link *l)
408 {
409 	return l->acked;
410 }
411 
412 char *tipc_link_name(struct tipc_link *l)
413 {
414 	return l->name;
415 }
416 
417 u32 tipc_link_state(struct tipc_link *l)
418 {
419 	return l->state;
420 }
421 
422 /**
423  * tipc_link_create - create a new link
424  * @n: pointer to associated node
425  * @if_name: associated interface name
426  * @bearer_id: id (index) of associated bearer
427  * @tolerance: link tolerance to be used by link
428  * @net_plane: network plane (A,B,c..) this link belongs to
429  * @mtu: mtu to be advertised by link
430  * @priority: priority to be used by link
431  * @window: send window to be used by link
432  * @session: session to be used by link
433  * @ownnode: identity of own node
434  * @peer: node id of peer node
435  * @peer_caps: bitmap describing peer node capabilities
436  * @bc_sndlink: the namespace global link used for broadcast sending
437  * @bc_rcvlink: the peer specific link used for broadcast reception
438  * @inputq: queue to put messages ready for delivery
439  * @namedq: queue to put binding table update messages ready for delivery
440  * @link: return value, pointer to put the created link
441  *
442  * Returns true if link was created, otherwise false
443  */
444 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
445 		      int tolerance, char net_plane, u32 mtu, int priority,
446 		      int window, u32 session, u32 self,
447 		      u32 peer, u8 *peer_id, u16 peer_caps,
448 		      struct tipc_link *bc_sndlink,
449 		      struct tipc_link *bc_rcvlink,
450 		      struct sk_buff_head *inputq,
451 		      struct sk_buff_head *namedq,
452 		      struct tipc_link **link)
453 {
454 	char peer_str[NODE_ID_STR_LEN] = {0,};
455 	char self_str[NODE_ID_STR_LEN] = {0,};
456 	struct tipc_link *l;
457 
458 	l = kzalloc(sizeof(*l), GFP_ATOMIC);
459 	if (!l)
460 		return false;
461 	*link = l;
462 	l->session = session;
463 
464 	/* Set link name for unicast links only */
465 	if (peer_id) {
466 		tipc_nodeid2string(self_str, tipc_own_id(net));
467 		if (strlen(self_str) > 16)
468 			sprintf(self_str, "%x", self);
469 		tipc_nodeid2string(peer_str, peer_id);
470 		if (strlen(peer_str) > 16)
471 			sprintf(peer_str, "%x", peer);
472 	}
473 	/* Peer i/f name will be completed by reset/activate message */
474 	snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
475 		 self_str, if_name, peer_str);
476 
477 	strcpy(l->if_name, if_name);
478 	l->addr = peer;
479 	l->peer_caps = peer_caps;
480 	l->net = net;
481 	l->in_session = false;
482 	l->bearer_id = bearer_id;
483 	l->tolerance = tolerance;
484 	if (bc_rcvlink)
485 		bc_rcvlink->tolerance = tolerance;
486 	l->net_plane = net_plane;
487 	l->advertised_mtu = mtu;
488 	l->mtu = mtu;
489 	l->priority = priority;
490 	tipc_link_set_queue_limits(l, window);
491 	l->ackers = 1;
492 	l->bc_sndlink = bc_sndlink;
493 	l->bc_rcvlink = bc_rcvlink;
494 	l->inputq = inputq;
495 	l->namedq = namedq;
496 	l->state = LINK_RESETTING;
497 	__skb_queue_head_init(&l->transmq);
498 	__skb_queue_head_init(&l->backlogq);
499 	__skb_queue_head_init(&l->deferdq);
500 	__skb_queue_head_init(&l->failover_deferdq);
501 	skb_queue_head_init(&l->wakeupq);
502 	skb_queue_head_init(l->inputq);
503 	return true;
504 }
505 
506 /**
507  * tipc_link_bc_create - create new link to be used for broadcast
508  * @n: pointer to associated node
509  * @mtu: mtu to be used initially if no peers
510  * @window: send window to be used
511  * @inputq: queue to put messages ready for delivery
512  * @namedq: queue to put binding table update messages ready for delivery
513  * @link: return value, pointer to put the created link
514  *
515  * Returns true if link was created, otherwise false
516  */
517 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
518 			 int mtu, int window, u16 peer_caps,
519 			 struct sk_buff_head *inputq,
520 			 struct sk_buff_head *namedq,
521 			 struct tipc_link *bc_sndlink,
522 			 struct tipc_link **link)
523 {
524 	struct tipc_link *l;
525 
526 	if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
527 			      0, ownnode, peer, NULL, peer_caps, bc_sndlink,
528 			      NULL, inputq, namedq, link))
529 		return false;
530 
531 	l = *link;
532 	strcpy(l->name, tipc_bclink_name);
533 	trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
534 	tipc_link_reset(l);
535 	l->state = LINK_RESET;
536 	l->ackers = 0;
537 	l->bc_rcvlink = l;
538 
539 	/* Broadcast send link is always up */
540 	if (link_is_bc_sndlink(l))
541 		l->state = LINK_ESTABLISHED;
542 
543 	/* Disable replicast if even a single peer doesn't support it */
544 	if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
545 		tipc_bcast_disable_rcast(net);
546 
547 	return true;
548 }
549 
550 /**
551  * tipc_link_fsm_evt - link finite state machine
552  * @l: pointer to link
553  * @evt: state machine event to be processed
554  */
555 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
556 {
557 	int rc = 0;
558 	int old_state = l->state;
559 
560 	switch (l->state) {
561 	case LINK_RESETTING:
562 		switch (evt) {
563 		case LINK_PEER_RESET_EVT:
564 			l->state = LINK_PEER_RESET;
565 			break;
566 		case LINK_RESET_EVT:
567 			l->state = LINK_RESET;
568 			break;
569 		case LINK_FAILURE_EVT:
570 		case LINK_FAILOVER_BEGIN_EVT:
571 		case LINK_ESTABLISH_EVT:
572 		case LINK_FAILOVER_END_EVT:
573 		case LINK_SYNCH_BEGIN_EVT:
574 		case LINK_SYNCH_END_EVT:
575 		default:
576 			goto illegal_evt;
577 		}
578 		break;
579 	case LINK_RESET:
580 		switch (evt) {
581 		case LINK_PEER_RESET_EVT:
582 			l->state = LINK_ESTABLISHING;
583 			break;
584 		case LINK_FAILOVER_BEGIN_EVT:
585 			l->state = LINK_FAILINGOVER;
586 		case LINK_FAILURE_EVT:
587 		case LINK_RESET_EVT:
588 		case LINK_ESTABLISH_EVT:
589 		case LINK_FAILOVER_END_EVT:
590 			break;
591 		case LINK_SYNCH_BEGIN_EVT:
592 		case LINK_SYNCH_END_EVT:
593 		default:
594 			goto illegal_evt;
595 		}
596 		break;
597 	case LINK_PEER_RESET:
598 		switch (evt) {
599 		case LINK_RESET_EVT:
600 			l->state = LINK_ESTABLISHING;
601 			break;
602 		case LINK_PEER_RESET_EVT:
603 		case LINK_ESTABLISH_EVT:
604 		case LINK_FAILURE_EVT:
605 			break;
606 		case LINK_SYNCH_BEGIN_EVT:
607 		case LINK_SYNCH_END_EVT:
608 		case LINK_FAILOVER_BEGIN_EVT:
609 		case LINK_FAILOVER_END_EVT:
610 		default:
611 			goto illegal_evt;
612 		}
613 		break;
614 	case LINK_FAILINGOVER:
615 		switch (evt) {
616 		case LINK_FAILOVER_END_EVT:
617 			l->state = LINK_RESET;
618 			break;
619 		case LINK_PEER_RESET_EVT:
620 		case LINK_RESET_EVT:
621 		case LINK_ESTABLISH_EVT:
622 		case LINK_FAILURE_EVT:
623 			break;
624 		case LINK_FAILOVER_BEGIN_EVT:
625 		case LINK_SYNCH_BEGIN_EVT:
626 		case LINK_SYNCH_END_EVT:
627 		default:
628 			goto illegal_evt;
629 		}
630 		break;
631 	case LINK_ESTABLISHING:
632 		switch (evt) {
633 		case LINK_ESTABLISH_EVT:
634 			l->state = LINK_ESTABLISHED;
635 			break;
636 		case LINK_FAILOVER_BEGIN_EVT:
637 			l->state = LINK_FAILINGOVER;
638 			break;
639 		case LINK_RESET_EVT:
640 			l->state = LINK_RESET;
641 			break;
642 		case LINK_FAILURE_EVT:
643 		case LINK_PEER_RESET_EVT:
644 		case LINK_SYNCH_BEGIN_EVT:
645 		case LINK_FAILOVER_END_EVT:
646 			break;
647 		case LINK_SYNCH_END_EVT:
648 		default:
649 			goto illegal_evt;
650 		}
651 		break;
652 	case LINK_ESTABLISHED:
653 		switch (evt) {
654 		case LINK_PEER_RESET_EVT:
655 			l->state = LINK_PEER_RESET;
656 			rc |= TIPC_LINK_DOWN_EVT;
657 			break;
658 		case LINK_FAILURE_EVT:
659 			l->state = LINK_RESETTING;
660 			rc |= TIPC_LINK_DOWN_EVT;
661 			break;
662 		case LINK_RESET_EVT:
663 			l->state = LINK_RESET;
664 			break;
665 		case LINK_ESTABLISH_EVT:
666 		case LINK_SYNCH_END_EVT:
667 			break;
668 		case LINK_SYNCH_BEGIN_EVT:
669 			l->state = LINK_SYNCHING;
670 			break;
671 		case LINK_FAILOVER_BEGIN_EVT:
672 		case LINK_FAILOVER_END_EVT:
673 		default:
674 			goto illegal_evt;
675 		}
676 		break;
677 	case LINK_SYNCHING:
678 		switch (evt) {
679 		case LINK_PEER_RESET_EVT:
680 			l->state = LINK_PEER_RESET;
681 			rc |= TIPC_LINK_DOWN_EVT;
682 			break;
683 		case LINK_FAILURE_EVT:
684 			l->state = LINK_RESETTING;
685 			rc |= TIPC_LINK_DOWN_EVT;
686 			break;
687 		case LINK_RESET_EVT:
688 			l->state = LINK_RESET;
689 			break;
690 		case LINK_ESTABLISH_EVT:
691 		case LINK_SYNCH_BEGIN_EVT:
692 			break;
693 		case LINK_SYNCH_END_EVT:
694 			l->state = LINK_ESTABLISHED;
695 			break;
696 		case LINK_FAILOVER_BEGIN_EVT:
697 		case LINK_FAILOVER_END_EVT:
698 		default:
699 			goto illegal_evt;
700 		}
701 		break;
702 	default:
703 		pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
704 	}
705 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
706 	return rc;
707 illegal_evt:
708 	pr_err("Illegal FSM event %x in state %x on link %s\n",
709 	       evt, l->state, l->name);
710 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
711 	return rc;
712 }
713 
714 /* link_profile_stats - update statistical profiling of traffic
715  */
716 static void link_profile_stats(struct tipc_link *l)
717 {
718 	struct sk_buff *skb;
719 	struct tipc_msg *msg;
720 	int length;
721 
722 	/* Update counters used in statistical profiling of send traffic */
723 	l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
724 	l->stats.queue_sz_counts++;
725 
726 	skb = skb_peek(&l->transmq);
727 	if (!skb)
728 		return;
729 	msg = buf_msg(skb);
730 	length = msg_size(msg);
731 
732 	if (msg_user(msg) == MSG_FRAGMENTER) {
733 		if (msg_type(msg) != FIRST_FRAGMENT)
734 			return;
735 		length = msg_size(msg_inner_hdr(msg));
736 	}
737 	l->stats.msg_lengths_total += length;
738 	l->stats.msg_length_counts++;
739 	if (length <= 64)
740 		l->stats.msg_length_profile[0]++;
741 	else if (length <= 256)
742 		l->stats.msg_length_profile[1]++;
743 	else if (length <= 1024)
744 		l->stats.msg_length_profile[2]++;
745 	else if (length <= 4096)
746 		l->stats.msg_length_profile[3]++;
747 	else if (length <= 16384)
748 		l->stats.msg_length_profile[4]++;
749 	else if (length <= 32768)
750 		l->stats.msg_length_profile[5]++;
751 	else
752 		l->stats.msg_length_profile[6]++;
753 }
754 
755 /**
756  * tipc_link_too_silent - check if link is "too silent"
757  * @l: tipc link to be checked
758  *
759  * Returns true if the link 'silent_intv_cnt' is about to reach the
760  * 'abort_limit' value, otherwise false
761  */
762 bool tipc_link_too_silent(struct tipc_link *l)
763 {
764 	return (l->silent_intv_cnt + 2 > l->abort_limit);
765 }
766 
767 /* tipc_link_timeout - perform periodic task as instructed from node timeout
768  */
769 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
770 {
771 	int mtyp = 0;
772 	int rc = 0;
773 	bool state = false;
774 	bool probe = false;
775 	bool setup = false;
776 	u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
777 	u16 bc_acked = l->bc_rcvlink->acked;
778 	struct tipc_mon_state *mstate = &l->mon_state;
779 
780 	trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
781 	trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
782 	switch (l->state) {
783 	case LINK_ESTABLISHED:
784 	case LINK_SYNCHING:
785 		mtyp = STATE_MSG;
786 		link_profile_stats(l);
787 		tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
788 		if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
789 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
790 		state = bc_acked != bc_snt;
791 		state |= l->bc_rcvlink->rcv_unacked;
792 		state |= l->rcv_unacked;
793 		state |= !skb_queue_empty(&l->transmq);
794 		state |= !skb_queue_empty(&l->deferdq);
795 		probe = mstate->probing;
796 		probe |= l->silent_intv_cnt;
797 		if (probe || mstate->monitoring)
798 			l->silent_intv_cnt++;
799 		break;
800 	case LINK_RESET:
801 		setup = l->rst_cnt++ <= 4;
802 		setup |= !(l->rst_cnt % 16);
803 		mtyp = RESET_MSG;
804 		break;
805 	case LINK_ESTABLISHING:
806 		setup = true;
807 		mtyp = ACTIVATE_MSG;
808 		break;
809 	case LINK_PEER_RESET:
810 	case LINK_RESETTING:
811 	case LINK_FAILINGOVER:
812 		break;
813 	default:
814 		break;
815 	}
816 
817 	if (state || probe || setup)
818 		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
819 
820 	return rc;
821 }
822 
823 /**
824  * link_schedule_user - schedule a message sender for wakeup after congestion
825  * @l: congested link
826  * @hdr: header of message that is being sent
827  * Create pseudo msg to send back to user when congestion abates
828  */
829 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
830 {
831 	u32 dnode = tipc_own_addr(l->net);
832 	u32 dport = msg_origport(hdr);
833 	struct sk_buff *skb;
834 
835 	/* Create and schedule wakeup pseudo message */
836 	skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
837 			      dnode, l->addr, dport, 0, 0);
838 	if (!skb)
839 		return -ENOBUFS;
840 	msg_set_dest_droppable(buf_msg(skb), true);
841 	TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
842 	skb_queue_tail(&l->wakeupq, skb);
843 	l->stats.link_congs++;
844 	trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
845 	return -ELINKCONG;
846 }
847 
848 /**
849  * link_prepare_wakeup - prepare users for wakeup after congestion
850  * @l: congested link
851  * Wake up a number of waiting users, as permitted by available space
852  * in the send queue
853  */
854 static void link_prepare_wakeup(struct tipc_link *l)
855 {
856 	struct sk_buff *skb, *tmp;
857 	int imp, i = 0;
858 
859 	skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
860 		imp = TIPC_SKB_CB(skb)->chain_imp;
861 		if (l->backlog[imp].len < l->backlog[imp].limit) {
862 			skb_unlink(skb, &l->wakeupq);
863 			skb_queue_tail(l->inputq, skb);
864 		} else if (i++ > 10) {
865 			break;
866 		}
867 	}
868 }
869 
870 void tipc_link_reset(struct tipc_link *l)
871 {
872 	struct sk_buff_head list;
873 
874 	__skb_queue_head_init(&list);
875 
876 	l->in_session = false;
877 	/* Force re-synch of peer session number before establishing */
878 	l->peer_session--;
879 	l->session++;
880 	l->mtu = l->advertised_mtu;
881 
882 	spin_lock_bh(&l->wakeupq.lock);
883 	skb_queue_splice_init(&l->wakeupq, &list);
884 	spin_unlock_bh(&l->wakeupq.lock);
885 
886 	spin_lock_bh(&l->inputq->lock);
887 	skb_queue_splice_init(&list, l->inputq);
888 	spin_unlock_bh(&l->inputq->lock);
889 
890 	__skb_queue_purge(&l->transmq);
891 	__skb_queue_purge(&l->deferdq);
892 	__skb_queue_purge(&l->backlogq);
893 	__skb_queue_purge(&l->failover_deferdq);
894 	l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
895 	l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
896 	l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
897 	l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
898 	l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
899 	kfree_skb(l->reasm_buf);
900 	kfree_skb(l->failover_reasm_skb);
901 	l->reasm_buf = NULL;
902 	l->failover_reasm_skb = NULL;
903 	l->rcv_unacked = 0;
904 	l->snd_nxt = 1;
905 	l->rcv_nxt = 1;
906 	l->snd_nxt_state = 1;
907 	l->rcv_nxt_state = 1;
908 	l->acked = 0;
909 	l->silent_intv_cnt = 0;
910 	l->rst_cnt = 0;
911 	l->bc_peer_is_up = false;
912 	memset(&l->mon_state, 0, sizeof(l->mon_state));
913 	tipc_link_reset_stats(l);
914 }
915 
916 /**
917  * tipc_link_xmit(): enqueue buffer list according to queue situation
918  * @link: link to use
919  * @list: chain of buffers containing message
920  * @xmitq: returned list of packets to be sent by caller
921  *
922  * Consumes the buffer chain.
923  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
924  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
925  */
926 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
927 		   struct sk_buff_head *xmitq)
928 {
929 	struct tipc_msg *hdr = buf_msg(skb_peek(list));
930 	unsigned int maxwin = l->window;
931 	int imp = msg_importance(hdr);
932 	unsigned int mtu = l->mtu;
933 	u16 ack = l->rcv_nxt - 1;
934 	u16 seqno = l->snd_nxt;
935 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
936 	struct sk_buff_head *transmq = &l->transmq;
937 	struct sk_buff_head *backlogq = &l->backlogq;
938 	struct sk_buff *skb, *_skb, *bskb;
939 	int pkt_cnt = skb_queue_len(list);
940 	int rc = 0;
941 
942 	if (unlikely(msg_size(hdr) > mtu)) {
943 		skb_queue_purge(list);
944 		return -EMSGSIZE;
945 	}
946 
947 	/* Allow oversubscription of one data msg per source at congestion */
948 	if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
949 		if (imp == TIPC_SYSTEM_IMPORTANCE) {
950 			pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
951 			return -ENOBUFS;
952 		}
953 		rc = link_schedule_user(l, hdr);
954 	}
955 
956 	if (pkt_cnt > 1) {
957 		l->stats.sent_fragmented++;
958 		l->stats.sent_fragments += pkt_cnt;
959 	}
960 
961 	/* Prepare each packet for sending, and add to relevant queue: */
962 	while (skb_queue_len(list)) {
963 		skb = skb_peek(list);
964 		hdr = buf_msg(skb);
965 		msg_set_seqno(hdr, seqno);
966 		msg_set_ack(hdr, ack);
967 		msg_set_bcast_ack(hdr, bc_ack);
968 
969 		if (likely(skb_queue_len(transmq) < maxwin)) {
970 			_skb = skb_clone(skb, GFP_ATOMIC);
971 			if (!_skb) {
972 				skb_queue_purge(list);
973 				return -ENOBUFS;
974 			}
975 			__skb_dequeue(list);
976 			__skb_queue_tail(transmq, skb);
977 			/* next retransmit attempt */
978 			if (link_is_bc_sndlink(l))
979 				TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
980 			__skb_queue_tail(xmitq, _skb);
981 			TIPC_SKB_CB(skb)->ackers = l->ackers;
982 			l->rcv_unacked = 0;
983 			l->stats.sent_pkts++;
984 			seqno++;
985 			continue;
986 		}
987 		if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
988 			kfree_skb(__skb_dequeue(list));
989 			l->stats.sent_bundled++;
990 			continue;
991 		}
992 		if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
993 			kfree_skb(__skb_dequeue(list));
994 			__skb_queue_tail(backlogq, bskb);
995 			l->backlog[msg_importance(buf_msg(bskb))].len++;
996 			l->stats.sent_bundled++;
997 			l->stats.sent_bundles++;
998 			continue;
999 		}
1000 		l->backlog[imp].len += skb_queue_len(list);
1001 		skb_queue_splice_tail_init(list, backlogq);
1002 	}
1003 	l->snd_nxt = seqno;
1004 	return rc;
1005 }
1006 
1007 static void tipc_link_advance_backlog(struct tipc_link *l,
1008 				      struct sk_buff_head *xmitq)
1009 {
1010 	struct sk_buff *skb, *_skb;
1011 	struct tipc_msg *hdr;
1012 	u16 seqno = l->snd_nxt;
1013 	u16 ack = l->rcv_nxt - 1;
1014 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1015 
1016 	while (skb_queue_len(&l->transmq) < l->window) {
1017 		skb = skb_peek(&l->backlogq);
1018 		if (!skb)
1019 			break;
1020 		_skb = skb_clone(skb, GFP_ATOMIC);
1021 		if (!_skb)
1022 			break;
1023 		__skb_dequeue(&l->backlogq);
1024 		hdr = buf_msg(skb);
1025 		l->backlog[msg_importance(hdr)].len--;
1026 		__skb_queue_tail(&l->transmq, skb);
1027 		/* next retransmit attempt */
1028 		if (link_is_bc_sndlink(l))
1029 			TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1030 
1031 		__skb_queue_tail(xmitq, _skb);
1032 		TIPC_SKB_CB(skb)->ackers = l->ackers;
1033 		msg_set_seqno(hdr, seqno);
1034 		msg_set_ack(hdr, ack);
1035 		msg_set_bcast_ack(hdr, bc_ack);
1036 		l->rcv_unacked = 0;
1037 		l->stats.sent_pkts++;
1038 		seqno++;
1039 	}
1040 	l->snd_nxt = seqno;
1041 }
1042 
1043 /**
1044  * link_retransmit_failure() - Detect repeated retransmit failures
1045  * @l: tipc link sender
1046  * @r: tipc link receiver (= l in case of unicast)
1047  * @from: seqno of the 1st packet in retransmit request
1048  * @rc: returned code
1049  *
1050  * Return: true if the repeated retransmit failures happens, otherwise
1051  * false
1052  */
1053 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1054 				    u16 from, int *rc)
1055 {
1056 	struct sk_buff *skb = skb_peek(&l->transmq);
1057 	struct tipc_msg *hdr;
1058 
1059 	if (!skb)
1060 		return false;
1061 	hdr = buf_msg(skb);
1062 
1063 	/* Detect repeated retransmit failures on same packet */
1064 	if (r->prev_from != from) {
1065 		r->prev_from = from;
1066 		r->stale_limit = jiffies + msecs_to_jiffies(r->tolerance);
1067 	} else if (time_after(jiffies, r->stale_limit)) {
1068 		pr_warn("Retransmission failure on link <%s>\n", l->name);
1069 		link_print(l, "State of link ");
1070 		pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1071 			msg_user(hdr), msg_type(hdr), msg_size(hdr),
1072 			msg_errcode(hdr));
1073 		pr_info("sqno %u, prev: %x, src: %x\n",
1074 			msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
1075 
1076 		trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1077 		trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1078 		trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1079 
1080 		if (link_is_bc_sndlink(l))
1081 			*rc = TIPC_LINK_DOWN_EVT;
1082 
1083 		*rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1084 		return true;
1085 	}
1086 
1087 	return false;
1088 }
1089 
1090 /* tipc_link_bc_retrans() - retransmit zero or more packets
1091  * @l: the link to transmit on
1092  * @r: the receiving link ordering the retransmit. Same as l if unicast
1093  * @from: retransmit from (inclusive) this sequence number
1094  * @to: retransmit to (inclusive) this sequence number
1095  * xmitq: queue for accumulating the retransmitted packets
1096  */
1097 static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r,
1098 				u16 from, u16 to, struct sk_buff_head *xmitq)
1099 {
1100 	struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1101 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1102 	u16 ack = l->rcv_nxt - 1;
1103 	struct tipc_msg *hdr;
1104 	int rc = 0;
1105 
1106 	if (!skb)
1107 		return 0;
1108 	if (less(to, from))
1109 		return 0;
1110 
1111 	trace_tipc_link_retrans(r, from, to, &l->transmq);
1112 
1113 	if (link_retransmit_failure(l, r, from, &rc))
1114 		return rc;
1115 
1116 	skb_queue_walk(&l->transmq, skb) {
1117 		hdr = buf_msg(skb);
1118 		if (less(msg_seqno(hdr), from))
1119 			continue;
1120 		if (more(msg_seqno(hdr), to))
1121 			break;
1122 		if (link_is_bc_sndlink(l)) {
1123 			if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1124 				continue;
1125 			TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1126 		}
1127 		_skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE, GFP_ATOMIC);
1128 		if (!_skb)
1129 			return 0;
1130 		hdr = buf_msg(_skb);
1131 		msg_set_ack(hdr, ack);
1132 		msg_set_bcast_ack(hdr, bc_ack);
1133 		_skb->priority = TC_PRIO_CONTROL;
1134 		__skb_queue_tail(xmitq, _skb);
1135 		l->stats.retransmitted++;
1136 	}
1137 	return 0;
1138 }
1139 
1140 /* tipc_data_input - deliver data and name distr msgs to upper layer
1141  *
1142  * Consumes buffer if message is of right type
1143  * Node lock must be held
1144  */
1145 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1146 			    struct sk_buff_head *inputq)
1147 {
1148 	struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1149 	struct tipc_msg *hdr = buf_msg(skb);
1150 
1151 	switch (msg_user(hdr)) {
1152 	case TIPC_LOW_IMPORTANCE:
1153 	case TIPC_MEDIUM_IMPORTANCE:
1154 	case TIPC_HIGH_IMPORTANCE:
1155 	case TIPC_CRITICAL_IMPORTANCE:
1156 		if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1157 			skb_queue_tail(mc_inputq, skb);
1158 			return true;
1159 		}
1160 		/* fall through */
1161 	case CONN_MANAGER:
1162 		skb_queue_tail(inputq, skb);
1163 		return true;
1164 	case GROUP_PROTOCOL:
1165 		skb_queue_tail(mc_inputq, skb);
1166 		return true;
1167 	case NAME_DISTRIBUTOR:
1168 		l->bc_rcvlink->state = LINK_ESTABLISHED;
1169 		skb_queue_tail(l->namedq, skb);
1170 		return true;
1171 	case MSG_BUNDLER:
1172 	case TUNNEL_PROTOCOL:
1173 	case MSG_FRAGMENTER:
1174 	case BCAST_PROTOCOL:
1175 		return false;
1176 	default:
1177 		pr_warn("Dropping received illegal msg type\n");
1178 		kfree_skb(skb);
1179 		return true;
1180 	};
1181 }
1182 
1183 /* tipc_link_input - process packet that has passed link protocol check
1184  *
1185  * Consumes buffer
1186  */
1187 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1188 			   struct sk_buff_head *inputq,
1189 			   struct sk_buff **reasm_skb)
1190 {
1191 	struct tipc_msg *hdr = buf_msg(skb);
1192 	struct sk_buff *iskb;
1193 	struct sk_buff_head tmpq;
1194 	int usr = msg_user(hdr);
1195 	int pos = 0;
1196 
1197 	if (usr == MSG_BUNDLER) {
1198 		skb_queue_head_init(&tmpq);
1199 		l->stats.recv_bundles++;
1200 		l->stats.recv_bundled += msg_msgcnt(hdr);
1201 		while (tipc_msg_extract(skb, &iskb, &pos))
1202 			tipc_data_input(l, iskb, &tmpq);
1203 		tipc_skb_queue_splice_tail(&tmpq, inputq);
1204 		return 0;
1205 	} else if (usr == MSG_FRAGMENTER) {
1206 		l->stats.recv_fragments++;
1207 		if (tipc_buf_append(reasm_skb, &skb)) {
1208 			l->stats.recv_fragmented++;
1209 			tipc_data_input(l, skb, inputq);
1210 		} else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1211 			pr_warn_ratelimited("Unable to build fragment list\n");
1212 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1213 		}
1214 		return 0;
1215 	} else if (usr == BCAST_PROTOCOL) {
1216 		tipc_bcast_lock(l->net);
1217 		tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1218 		tipc_bcast_unlock(l->net);
1219 	}
1220 
1221 	kfree_skb(skb);
1222 	return 0;
1223 }
1224 
1225 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1226  *			 inner message along with the ones in the old link's
1227  *			 deferdq
1228  * @l: tunnel link
1229  * @skb: TUNNEL_PROTOCOL message
1230  * @inputq: queue to put messages ready for delivery
1231  */
1232 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1233 			     struct sk_buff_head *inputq)
1234 {
1235 	struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1236 	struct sk_buff_head *fdefq = &l->failover_deferdq;
1237 	struct tipc_msg *hdr = buf_msg(skb);
1238 	struct sk_buff *iskb;
1239 	int ipos = 0;
1240 	int rc = 0;
1241 	u16 seqno;
1242 
1243 	/* SYNCH_MSG */
1244 	if (msg_type(hdr) == SYNCH_MSG)
1245 		goto drop;
1246 
1247 	/* FAILOVER_MSG */
1248 	if (!tipc_msg_extract(skb, &iskb, &ipos)) {
1249 		pr_warn_ratelimited("Cannot extract FAILOVER_MSG, defq: %d\n",
1250 				    skb_queue_len(fdefq));
1251 		return rc;
1252 	}
1253 
1254 	do {
1255 		seqno = buf_seqno(iskb);
1256 
1257 		if (unlikely(less(seqno, l->drop_point))) {
1258 			kfree_skb(iskb);
1259 			continue;
1260 		}
1261 
1262 		if (unlikely(seqno != l->drop_point)) {
1263 			__tipc_skb_queue_sorted(fdefq, seqno, iskb);
1264 			continue;
1265 		}
1266 
1267 		l->drop_point++;
1268 
1269 		if (!tipc_data_input(l, iskb, inputq))
1270 			rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1271 		if (unlikely(rc))
1272 			break;
1273 	} while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1274 
1275 drop:
1276 	kfree_skb(skb);
1277 	return rc;
1278 }
1279 
1280 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1281 {
1282 	bool released = false;
1283 	struct sk_buff *skb, *tmp;
1284 
1285 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1286 		if (more(buf_seqno(skb), acked))
1287 			break;
1288 		__skb_unlink(skb, &l->transmq);
1289 		kfree_skb(skb);
1290 		released = true;
1291 	}
1292 	return released;
1293 }
1294 
1295 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1296  * @l: tipc link that data have come with gaps in sequence if any
1297  * @data: data buffer to store the Gap ACK blocks after built
1298  *
1299  * returns the actual allocated memory size
1300  */
1301 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data)
1302 {
1303 	struct sk_buff *skb = skb_peek(&l->deferdq);
1304 	struct tipc_gap_ack_blks *ga = data;
1305 	u16 len, expect, seqno = 0;
1306 	u8 n = 0;
1307 
1308 	if (!skb)
1309 		goto exit;
1310 
1311 	expect = buf_seqno(skb);
1312 	skb_queue_walk(&l->deferdq, skb) {
1313 		seqno = buf_seqno(skb);
1314 		if (unlikely(more(seqno, expect))) {
1315 			ga->gacks[n].ack = htons(expect - 1);
1316 			ga->gacks[n].gap = htons(seqno - expect);
1317 			if (++n >= MAX_GAP_ACK_BLKS) {
1318 				pr_info_ratelimited("Too few Gap ACK blocks!\n");
1319 				goto exit;
1320 			}
1321 		} else if (unlikely(less(seqno, expect))) {
1322 			pr_warn("Unexpected skb in deferdq!\n");
1323 			continue;
1324 		}
1325 		expect = seqno + 1;
1326 	}
1327 
1328 	/* last block */
1329 	ga->gacks[n].ack = htons(seqno);
1330 	ga->gacks[n].gap = 0;
1331 	n++;
1332 
1333 exit:
1334 	len = tipc_gap_ack_blks_sz(n);
1335 	ga->len = htons(len);
1336 	ga->gack_cnt = n;
1337 	return len;
1338 }
1339 
1340 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1341  *			       acked packets, also doing retransmissions if
1342  *			       gaps found
1343  * @l: tipc link with transmq queue to be advanced
1344  * @acked: seqno of last packet acked by peer without any gaps before
1345  * @gap: # of gap packets
1346  * @ga: buffer pointer to Gap ACK blocks from peer
1347  * @xmitq: queue for accumulating the retransmitted packets if any
1348  *
1349  * In case of a repeated retransmit failures, the call will return shortly
1350  * with a returned code (e.g. TIPC_LINK_DOWN_EVT)
1351  */
1352 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
1353 				     struct tipc_gap_ack_blks *ga,
1354 				     struct sk_buff_head *xmitq)
1355 {
1356 	struct sk_buff *skb, *_skb, *tmp;
1357 	struct tipc_msg *hdr;
1358 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1359 	u16 ack = l->rcv_nxt - 1;
1360 	u16 seqno, n = 0;
1361 	int rc = 0;
1362 
1363 	if (gap && link_retransmit_failure(l, l, acked + 1, &rc))
1364 		return rc;
1365 
1366 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1367 		seqno = buf_seqno(skb);
1368 
1369 next_gap_ack:
1370 		if (less_eq(seqno, acked)) {
1371 			/* release skb */
1372 			__skb_unlink(skb, &l->transmq);
1373 			kfree_skb(skb);
1374 		} else if (less_eq(seqno, acked + gap)) {
1375 			/* retransmit skb */
1376 			if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1377 				continue;
1378 			TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
1379 
1380 			_skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1381 			if (!_skb)
1382 				continue;
1383 			hdr = buf_msg(_skb);
1384 			msg_set_ack(hdr, ack);
1385 			msg_set_bcast_ack(hdr, bc_ack);
1386 			_skb->priority = TC_PRIO_CONTROL;
1387 			__skb_queue_tail(xmitq, _skb);
1388 			l->stats.retransmitted++;
1389 		} else {
1390 			/* retry with Gap ACK blocks if any */
1391 			if (!ga || n >= ga->gack_cnt)
1392 				break;
1393 			acked = ntohs(ga->gacks[n].ack);
1394 			gap = ntohs(ga->gacks[n].gap);
1395 			n++;
1396 			goto next_gap_ack;
1397 		}
1398 	}
1399 
1400 	return 0;
1401 }
1402 
1403 /* tipc_link_build_state_msg: prepare link state message for transmission
1404  *
1405  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1406  * risk of ack storms towards the sender
1407  */
1408 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1409 {
1410 	if (!l)
1411 		return 0;
1412 
1413 	/* Broadcast ACK must be sent via a unicast link => defer to caller */
1414 	if (link_is_bc_rcvlink(l)) {
1415 		if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1416 			return 0;
1417 		l->rcv_unacked = 0;
1418 
1419 		/* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1420 		l->snd_nxt = l->rcv_nxt;
1421 		return TIPC_LINK_SND_STATE;
1422 	}
1423 
1424 	/* Unicast ACK */
1425 	l->rcv_unacked = 0;
1426 	l->stats.sent_acks++;
1427 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1428 	return 0;
1429 }
1430 
1431 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1432  */
1433 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1434 {
1435 	int mtyp = RESET_MSG;
1436 	struct sk_buff *skb;
1437 
1438 	if (l->state == LINK_ESTABLISHING)
1439 		mtyp = ACTIVATE_MSG;
1440 
1441 	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1442 
1443 	/* Inform peer that this endpoint is going down if applicable */
1444 	skb = skb_peek_tail(xmitq);
1445 	if (skb && (l->state == LINK_RESET))
1446 		msg_set_peer_stopping(buf_msg(skb), 1);
1447 }
1448 
1449 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1450  * Note that sending of broadcast NACK is coordinated among nodes, to
1451  * reduce the risk of NACK storms towards the sender
1452  */
1453 static int tipc_link_build_nack_msg(struct tipc_link *l,
1454 				    struct sk_buff_head *xmitq)
1455 {
1456 	u32 def_cnt = ++l->stats.deferred_recv;
1457 	u32 defq_len = skb_queue_len(&l->deferdq);
1458 	int match1, match2;
1459 
1460 	if (link_is_bc_rcvlink(l)) {
1461 		match1 = def_cnt & 0xf;
1462 		match2 = tipc_own_addr(l->net) & 0xf;
1463 		if (match1 == match2)
1464 			return TIPC_LINK_SND_STATE;
1465 		return 0;
1466 	}
1467 
1468 	if (defq_len >= 3 && !((defq_len - 3) % 16))
1469 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1470 	return 0;
1471 }
1472 
1473 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1474  * @l: the link that should handle the message
1475  * @skb: TIPC packet
1476  * @xmitq: queue to place packets to be sent after this call
1477  */
1478 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1479 		  struct sk_buff_head *xmitq)
1480 {
1481 	struct sk_buff_head *defq = &l->deferdq;
1482 	struct tipc_msg *hdr = buf_msg(skb);
1483 	u16 seqno, rcv_nxt, win_lim;
1484 	int rc = 0;
1485 
1486 	/* Verify and update link state */
1487 	if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1488 		return tipc_link_proto_rcv(l, skb, xmitq);
1489 
1490 	/* Don't send probe at next timeout expiration */
1491 	l->silent_intv_cnt = 0;
1492 
1493 	do {
1494 		hdr = buf_msg(skb);
1495 		seqno = msg_seqno(hdr);
1496 		rcv_nxt = l->rcv_nxt;
1497 		win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1498 
1499 		if (unlikely(!link_is_up(l))) {
1500 			if (l->state == LINK_ESTABLISHING)
1501 				rc = TIPC_LINK_UP_EVT;
1502 			goto drop;
1503 		}
1504 
1505 		/* Drop if outside receive window */
1506 		if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1507 			l->stats.duplicates++;
1508 			goto drop;
1509 		}
1510 
1511 		/* Forward queues and wake up waiting users */
1512 		if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1513 			tipc_link_advance_backlog(l, xmitq);
1514 			if (unlikely(!skb_queue_empty(&l->wakeupq)))
1515 				link_prepare_wakeup(l);
1516 		}
1517 
1518 		/* Defer delivery if sequence gap */
1519 		if (unlikely(seqno != rcv_nxt)) {
1520 			__tipc_skb_queue_sorted(defq, seqno, skb);
1521 			rc |= tipc_link_build_nack_msg(l, xmitq);
1522 			break;
1523 		}
1524 
1525 		/* Deliver packet */
1526 		l->rcv_nxt++;
1527 		l->stats.recv_pkts++;
1528 
1529 		if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1530 			rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1531 		else if (!tipc_data_input(l, skb, l->inputq))
1532 			rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1533 		if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1534 			rc |= tipc_link_build_state_msg(l, xmitq);
1535 		if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1536 			break;
1537 	} while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1538 
1539 	return rc;
1540 drop:
1541 	kfree_skb(skb);
1542 	return rc;
1543 }
1544 
1545 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1546 				      bool probe_reply, u16 rcvgap,
1547 				      int tolerance, int priority,
1548 				      struct sk_buff_head *xmitq)
1549 {
1550 	struct tipc_link *bcl = l->bc_rcvlink;
1551 	struct sk_buff *skb;
1552 	struct tipc_msg *hdr;
1553 	struct sk_buff_head *dfq = &l->deferdq;
1554 	bool node_up = link_is_up(bcl);
1555 	struct tipc_mon_state *mstate = &l->mon_state;
1556 	int dlen = 0;
1557 	void *data;
1558 	u16 glen = 0;
1559 
1560 	/* Don't send protocol message during reset or link failover */
1561 	if (tipc_link_is_blocked(l))
1562 		return;
1563 
1564 	if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1565 		return;
1566 
1567 	if (!skb_queue_empty(dfq))
1568 		rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1569 
1570 	skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1571 			      tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1572 			      l->addr, tipc_own_addr(l->net), 0, 0, 0);
1573 	if (!skb)
1574 		return;
1575 
1576 	hdr = buf_msg(skb);
1577 	data = msg_data(hdr);
1578 	msg_set_session(hdr, l->session);
1579 	msg_set_bearer_id(hdr, l->bearer_id);
1580 	msg_set_net_plane(hdr, l->net_plane);
1581 	msg_set_next_sent(hdr, l->snd_nxt);
1582 	msg_set_ack(hdr, l->rcv_nxt - 1);
1583 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1584 	msg_set_bc_ack_invalid(hdr, !node_up);
1585 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1586 	msg_set_link_tolerance(hdr, tolerance);
1587 	msg_set_linkprio(hdr, priority);
1588 	msg_set_redundant_link(hdr, node_up);
1589 	msg_set_seq_gap(hdr, 0);
1590 	msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1591 
1592 	if (mtyp == STATE_MSG) {
1593 		if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1594 			msg_set_seqno(hdr, l->snd_nxt_state++);
1595 		msg_set_seq_gap(hdr, rcvgap);
1596 		msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1597 		msg_set_probe(hdr, probe);
1598 		msg_set_is_keepalive(hdr, probe || probe_reply);
1599 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1600 			glen = tipc_build_gap_ack_blks(l, data);
1601 		tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1602 		msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1603 		skb_trim(skb, INT_H_SIZE + glen + dlen);
1604 		l->stats.sent_states++;
1605 		l->rcv_unacked = 0;
1606 	} else {
1607 		/* RESET_MSG or ACTIVATE_MSG */
1608 		if (mtyp == ACTIVATE_MSG) {
1609 			msg_set_dest_session_valid(hdr, 1);
1610 			msg_set_dest_session(hdr, l->peer_session);
1611 		}
1612 		msg_set_max_pkt(hdr, l->advertised_mtu);
1613 		strcpy(data, l->if_name);
1614 		msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1615 		skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1616 	}
1617 	if (probe)
1618 		l->stats.sent_probes++;
1619 	if (rcvgap)
1620 		l->stats.sent_nacks++;
1621 	skb->priority = TC_PRIO_CONTROL;
1622 	__skb_queue_tail(xmitq, skb);
1623 	trace_tipc_proto_build(skb, false, l->name);
1624 }
1625 
1626 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1627 				    struct sk_buff_head *xmitq)
1628 {
1629 	u32 onode = tipc_own_addr(l->net);
1630 	struct tipc_msg *hdr, *ihdr;
1631 	struct sk_buff_head tnlq;
1632 	struct sk_buff *skb;
1633 	u32 dnode = l->addr;
1634 
1635 	skb_queue_head_init(&tnlq);
1636 	skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1637 			      INT_H_SIZE, BASIC_H_SIZE,
1638 			      dnode, onode, 0, 0, 0);
1639 	if (!skb) {
1640 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1641 		return;
1642 	}
1643 
1644 	hdr = buf_msg(skb);
1645 	msg_set_msgcnt(hdr, 1);
1646 	msg_set_bearer_id(hdr, l->peer_bearer_id);
1647 
1648 	ihdr = (struct tipc_msg *)msg_data(hdr);
1649 	tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1650 		      BASIC_H_SIZE, dnode);
1651 	msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1652 	__skb_queue_tail(&tnlq, skb);
1653 	tipc_link_xmit(l, &tnlq, xmitq);
1654 }
1655 
1656 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1657  * with contents of the link's transmit and backlog queues.
1658  */
1659 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1660 			   int mtyp, struct sk_buff_head *xmitq)
1661 {
1662 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1663 	struct sk_buff *skb, *tnlskb;
1664 	struct tipc_msg *hdr, tnlhdr;
1665 	struct sk_buff_head *queue = &l->transmq;
1666 	struct sk_buff_head tmpxq, tnlq;
1667 	u16 pktlen, pktcnt, seqno = l->snd_nxt;
1668 
1669 	if (!tnl)
1670 		return;
1671 
1672 	skb_queue_head_init(&tnlq);
1673 	skb_queue_head_init(&tmpxq);
1674 
1675 	/* At least one packet required for safe algorithm => add dummy */
1676 	skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1677 			      BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1678 			      0, 0, TIPC_ERR_NO_PORT);
1679 	if (!skb) {
1680 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1681 		return;
1682 	}
1683 	skb_queue_tail(&tnlq, skb);
1684 	tipc_link_xmit(l, &tnlq, &tmpxq);
1685 	__skb_queue_purge(&tmpxq);
1686 
1687 	/* Initialize reusable tunnel packet header */
1688 	tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1689 		      mtyp, INT_H_SIZE, l->addr);
1690 	if (mtyp == SYNCH_MSG)
1691 		pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
1692 	else
1693 		pktcnt = skb_queue_len(&l->transmq);
1694 	pktcnt += skb_queue_len(&l->backlogq);
1695 	msg_set_msgcnt(&tnlhdr, pktcnt);
1696 	msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1697 tnl:
1698 	/* Wrap each packet into a tunnel packet */
1699 	skb_queue_walk(queue, skb) {
1700 		hdr = buf_msg(skb);
1701 		if (queue == &l->backlogq)
1702 			msg_set_seqno(hdr, seqno++);
1703 		pktlen = msg_size(hdr);
1704 		msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1705 		tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1706 		if (!tnlskb) {
1707 			pr_warn("%sunable to send packet\n", link_co_err);
1708 			return;
1709 		}
1710 		skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1711 		skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1712 		__skb_queue_tail(&tnlq, tnlskb);
1713 	}
1714 	if (queue != &l->backlogq) {
1715 		queue = &l->backlogq;
1716 		goto tnl;
1717 	}
1718 
1719 	tipc_link_xmit(tnl, &tnlq, xmitq);
1720 
1721 	if (mtyp == FAILOVER_MSG) {
1722 		tnl->drop_point = l->rcv_nxt;
1723 		tnl->failover_reasm_skb = l->reasm_buf;
1724 		l->reasm_buf = NULL;
1725 
1726 		/* Failover the link's deferdq */
1727 		if (unlikely(!skb_queue_empty(fdefq))) {
1728 			pr_warn("Link failover deferdq not empty: %d!\n",
1729 				skb_queue_len(fdefq));
1730 			__skb_queue_purge(fdefq);
1731 		}
1732 		skb_queue_splice_init(&l->deferdq, fdefq);
1733 	}
1734 }
1735 
1736 /**
1737  * tipc_link_failover_prepare() - prepare tnl for link failover
1738  *
1739  * This is a special version of the precursor - tipc_link_tnl_prepare(),
1740  * see the tipc_node_link_failover() for details
1741  *
1742  * @l: failover link
1743  * @tnl: tunnel link
1744  * @xmitq: queue for messages to be xmited
1745  */
1746 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
1747 				struct sk_buff_head *xmitq)
1748 {
1749 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1750 
1751 	tipc_link_create_dummy_tnl_msg(tnl, xmitq);
1752 
1753 	/* This failover link enpoint was never established before,
1754 	 * so it has not received anything from peer.
1755 	 * Otherwise, it must be a normal failover situation or the
1756 	 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
1757 	 * would have to start over from scratch instead.
1758 	 */
1759 	tnl->drop_point = 1;
1760 	tnl->failover_reasm_skb = NULL;
1761 
1762 	/* Initiate the link's failover deferdq */
1763 	if (unlikely(!skb_queue_empty(fdefq))) {
1764 		pr_warn("Link failover deferdq not empty: %d!\n",
1765 			skb_queue_len(fdefq));
1766 		__skb_queue_purge(fdefq);
1767 	}
1768 }
1769 
1770 /* tipc_link_validate_msg(): validate message against current link state
1771  * Returns true if message should be accepted, otherwise false
1772  */
1773 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
1774 {
1775 	u16 curr_session = l->peer_session;
1776 	u16 session = msg_session(hdr);
1777 	int mtyp = msg_type(hdr);
1778 
1779 	if (msg_user(hdr) != LINK_PROTOCOL)
1780 		return true;
1781 
1782 	switch (mtyp) {
1783 	case RESET_MSG:
1784 		if (!l->in_session)
1785 			return true;
1786 		/* Accept only RESET with new session number */
1787 		return more(session, curr_session);
1788 	case ACTIVATE_MSG:
1789 		if (!l->in_session)
1790 			return true;
1791 		/* Accept only ACTIVATE with new or current session number */
1792 		return !less(session, curr_session);
1793 	case STATE_MSG:
1794 		/* Accept only STATE with current session number */
1795 		if (!l->in_session)
1796 			return false;
1797 		if (session != curr_session)
1798 			return false;
1799 		/* Extra sanity check */
1800 		if (!link_is_up(l) && msg_ack(hdr))
1801 			return false;
1802 		if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
1803 			return true;
1804 		/* Accept only STATE with new sequence number */
1805 		return !less(msg_seqno(hdr), l->rcv_nxt_state);
1806 	default:
1807 		return false;
1808 	}
1809 }
1810 
1811 /* tipc_link_proto_rcv(): receive link level protocol message :
1812  * Note that network plane id propagates through the network, and may
1813  * change at any time. The node with lowest numerical id determines
1814  * network plane
1815  */
1816 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1817 			       struct sk_buff_head *xmitq)
1818 {
1819 	struct tipc_msg *hdr = buf_msg(skb);
1820 	struct tipc_gap_ack_blks *ga = NULL;
1821 	u16 rcvgap = 0;
1822 	u16 ack = msg_ack(hdr);
1823 	u16 gap = msg_seq_gap(hdr);
1824 	u16 peers_snd_nxt =  msg_next_sent(hdr);
1825 	u16 peers_tol = msg_link_tolerance(hdr);
1826 	u16 peers_prio = msg_linkprio(hdr);
1827 	u16 rcv_nxt = l->rcv_nxt;
1828 	u16 dlen = msg_data_sz(hdr);
1829 	int mtyp = msg_type(hdr);
1830 	bool reply = msg_probe(hdr);
1831 	u16 glen = 0;
1832 	void *data;
1833 	char *if_name;
1834 	int rc = 0;
1835 
1836 	trace_tipc_proto_rcv(skb, false, l->name);
1837 	if (tipc_link_is_blocked(l) || !xmitq)
1838 		goto exit;
1839 
1840 	if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1841 		l->net_plane = msg_net_plane(hdr);
1842 
1843 	skb_linearize(skb);
1844 	hdr = buf_msg(skb);
1845 	data = msg_data(hdr);
1846 
1847 	if (!tipc_link_validate_msg(l, hdr)) {
1848 		trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
1849 		trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
1850 		goto exit;
1851 	}
1852 
1853 	switch (mtyp) {
1854 	case RESET_MSG:
1855 	case ACTIVATE_MSG:
1856 		/* Complete own link name with peer's interface name */
1857 		if_name =  strrchr(l->name, ':') + 1;
1858 		if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1859 			break;
1860 		if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1861 			break;
1862 		strncpy(if_name, data, TIPC_MAX_IF_NAME);
1863 
1864 		/* Update own tolerance if peer indicates a non-zero value */
1865 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1866 			l->tolerance = peers_tol;
1867 			l->bc_rcvlink->tolerance = peers_tol;
1868 		}
1869 		/* Update own priority if peer's priority is higher */
1870 		if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1871 			l->priority = peers_prio;
1872 
1873 		/* If peer is going down we want full re-establish cycle */
1874 		if (msg_peer_stopping(hdr)) {
1875 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1876 			break;
1877 		}
1878 
1879 		/* If this endpoint was re-created while peer was ESTABLISHING
1880 		 * it doesn't know current session number. Force re-synch.
1881 		 */
1882 		if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
1883 		    l->session != msg_dest_session(hdr)) {
1884 			if (less(l->session, msg_dest_session(hdr)))
1885 				l->session = msg_dest_session(hdr) + 1;
1886 			break;
1887 		}
1888 
1889 		/* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1890 		if (mtyp == RESET_MSG || !link_is_up(l))
1891 			rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1892 
1893 		/* ACTIVATE_MSG takes up link if it was already locally reset */
1894 		if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
1895 			rc = TIPC_LINK_UP_EVT;
1896 
1897 		l->peer_session = msg_session(hdr);
1898 		l->in_session = true;
1899 		l->peer_bearer_id = msg_bearer_id(hdr);
1900 		if (l->mtu > msg_max_pkt(hdr))
1901 			l->mtu = msg_max_pkt(hdr);
1902 		break;
1903 
1904 	case STATE_MSG:
1905 		l->rcv_nxt_state = msg_seqno(hdr) + 1;
1906 
1907 		/* Update own tolerance if peer indicates a non-zero value */
1908 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1909 			l->tolerance = peers_tol;
1910 			l->bc_rcvlink->tolerance = peers_tol;
1911 		}
1912 		/* Update own prio if peer indicates a different value */
1913 		if ((peers_prio != l->priority) &&
1914 		    in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
1915 			l->priority = peers_prio;
1916 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1917 		}
1918 
1919 		l->silent_intv_cnt = 0;
1920 		l->stats.recv_states++;
1921 		if (msg_probe(hdr))
1922 			l->stats.recv_probes++;
1923 
1924 		if (!link_is_up(l)) {
1925 			if (l->state == LINK_ESTABLISHING)
1926 				rc = TIPC_LINK_UP_EVT;
1927 			break;
1928 		}
1929 
1930 		/* Receive Gap ACK blocks from peer if any */
1931 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
1932 			ga = (struct tipc_gap_ack_blks *)data;
1933 			glen = ntohs(ga->len);
1934 			/* sanity check: if failed, ignore Gap ACK blocks */
1935 			if (glen != tipc_gap_ack_blks_sz(ga->gack_cnt))
1936 				ga = NULL;
1937 		}
1938 
1939 		tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
1940 			     &l->mon_state, l->bearer_id);
1941 
1942 		/* Send NACK if peer has sent pkts we haven't received yet */
1943 		if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1944 			rcvgap = peers_snd_nxt - l->rcv_nxt;
1945 		if (rcvgap || reply)
1946 			tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
1947 						  rcvgap, 0, 0, xmitq);
1948 
1949 		rc |= tipc_link_advance_transmq(l, ack, gap, ga, xmitq);
1950 
1951 		/* If NACK, retransmit will now start at right position */
1952 		if (gap)
1953 			l->stats.recv_nacks++;
1954 
1955 		tipc_link_advance_backlog(l, xmitq);
1956 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
1957 			link_prepare_wakeup(l);
1958 	}
1959 exit:
1960 	kfree_skb(skb);
1961 	return rc;
1962 }
1963 
1964 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1965  */
1966 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1967 					 u16 peers_snd_nxt,
1968 					 struct sk_buff_head *xmitq)
1969 {
1970 	struct sk_buff *skb;
1971 	struct tipc_msg *hdr;
1972 	struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1973 	u16 ack = l->rcv_nxt - 1;
1974 	u16 gap_to = peers_snd_nxt - 1;
1975 
1976 	skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1977 			      0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1978 	if (!skb)
1979 		return false;
1980 	hdr = buf_msg(skb);
1981 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1982 	msg_set_bcast_ack(hdr, ack);
1983 	msg_set_bcgap_after(hdr, ack);
1984 	if (dfrd_skb)
1985 		gap_to = buf_seqno(dfrd_skb) - 1;
1986 	msg_set_bcgap_to(hdr, gap_to);
1987 	msg_set_non_seq(hdr, bcast);
1988 	__skb_queue_tail(xmitq, skb);
1989 	return true;
1990 }
1991 
1992 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1993  *
1994  * Give a newly added peer node the sequence number where it should
1995  * start receiving and acking broadcast packets.
1996  */
1997 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1998 					struct sk_buff_head *xmitq)
1999 {
2000 	struct sk_buff_head list;
2001 
2002 	__skb_queue_head_init(&list);
2003 	if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2004 		return;
2005 	msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2006 	tipc_link_xmit(l, &list, xmitq);
2007 }
2008 
2009 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2010  */
2011 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2012 {
2013 	int mtyp = msg_type(hdr);
2014 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2015 
2016 	if (link_is_up(l))
2017 		return;
2018 
2019 	if (msg_user(hdr) == BCAST_PROTOCOL) {
2020 		l->rcv_nxt = peers_snd_nxt;
2021 		l->state = LINK_ESTABLISHED;
2022 		return;
2023 	}
2024 
2025 	if (l->peer_caps & TIPC_BCAST_SYNCH)
2026 		return;
2027 
2028 	if (msg_peer_node_is_up(hdr))
2029 		return;
2030 
2031 	/* Compatibility: accept older, less safe initial synch data */
2032 	if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2033 		l->rcv_nxt = peers_snd_nxt;
2034 }
2035 
2036 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2037  */
2038 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2039 			  struct sk_buff_head *xmitq)
2040 {
2041 	struct tipc_link *snd_l = l->bc_sndlink;
2042 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2043 	u16 from = msg_bcast_ack(hdr) + 1;
2044 	u16 to = from + msg_bc_gap(hdr) - 1;
2045 	int rc = 0;
2046 
2047 	if (!link_is_up(l))
2048 		return rc;
2049 
2050 	if (!msg_peer_node_is_up(hdr))
2051 		return rc;
2052 
2053 	/* Open when peer ackowledges our bcast init msg (pkt #1) */
2054 	if (msg_ack(hdr))
2055 		l->bc_peer_is_up = true;
2056 
2057 	if (!l->bc_peer_is_up)
2058 		return rc;
2059 
2060 	l->stats.recv_nacks++;
2061 
2062 	/* Ignore if peers_snd_nxt goes beyond receive window */
2063 	if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2064 		return rc;
2065 
2066 	rc = tipc_link_bc_retrans(snd_l, l, from, to, xmitq);
2067 
2068 	l->snd_nxt = peers_snd_nxt;
2069 	if (link_bc_rcv_gap(l))
2070 		rc |= TIPC_LINK_SND_STATE;
2071 
2072 	/* Return now if sender supports nack via STATE messages */
2073 	if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2074 		return rc;
2075 
2076 	/* Otherwise, be backwards compatible */
2077 
2078 	if (!more(peers_snd_nxt, l->rcv_nxt)) {
2079 		l->nack_state = BC_NACK_SND_CONDITIONAL;
2080 		return 0;
2081 	}
2082 
2083 	/* Don't NACK if one was recently sent or peeked */
2084 	if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2085 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2086 		return 0;
2087 	}
2088 
2089 	/* Conditionally delay NACK sending until next synch rcv */
2090 	if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2091 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2092 		if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2093 			return 0;
2094 	}
2095 
2096 	/* Send NACK now but suppress next one */
2097 	tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2098 	l->nack_state = BC_NACK_SND_SUPPRESS;
2099 	return 0;
2100 }
2101 
2102 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
2103 			  struct sk_buff_head *xmitq)
2104 {
2105 	struct sk_buff *skb, *tmp;
2106 	struct tipc_link *snd_l = l->bc_sndlink;
2107 
2108 	if (!link_is_up(l) || !l->bc_peer_is_up)
2109 		return;
2110 
2111 	if (!more(acked, l->acked))
2112 		return;
2113 
2114 	trace_tipc_link_bc_ack(l, l->acked, acked, &snd_l->transmq);
2115 	/* Skip over packets peer has already acked */
2116 	skb_queue_walk(&snd_l->transmq, skb) {
2117 		if (more(buf_seqno(skb), l->acked))
2118 			break;
2119 	}
2120 
2121 	/* Update/release the packets peer is acking now */
2122 	skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
2123 		if (more(buf_seqno(skb), acked))
2124 			break;
2125 		if (!--TIPC_SKB_CB(skb)->ackers) {
2126 			__skb_unlink(skb, &snd_l->transmq);
2127 			kfree_skb(skb);
2128 		}
2129 	}
2130 	l->acked = acked;
2131 	tipc_link_advance_backlog(snd_l, xmitq);
2132 	if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
2133 		link_prepare_wakeup(snd_l);
2134 }
2135 
2136 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2137  * This function is here for backwards compatibility, since
2138  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2139  */
2140 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2141 			  struct sk_buff_head *xmitq)
2142 {
2143 	struct tipc_msg *hdr = buf_msg(skb);
2144 	u32 dnode = msg_destnode(hdr);
2145 	int mtyp = msg_type(hdr);
2146 	u16 acked = msg_bcast_ack(hdr);
2147 	u16 from = acked + 1;
2148 	u16 to = msg_bcgap_to(hdr);
2149 	u16 peers_snd_nxt = to + 1;
2150 	int rc = 0;
2151 
2152 	kfree_skb(skb);
2153 
2154 	if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2155 		return 0;
2156 
2157 	if (mtyp != STATE_MSG)
2158 		return 0;
2159 
2160 	if (dnode == tipc_own_addr(l->net)) {
2161 		tipc_link_bc_ack_rcv(l, acked, xmitq);
2162 		rc = tipc_link_bc_retrans(l->bc_sndlink, l, from, to, xmitq);
2163 		l->stats.recv_nacks++;
2164 		return rc;
2165 	}
2166 
2167 	/* Msg for other node => suppress own NACK at next sync if applicable */
2168 	if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2169 		l->nack_state = BC_NACK_SND_SUPPRESS;
2170 
2171 	return 0;
2172 }
2173 
2174 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
2175 {
2176 	int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2177 
2178 	l->window = win;
2179 	l->backlog[TIPC_LOW_IMPORTANCE].limit      = max_t(u16, 50, win);
2180 	l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = max_t(u16, 100, win * 2);
2181 	l->backlog[TIPC_HIGH_IMPORTANCE].limit     = max_t(u16, 150, win * 3);
2182 	l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
2183 	l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
2184 }
2185 
2186 /**
2187  * link_reset_stats - reset link statistics
2188  * @l: pointer to link
2189  */
2190 void tipc_link_reset_stats(struct tipc_link *l)
2191 {
2192 	memset(&l->stats, 0, sizeof(l->stats));
2193 }
2194 
2195 static void link_print(struct tipc_link *l, const char *str)
2196 {
2197 	struct sk_buff *hskb = skb_peek(&l->transmq);
2198 	u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2199 	u16 tail = l->snd_nxt - 1;
2200 
2201 	pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2202 	pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2203 		skb_queue_len(&l->transmq), head, tail,
2204 		skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2205 }
2206 
2207 /* Parse and validate nested (link) properties valid for media, bearer and link
2208  */
2209 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2210 {
2211 	int err;
2212 
2213 	err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2214 					  tipc_nl_prop_policy, NULL);
2215 	if (err)
2216 		return err;
2217 
2218 	if (props[TIPC_NLA_PROP_PRIO]) {
2219 		u32 prio;
2220 
2221 		prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2222 		if (prio > TIPC_MAX_LINK_PRI)
2223 			return -EINVAL;
2224 	}
2225 
2226 	if (props[TIPC_NLA_PROP_TOL]) {
2227 		u32 tol;
2228 
2229 		tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2230 		if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2231 			return -EINVAL;
2232 	}
2233 
2234 	if (props[TIPC_NLA_PROP_WIN]) {
2235 		u32 win;
2236 
2237 		win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2238 		if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
2239 			return -EINVAL;
2240 	}
2241 
2242 	return 0;
2243 }
2244 
2245 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2246 {
2247 	int i;
2248 	struct nlattr *stats;
2249 
2250 	struct nla_map {
2251 		u32 key;
2252 		u32 val;
2253 	};
2254 
2255 	struct nla_map map[] = {
2256 		{TIPC_NLA_STATS_RX_INFO, 0},
2257 		{TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2258 		{TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2259 		{TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2260 		{TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2261 		{TIPC_NLA_STATS_TX_INFO, 0},
2262 		{TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2263 		{TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2264 		{TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2265 		{TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2266 		{TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2267 			s->msg_length_counts : 1},
2268 		{TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2269 		{TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2270 		{TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2271 		{TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2272 		{TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2273 		{TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2274 		{TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2275 		{TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2276 		{TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2277 		{TIPC_NLA_STATS_RX_STATES, s->recv_states},
2278 		{TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2279 		{TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2280 		{TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2281 		{TIPC_NLA_STATS_TX_STATES, s->sent_states},
2282 		{TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2283 		{TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2284 		{TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2285 		{TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2286 		{TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2287 		{TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2288 		{TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2289 		{TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2290 			(s->accu_queue_sz / s->queue_sz_counts) : 0}
2291 	};
2292 
2293 	stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2294 	if (!stats)
2295 		return -EMSGSIZE;
2296 
2297 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2298 		if (nla_put_u32(skb, map[i].key, map[i].val))
2299 			goto msg_full;
2300 
2301 	nla_nest_end(skb, stats);
2302 
2303 	return 0;
2304 msg_full:
2305 	nla_nest_cancel(skb, stats);
2306 
2307 	return -EMSGSIZE;
2308 }
2309 
2310 /* Caller should hold appropriate locks to protect the link */
2311 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2312 		       struct tipc_link *link, int nlflags)
2313 {
2314 	u32 self = tipc_own_addr(net);
2315 	struct nlattr *attrs;
2316 	struct nlattr *prop;
2317 	void *hdr;
2318 	int err;
2319 
2320 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2321 			  nlflags, TIPC_NL_LINK_GET);
2322 	if (!hdr)
2323 		return -EMSGSIZE;
2324 
2325 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2326 	if (!attrs)
2327 		goto msg_full;
2328 
2329 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2330 		goto attr_msg_full;
2331 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2332 		goto attr_msg_full;
2333 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2334 		goto attr_msg_full;
2335 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2336 		goto attr_msg_full;
2337 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2338 		goto attr_msg_full;
2339 
2340 	if (tipc_link_is_up(link))
2341 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2342 			goto attr_msg_full;
2343 	if (link->active)
2344 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2345 			goto attr_msg_full;
2346 
2347 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2348 	if (!prop)
2349 		goto attr_msg_full;
2350 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2351 		goto prop_msg_full;
2352 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2353 		goto prop_msg_full;
2354 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2355 			link->window))
2356 		goto prop_msg_full;
2357 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2358 		goto prop_msg_full;
2359 	nla_nest_end(msg->skb, prop);
2360 
2361 	err = __tipc_nl_add_stats(msg->skb, &link->stats);
2362 	if (err)
2363 		goto attr_msg_full;
2364 
2365 	nla_nest_end(msg->skb, attrs);
2366 	genlmsg_end(msg->skb, hdr);
2367 
2368 	return 0;
2369 
2370 prop_msg_full:
2371 	nla_nest_cancel(msg->skb, prop);
2372 attr_msg_full:
2373 	nla_nest_cancel(msg->skb, attrs);
2374 msg_full:
2375 	genlmsg_cancel(msg->skb, hdr);
2376 
2377 	return -EMSGSIZE;
2378 }
2379 
2380 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2381 				      struct tipc_stats *stats)
2382 {
2383 	int i;
2384 	struct nlattr *nest;
2385 
2386 	struct nla_map {
2387 		__u32 key;
2388 		__u32 val;
2389 	};
2390 
2391 	struct nla_map map[] = {
2392 		{TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2393 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2394 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2395 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2396 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2397 		{TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2398 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2399 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2400 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2401 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2402 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2403 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2404 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2405 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2406 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2407 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2408 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2409 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2410 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2411 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2412 	};
2413 
2414 	nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2415 	if (!nest)
2416 		return -EMSGSIZE;
2417 
2418 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2419 		if (nla_put_u32(skb, map[i].key, map[i].val))
2420 			goto msg_full;
2421 
2422 	nla_nest_end(skb, nest);
2423 
2424 	return 0;
2425 msg_full:
2426 	nla_nest_cancel(skb, nest);
2427 
2428 	return -EMSGSIZE;
2429 }
2430 
2431 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2432 {
2433 	int err;
2434 	void *hdr;
2435 	struct nlattr *attrs;
2436 	struct nlattr *prop;
2437 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2438 	u32 bc_mode = tipc_bcast_get_broadcast_mode(net);
2439 	u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2440 	struct tipc_link *bcl = tn->bcl;
2441 
2442 	if (!bcl)
2443 		return 0;
2444 
2445 	tipc_bcast_lock(net);
2446 
2447 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2448 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
2449 	if (!hdr) {
2450 		tipc_bcast_unlock(net);
2451 		return -EMSGSIZE;
2452 	}
2453 
2454 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2455 	if (!attrs)
2456 		goto msg_full;
2457 
2458 	/* The broadcast link is always up */
2459 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2460 		goto attr_msg_full;
2461 
2462 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2463 		goto attr_msg_full;
2464 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2465 		goto attr_msg_full;
2466 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2467 		goto attr_msg_full;
2468 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2469 		goto attr_msg_full;
2470 
2471 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2472 	if (!prop)
2473 		goto attr_msg_full;
2474 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2475 		goto prop_msg_full;
2476 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2477 		goto prop_msg_full;
2478 	if (bc_mode & BCLINK_MODE_SEL)
2479 		if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2480 				bc_ratio))
2481 			goto prop_msg_full;
2482 	nla_nest_end(msg->skb, prop);
2483 
2484 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2485 	if (err)
2486 		goto attr_msg_full;
2487 
2488 	tipc_bcast_unlock(net);
2489 	nla_nest_end(msg->skb, attrs);
2490 	genlmsg_end(msg->skb, hdr);
2491 
2492 	return 0;
2493 
2494 prop_msg_full:
2495 	nla_nest_cancel(msg->skb, prop);
2496 attr_msg_full:
2497 	nla_nest_cancel(msg->skb, attrs);
2498 msg_full:
2499 	tipc_bcast_unlock(net);
2500 	genlmsg_cancel(msg->skb, hdr);
2501 
2502 	return -EMSGSIZE;
2503 }
2504 
2505 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2506 			     struct sk_buff_head *xmitq)
2507 {
2508 	l->tolerance = tol;
2509 	if (l->bc_rcvlink)
2510 		l->bc_rcvlink->tolerance = tol;
2511 	if (link_is_up(l))
2512 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2513 }
2514 
2515 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2516 			struct sk_buff_head *xmitq)
2517 {
2518 	l->priority = prio;
2519 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2520 }
2521 
2522 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2523 {
2524 	l->abort_limit = limit;
2525 }
2526 
2527 char *tipc_link_name_ext(struct tipc_link *l, char *buf)
2528 {
2529 	if (!l)
2530 		scnprintf(buf, TIPC_MAX_LINK_NAME, "null");
2531 	else if (link_is_bc_sndlink(l))
2532 		scnprintf(buf, TIPC_MAX_LINK_NAME, "broadcast-sender");
2533 	else if (link_is_bc_rcvlink(l))
2534 		scnprintf(buf, TIPC_MAX_LINK_NAME,
2535 			  "broadcast-receiver, peer %x", l->addr);
2536 	else
2537 		memcpy(buf, l->name, TIPC_MAX_LINK_NAME);
2538 
2539 	return buf;
2540 }
2541 
2542 /**
2543  * tipc_link_dump - dump TIPC link data
2544  * @l: tipc link to be dumped
2545  * @dqueues: bitmask to decide if any link queue to be dumped?
2546  *           - TIPC_DUMP_NONE: don't dump link queues
2547  *           - TIPC_DUMP_TRANSMQ: dump link transmq queue
2548  *           - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2549  *           - TIPC_DUMP_DEFERDQ: dump link deferd queue
2550  *           - TIPC_DUMP_INPUTQ: dump link input queue
2551  *           - TIPC_DUMP_WAKEUP: dump link wakeup queue
2552  *           - TIPC_DUMP_ALL: dump all the link queues above
2553  * @buf: returned buffer of dump data in format
2554  */
2555 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2556 {
2557 	int i = 0;
2558 	size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2559 	struct sk_buff_head *list;
2560 	struct sk_buff *hskb, *tskb;
2561 	u32 len;
2562 
2563 	if (!l) {
2564 		i += scnprintf(buf, sz, "link data: (null)\n");
2565 		return i;
2566 	}
2567 
2568 	i += scnprintf(buf, sz, "link data: %x", l->addr);
2569 	i += scnprintf(buf + i, sz - i, " %x", l->state);
2570 	i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2571 	i += scnprintf(buf + i, sz - i, " %u", l->session);
2572 	i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2573 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2574 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2575 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2576 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2577 	i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2578 	i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2579 	i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2580 	i += scnprintf(buf + i, sz - i, " %u", l->prev_from);
2581 	i += scnprintf(buf + i, sz - i, " %u", 0);
2582 	i += scnprintf(buf + i, sz - i, " %u", l->acked);
2583 
2584 	list = &l->transmq;
2585 	len = skb_queue_len(list);
2586 	hskb = skb_peek(list);
2587 	tskb = skb_peek_tail(list);
2588 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2589 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2590 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2591 
2592 	list = &l->deferdq;
2593 	len = skb_queue_len(list);
2594 	hskb = skb_peek(list);
2595 	tskb = skb_peek_tail(list);
2596 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2597 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2598 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2599 
2600 	list = &l->backlogq;
2601 	len = skb_queue_len(list);
2602 	hskb = skb_peek(list);
2603 	tskb = skb_peek_tail(list);
2604 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2605 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2606 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2607 
2608 	list = l->inputq;
2609 	len = skb_queue_len(list);
2610 	hskb = skb_peek(list);
2611 	tskb = skb_peek_tail(list);
2612 	i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2613 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2614 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2615 
2616 	if (dqueues & TIPC_DUMP_TRANSMQ) {
2617 		i += scnprintf(buf + i, sz - i, "transmq: ");
2618 		i += tipc_list_dump(&l->transmq, false, buf + i);
2619 	}
2620 	if (dqueues & TIPC_DUMP_BACKLOGQ) {
2621 		i += scnprintf(buf + i, sz - i,
2622 			       "backlogq: <%u %u %u %u %u>, ",
2623 			       l->backlog[TIPC_LOW_IMPORTANCE].len,
2624 			       l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2625 			       l->backlog[TIPC_HIGH_IMPORTANCE].len,
2626 			       l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2627 			       l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2628 		i += tipc_list_dump(&l->backlogq, false, buf + i);
2629 	}
2630 	if (dqueues & TIPC_DUMP_DEFERDQ) {
2631 		i += scnprintf(buf + i, sz - i, "deferdq: ");
2632 		i += tipc_list_dump(&l->deferdq, false, buf + i);
2633 	}
2634 	if (dqueues & TIPC_DUMP_INPUTQ) {
2635 		i += scnprintf(buf + i, sz - i, "inputq: ");
2636 		i += tipc_list_dump(l->inputq, false, buf + i);
2637 	}
2638 	if (dqueues & TIPC_DUMP_WAKEUP) {
2639 		i += scnprintf(buf + i, sz - i, "wakeup: ");
2640 		i += tipc_list_dump(&l->wakeupq, false, buf + i);
2641 	}
2642 
2643 	return i;
2644 }
2645