• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * net/tipc/link.c: TIPC link code
3  *
4  * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5  * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #include "core.h"
38 #include "subscr.h"
39 #include "link.h"
40 #include "bcast.h"
41 #include "socket.h"
42 #include "name_distr.h"
43 #include "discover.h"
44 #include "netlink.h"
45 #include "monitor.h"
46 
47 #include <linux/pkt_sched.h>
48 
49 struct tipc_stats {
50 	u32 sent_pkts;
51 	u32 recv_pkts;
52 	u32 sent_states;
53 	u32 recv_states;
54 	u32 sent_probes;
55 	u32 recv_probes;
56 	u32 sent_nacks;
57 	u32 recv_nacks;
58 	u32 sent_acks;
59 	u32 sent_bundled;
60 	u32 sent_bundles;
61 	u32 recv_bundled;
62 	u32 recv_bundles;
63 	u32 retransmitted;
64 	u32 sent_fragmented;
65 	u32 sent_fragments;
66 	u32 recv_fragmented;
67 	u32 recv_fragments;
68 	u32 link_congs;		/* # port sends blocked by congestion */
69 	u32 deferred_recv;
70 	u32 duplicates;
71 	u32 max_queue_sz;	/* send queue size high water mark */
72 	u32 accu_queue_sz;	/* used for send queue size profiling */
73 	u32 queue_sz_counts;	/* used for send queue size profiling */
74 	u32 msg_length_counts;	/* used for message length profiling */
75 	u32 msg_lengths_total;	/* used for message length profiling */
76 	u32 msg_length_profile[7]; /* used for msg. length profiling */
77 };
78 
79 /**
80  * struct tipc_link - TIPC link data structure
81  * @addr: network address of link's peer node
82  * @name: link name character string
83  * @media_addr: media address to use when sending messages over link
84  * @timer: link timer
85  * @net: pointer to namespace struct
86  * @refcnt: reference counter for permanent references (owner node & timer)
87  * @peer_session: link session # being used by peer end of link
88  * @peer_bearer_id: bearer id used by link's peer endpoint
89  * @bearer_id: local bearer id used by link
90  * @tolerance: minimum link continuity loss needed to reset link [in ms]
91  * @abort_limit: # of unacknowledged continuity probes needed to reset link
92  * @state: current state of link FSM
93  * @peer_caps: bitmap describing capabilities of peer node
94  * @silent_intv_cnt: # of timer intervals without any reception from peer
95  * @proto_msg: template for control messages generated by link
96  * @pmsg: convenience pointer to "proto_msg" field
97  * @priority: current link priority
98  * @net_plane: current link network plane ('A' through 'H')
99  * @mon_state: cookie with information needed by link monitor
100  * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
101  * @exp_msg_count: # of tunnelled messages expected during link changeover
102  * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
103  * @mtu: current maximum packet size for this link
104  * @advertised_mtu: advertised own mtu when link is being established
105  * @transmitq: queue for sent, non-acked messages
106  * @backlogq: queue for messages waiting to be sent
107  * @snt_nxt: next sequence number to use for outbound messages
108  * @last_retransmitted: sequence number of most recently retransmitted message
109  * @stale_count: # of identical retransmit requests made by peer
110  * @ackers: # of peers that needs to ack each packet before it can be released
111  * @acked: # last packet acked by a certain peer. Used for broadcast.
112  * @rcv_nxt: next sequence number to expect for inbound messages
113  * @deferred_queue: deferred queue saved OOS b'cast message received from node
114  * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
115  * @inputq: buffer queue for messages to be delivered upwards
116  * @namedq: buffer queue for name table messages to be delivered upwards
117  * @next_out: ptr to first unsent outbound message in queue
118  * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
119  * @long_msg_seq_no: next identifier to use for outbound fragmented messages
120  * @reasm_buf: head of partially reassembled inbound message fragments
121  * @bc_rcvr: marks that this is a broadcast receiver link
122  * @stats: collects statistics regarding link activity
123  */
124 struct tipc_link {
125 	u32 addr;
126 	char name[TIPC_MAX_LINK_NAME];
127 	struct net *net;
128 
129 	/* Management and link supervision data */
130 	u32 peer_session;
131 	u32 session;
132 	u32 peer_bearer_id;
133 	u32 bearer_id;
134 	u32 tolerance;
135 	u32 abort_limit;
136 	u32 state;
137 	u16 peer_caps;
138 	bool active;
139 	u32 silent_intv_cnt;
140 	char if_name[TIPC_MAX_IF_NAME];
141 	u32 priority;
142 	char net_plane;
143 	struct tipc_mon_state mon_state;
144 	u16 rst_cnt;
145 
146 	/* Failover/synch */
147 	u16 drop_point;
148 	struct sk_buff *failover_reasm_skb;
149 
150 	/* Max packet negotiation */
151 	u16 mtu;
152 	u16 advertised_mtu;
153 
154 	/* Sending */
155 	struct sk_buff_head transmq;
156 	struct sk_buff_head backlogq;
157 	struct {
158 		u16 len;
159 		u16 limit;
160 		struct sk_buff *target_bskb;
161 	} backlog[5];
162 	u16 snd_nxt;
163 	u16 last_retransm;
164 	u16 window;
165 	u32 stale_count;
166 
167 	/* Reception */
168 	u16 rcv_nxt;
169 	u32 rcv_unacked;
170 	struct sk_buff_head deferdq;
171 	struct sk_buff_head *inputq;
172 	struct sk_buff_head *namedq;
173 
174 	/* Congestion handling */
175 	struct sk_buff_head wakeupq;
176 
177 	/* Fragmentation/reassembly */
178 	struct sk_buff *reasm_buf;
179 
180 	/* Broadcast */
181 	u16 ackers;
182 	u16 acked;
183 	struct tipc_link *bc_rcvlink;
184 	struct tipc_link *bc_sndlink;
185 	unsigned long prev_retr;
186 	u16 prev_from;
187 	u16 prev_to;
188 	u8 nack_state;
189 	bool bc_peer_is_up;
190 
191 	/* Statistics */
192 	struct tipc_stats stats;
193 };
194 
195 /*
196  * Error message prefixes
197  */
198 static const char *link_co_err = "Link tunneling error, ";
199 static const char *link_rst_msg = "Resetting link ";
200 
201 /* Send states for broadcast NACKs
202  */
203 enum {
204 	BC_NACK_SND_CONDITIONAL,
205 	BC_NACK_SND_UNCONDITIONAL,
206 	BC_NACK_SND_SUPPRESS,
207 };
208 
209 #define TIPC_BC_RETR_LIMIT 10   /* [ms] */
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 /* Wildcard value for link session numbers. When it is known that
217  * peer endpoint is down, any session number must be accepted.
218  */
219 #define ANY_SESSION 0x10000
220 
221 /* Link FSM states:
222  */
223 enum {
224 	LINK_ESTABLISHED     = 0xe,
225 	LINK_ESTABLISHING    = 0xe  << 4,
226 	LINK_RESET           = 0x1  << 8,
227 	LINK_RESETTING       = 0x2  << 12,
228 	LINK_PEER_RESET      = 0xd  << 16,
229 	LINK_FAILINGOVER     = 0xf  << 20,
230 	LINK_SYNCHING        = 0xc  << 24
231 };
232 
233 /* Link FSM state checking routines
234  */
link_is_up(struct tipc_link * l)235 static int link_is_up(struct tipc_link *l)
236 {
237 	return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
238 }
239 
240 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
241 			       struct sk_buff_head *xmitq);
242 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
243 				      u16 rcvgap, int tolerance, int priority,
244 				      struct sk_buff_head *xmitq);
245 static void link_print(struct tipc_link *l, const char *str);
246 static int tipc_link_build_nack_msg(struct tipc_link *l,
247 				    struct sk_buff_head *xmitq);
248 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
249 					struct sk_buff_head *xmitq);
250 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
251 
252 /*
253  *  Simple non-static link routines (i.e. referenced outside this file)
254  */
tipc_link_is_up(struct tipc_link * l)255 bool tipc_link_is_up(struct tipc_link *l)
256 {
257 	return link_is_up(l);
258 }
259 
tipc_link_peer_is_down(struct tipc_link * l)260 bool tipc_link_peer_is_down(struct tipc_link *l)
261 {
262 	return l->state == LINK_PEER_RESET;
263 }
264 
tipc_link_is_reset(struct tipc_link * l)265 bool tipc_link_is_reset(struct tipc_link *l)
266 {
267 	return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
268 }
269 
tipc_link_is_establishing(struct tipc_link * l)270 bool tipc_link_is_establishing(struct tipc_link *l)
271 {
272 	return l->state == LINK_ESTABLISHING;
273 }
274 
tipc_link_is_synching(struct tipc_link * l)275 bool tipc_link_is_synching(struct tipc_link *l)
276 {
277 	return l->state == LINK_SYNCHING;
278 }
279 
tipc_link_is_failingover(struct tipc_link * l)280 bool tipc_link_is_failingover(struct tipc_link *l)
281 {
282 	return l->state == LINK_FAILINGOVER;
283 }
284 
tipc_link_is_blocked(struct tipc_link * l)285 bool tipc_link_is_blocked(struct tipc_link *l)
286 {
287 	return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
288 }
289 
link_is_bc_sndlink(struct tipc_link * l)290 static bool link_is_bc_sndlink(struct tipc_link *l)
291 {
292 	return !l->bc_sndlink;
293 }
294 
link_is_bc_rcvlink(struct tipc_link * l)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 
tipc_link_is_active(struct tipc_link * l)300 int tipc_link_is_active(struct tipc_link *l)
301 {
302 	return l->active;
303 }
304 
tipc_link_set_active(struct tipc_link * l,bool active)305 void tipc_link_set_active(struct tipc_link *l, bool active)
306 {
307 	l->active = active;
308 }
309 
tipc_link_id(struct tipc_link * l)310 u32 tipc_link_id(struct tipc_link *l)
311 {
312 	return l->peer_bearer_id << 16 | l->bearer_id;
313 }
314 
tipc_link_window(struct tipc_link * l)315 int tipc_link_window(struct tipc_link *l)
316 {
317 	return l->window;
318 }
319 
tipc_link_prio(struct tipc_link * l)320 int tipc_link_prio(struct tipc_link *l)
321 {
322 	return l->priority;
323 }
324 
tipc_link_tolerance(struct tipc_link * l)325 unsigned long tipc_link_tolerance(struct tipc_link *l)
326 {
327 	return l->tolerance;
328 }
329 
tipc_link_inputq(struct tipc_link * l)330 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
331 {
332 	return l->inputq;
333 }
334 
tipc_link_plane(struct tipc_link * l)335 char tipc_link_plane(struct tipc_link *l)
336 {
337 	return l->net_plane;
338 }
339 
tipc_link_add_bc_peer(struct tipc_link * snd_l,struct tipc_link * uc_l,struct sk_buff_head * xmitq)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 
tipc_link_remove_bc_peer(struct tipc_link * snd_l,struct tipc_link * rcv_l,struct sk_buff_head * xmitq)352 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
353 			      struct tipc_link *rcv_l,
354 			      struct sk_buff_head *xmitq)
355 {
356 	u16 ack = snd_l->snd_nxt - 1;
357 
358 	snd_l->ackers--;
359 	rcv_l->bc_peer_is_up = true;
360 	rcv_l->state = LINK_ESTABLISHED;
361 	tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
362 	tipc_link_reset(rcv_l);
363 	rcv_l->state = LINK_RESET;
364 	if (!snd_l->ackers) {
365 		tipc_link_reset(snd_l);
366 		snd_l->state = LINK_RESET;
367 		__skb_queue_purge(xmitq);
368 	}
369 }
370 
tipc_link_bc_peers(struct tipc_link * l)371 int tipc_link_bc_peers(struct tipc_link *l)
372 {
373 	return l->ackers;
374 }
375 
link_bc_rcv_gap(struct tipc_link * l)376 u16 link_bc_rcv_gap(struct tipc_link *l)
377 {
378 	struct sk_buff *skb = skb_peek(&l->deferdq);
379 	u16 gap = 0;
380 
381 	if (more(l->snd_nxt, l->rcv_nxt))
382 		gap = l->snd_nxt - l->rcv_nxt;
383 	if (skb)
384 		gap = buf_seqno(skb) - l->rcv_nxt;
385 	return gap;
386 }
387 
tipc_link_set_mtu(struct tipc_link * l,int mtu)388 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
389 {
390 	l->mtu = mtu;
391 }
392 
tipc_link_mtu(struct tipc_link * l)393 int tipc_link_mtu(struct tipc_link *l)
394 {
395 	return l->mtu;
396 }
397 
tipc_link_rcv_nxt(struct tipc_link * l)398 u16 tipc_link_rcv_nxt(struct tipc_link *l)
399 {
400 	return l->rcv_nxt;
401 }
402 
tipc_link_acked(struct tipc_link * l)403 u16 tipc_link_acked(struct tipc_link *l)
404 {
405 	return l->acked;
406 }
407 
tipc_link_name(struct tipc_link * l)408 char *tipc_link_name(struct tipc_link *l)
409 {
410 	return l->name;
411 }
412 
413 /**
414  * tipc_link_create - create a new link
415  * @n: pointer to associated node
416  * @if_name: associated interface name
417  * @bearer_id: id (index) of associated bearer
418  * @tolerance: link tolerance to be used by link
419  * @net_plane: network plane (A,B,c..) this link belongs to
420  * @mtu: mtu to be advertised by link
421  * @priority: priority to be used by link
422  * @window: send window to be used by link
423  * @session: session to be used by link
424  * @ownnode: identity of own node
425  * @peer: node id of peer node
426  * @peer_caps: bitmap describing peer node capabilities
427  * @bc_sndlink: the namespace global link used for broadcast sending
428  * @bc_rcvlink: the peer specific link used for broadcast reception
429  * @inputq: queue to put messages ready for delivery
430  * @namedq: queue to put binding table update messages ready for delivery
431  * @link: return value, pointer to put the created link
432  *
433  * Returns true if link was created, otherwise false
434  */
tipc_link_create(struct net * net,char * if_name,int bearer_id,int tolerance,char net_plane,u32 mtu,int priority,int window,u32 session,u32 ownnode,u32 peer,u16 peer_caps,struct tipc_link * bc_sndlink,struct tipc_link * bc_rcvlink,struct sk_buff_head * inputq,struct sk_buff_head * namedq,struct tipc_link ** link)435 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
436 		      int tolerance, char net_plane, u32 mtu, int priority,
437 		      int window, u32 session, u32 ownnode, u32 peer,
438 		      u16 peer_caps,
439 		      struct tipc_link *bc_sndlink,
440 		      struct tipc_link *bc_rcvlink,
441 		      struct sk_buff_head *inputq,
442 		      struct sk_buff_head *namedq,
443 		      struct tipc_link **link)
444 {
445 	struct tipc_link *l;
446 
447 	l = kzalloc(sizeof(*l), GFP_ATOMIC);
448 	if (!l)
449 		return false;
450 	*link = l;
451 	l->session = session;
452 
453 	/* Note: peer i/f name is completed by reset/activate message */
454 	sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
455 		tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
456 		if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
457 	strcpy(l->if_name, if_name);
458 	l->addr = peer;
459 	l->peer_caps = peer_caps;
460 	l->net = net;
461 	l->peer_session = ANY_SESSION;
462 	l->bearer_id = bearer_id;
463 	l->tolerance = tolerance;
464 	l->net_plane = net_plane;
465 	l->advertised_mtu = mtu;
466 	l->mtu = mtu;
467 	l->priority = priority;
468 	tipc_link_set_queue_limits(l, window);
469 	l->ackers = 1;
470 	l->bc_sndlink = bc_sndlink;
471 	l->bc_rcvlink = bc_rcvlink;
472 	l->inputq = inputq;
473 	l->namedq = namedq;
474 	l->state = LINK_RESETTING;
475 	__skb_queue_head_init(&l->transmq);
476 	__skb_queue_head_init(&l->backlogq);
477 	__skb_queue_head_init(&l->deferdq);
478 	skb_queue_head_init(&l->wakeupq);
479 	skb_queue_head_init(l->inputq);
480 	return true;
481 }
482 
483 /**
484  * tipc_link_bc_create - create new link to be used for broadcast
485  * @n: pointer to associated node
486  * @mtu: mtu to be used
487  * @window: send window to be used
488  * @inputq: queue to put messages ready for delivery
489  * @namedq: queue to put binding table update messages ready for delivery
490  * @link: return value, pointer to put the created link
491  *
492  * Returns true if link was created, otherwise false
493  */
tipc_link_bc_create(struct net * net,u32 ownnode,u32 peer,int mtu,int window,u16 peer_caps,struct sk_buff_head * inputq,struct sk_buff_head * namedq,struct tipc_link * bc_sndlink,struct tipc_link ** link)494 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
495 			 int mtu, int window, u16 peer_caps,
496 			 struct sk_buff_head *inputq,
497 			 struct sk_buff_head *namedq,
498 			 struct tipc_link *bc_sndlink,
499 			 struct tipc_link **link)
500 {
501 	struct tipc_link *l;
502 
503 	if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
504 			      0, ownnode, peer, peer_caps, bc_sndlink,
505 			      NULL, inputq, namedq, link))
506 		return false;
507 
508 	l = *link;
509 	strcpy(l->name, tipc_bclink_name);
510 	tipc_link_reset(l);
511 	l->state = LINK_RESET;
512 	l->ackers = 0;
513 	l->bc_rcvlink = l;
514 
515 	/* Broadcast send link is always up */
516 	if (link_is_bc_sndlink(l))
517 		l->state = LINK_ESTABLISHED;
518 
519 	/* Disable replicast if even a single peer doesn't support it */
520 	if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
521 		tipc_bcast_disable_rcast(net);
522 
523 	return true;
524 }
525 
526 /**
527  * tipc_link_fsm_evt - link finite state machine
528  * @l: pointer to link
529  * @evt: state machine event to be processed
530  */
tipc_link_fsm_evt(struct tipc_link * l,int evt)531 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
532 {
533 	int rc = 0;
534 
535 	switch (l->state) {
536 	case LINK_RESETTING:
537 		switch (evt) {
538 		case LINK_PEER_RESET_EVT:
539 			l->state = LINK_PEER_RESET;
540 			break;
541 		case LINK_RESET_EVT:
542 			l->state = LINK_RESET;
543 			break;
544 		case LINK_FAILURE_EVT:
545 		case LINK_FAILOVER_BEGIN_EVT:
546 		case LINK_ESTABLISH_EVT:
547 		case LINK_FAILOVER_END_EVT:
548 		case LINK_SYNCH_BEGIN_EVT:
549 		case LINK_SYNCH_END_EVT:
550 		default:
551 			goto illegal_evt;
552 		}
553 		break;
554 	case LINK_RESET:
555 		switch (evt) {
556 		case LINK_PEER_RESET_EVT:
557 			l->state = LINK_ESTABLISHING;
558 			break;
559 		case LINK_FAILOVER_BEGIN_EVT:
560 			l->state = LINK_FAILINGOVER;
561 		case LINK_FAILURE_EVT:
562 		case LINK_RESET_EVT:
563 		case LINK_ESTABLISH_EVT:
564 		case LINK_FAILOVER_END_EVT:
565 			break;
566 		case LINK_SYNCH_BEGIN_EVT:
567 		case LINK_SYNCH_END_EVT:
568 		default:
569 			goto illegal_evt;
570 		}
571 		break;
572 	case LINK_PEER_RESET:
573 		switch (evt) {
574 		case LINK_RESET_EVT:
575 			l->state = LINK_ESTABLISHING;
576 			break;
577 		case LINK_PEER_RESET_EVT:
578 		case LINK_ESTABLISH_EVT:
579 		case LINK_FAILURE_EVT:
580 			break;
581 		case LINK_SYNCH_BEGIN_EVT:
582 		case LINK_SYNCH_END_EVT:
583 		case LINK_FAILOVER_BEGIN_EVT:
584 		case LINK_FAILOVER_END_EVT:
585 		default:
586 			goto illegal_evt;
587 		}
588 		break;
589 	case LINK_FAILINGOVER:
590 		switch (evt) {
591 		case LINK_FAILOVER_END_EVT:
592 			l->state = LINK_RESET;
593 			break;
594 		case LINK_PEER_RESET_EVT:
595 		case LINK_RESET_EVT:
596 		case LINK_ESTABLISH_EVT:
597 		case LINK_FAILURE_EVT:
598 			break;
599 		case LINK_FAILOVER_BEGIN_EVT:
600 		case LINK_SYNCH_BEGIN_EVT:
601 		case LINK_SYNCH_END_EVT:
602 		default:
603 			goto illegal_evt;
604 		}
605 		break;
606 	case LINK_ESTABLISHING:
607 		switch (evt) {
608 		case LINK_ESTABLISH_EVT:
609 			l->state = LINK_ESTABLISHED;
610 			break;
611 		case LINK_FAILOVER_BEGIN_EVT:
612 			l->state = LINK_FAILINGOVER;
613 			break;
614 		case LINK_RESET_EVT:
615 			l->state = LINK_RESET;
616 			break;
617 		case LINK_FAILURE_EVT:
618 		case LINK_PEER_RESET_EVT:
619 		case LINK_SYNCH_BEGIN_EVT:
620 		case LINK_FAILOVER_END_EVT:
621 			break;
622 		case LINK_SYNCH_END_EVT:
623 		default:
624 			goto illegal_evt;
625 		}
626 		break;
627 	case LINK_ESTABLISHED:
628 		switch (evt) {
629 		case LINK_PEER_RESET_EVT:
630 			l->state = LINK_PEER_RESET;
631 			rc |= TIPC_LINK_DOWN_EVT;
632 			break;
633 		case LINK_FAILURE_EVT:
634 			l->state = LINK_RESETTING;
635 			rc |= TIPC_LINK_DOWN_EVT;
636 			break;
637 		case LINK_RESET_EVT:
638 			l->state = LINK_RESET;
639 			break;
640 		case LINK_ESTABLISH_EVT:
641 		case LINK_SYNCH_END_EVT:
642 			break;
643 		case LINK_SYNCH_BEGIN_EVT:
644 			l->state = LINK_SYNCHING;
645 			break;
646 		case LINK_FAILOVER_BEGIN_EVT:
647 		case LINK_FAILOVER_END_EVT:
648 		default:
649 			goto illegal_evt;
650 		}
651 		break;
652 	case LINK_SYNCHING:
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_BEGIN_EVT:
667 			break;
668 		case LINK_SYNCH_END_EVT:
669 			l->state = LINK_ESTABLISHED;
670 			break;
671 		case LINK_FAILOVER_BEGIN_EVT:
672 		case LINK_FAILOVER_END_EVT:
673 		default:
674 			goto illegal_evt;
675 		}
676 		break;
677 	default:
678 		pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
679 	}
680 	return rc;
681 illegal_evt:
682 	pr_err("Illegal FSM event %x in state %x on link %s\n",
683 	       evt, l->state, l->name);
684 	return rc;
685 }
686 
687 /* link_profile_stats - update statistical profiling of traffic
688  */
link_profile_stats(struct tipc_link * l)689 static void link_profile_stats(struct tipc_link *l)
690 {
691 	struct sk_buff *skb;
692 	struct tipc_msg *msg;
693 	int length;
694 
695 	/* Update counters used in statistical profiling of send traffic */
696 	l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
697 	l->stats.queue_sz_counts++;
698 
699 	skb = skb_peek(&l->transmq);
700 	if (!skb)
701 		return;
702 	msg = buf_msg(skb);
703 	length = msg_size(msg);
704 
705 	if (msg_user(msg) == MSG_FRAGMENTER) {
706 		if (msg_type(msg) != FIRST_FRAGMENT)
707 			return;
708 		length = msg_size(msg_get_wrapped(msg));
709 	}
710 	l->stats.msg_lengths_total += length;
711 	l->stats.msg_length_counts++;
712 	if (length <= 64)
713 		l->stats.msg_length_profile[0]++;
714 	else if (length <= 256)
715 		l->stats.msg_length_profile[1]++;
716 	else if (length <= 1024)
717 		l->stats.msg_length_profile[2]++;
718 	else if (length <= 4096)
719 		l->stats.msg_length_profile[3]++;
720 	else if (length <= 16384)
721 		l->stats.msg_length_profile[4]++;
722 	else if (length <= 32768)
723 		l->stats.msg_length_profile[5]++;
724 	else
725 		l->stats.msg_length_profile[6]++;
726 }
727 
728 /* tipc_link_timeout - perform periodic task as instructed from node timeout
729  */
tipc_link_timeout(struct tipc_link * l,struct sk_buff_head * xmitq)730 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
731 {
732 	int mtyp = 0;
733 	int rc = 0;
734 	bool state = false;
735 	bool probe = false;
736 	bool setup = false;
737 	u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
738 	u16 bc_acked = l->bc_rcvlink->acked;
739 	struct tipc_mon_state *mstate = &l->mon_state;
740 
741 	switch (l->state) {
742 	case LINK_ESTABLISHED:
743 	case LINK_SYNCHING:
744 		mtyp = STATE_MSG;
745 		link_profile_stats(l);
746 		tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
747 		if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
748 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
749 		state = bc_acked != bc_snt;
750 		state |= l->bc_rcvlink->rcv_unacked;
751 		state |= l->rcv_unacked;
752 		state |= !skb_queue_empty(&l->transmq);
753 		state |= !skb_queue_empty(&l->deferdq);
754 		probe = mstate->probing;
755 		probe |= l->silent_intv_cnt;
756 		if (probe || mstate->monitoring)
757 			l->silent_intv_cnt++;
758 		break;
759 	case LINK_RESET:
760 		setup = l->rst_cnt++ <= 4;
761 		setup |= !(l->rst_cnt % 16);
762 		mtyp = RESET_MSG;
763 		break;
764 	case LINK_ESTABLISHING:
765 		setup = true;
766 		mtyp = ACTIVATE_MSG;
767 		break;
768 	case LINK_PEER_RESET:
769 	case LINK_RESETTING:
770 	case LINK_FAILINGOVER:
771 		break;
772 	default:
773 		break;
774 	}
775 
776 	if (state || probe || setup)
777 		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, xmitq);
778 
779 	return rc;
780 }
781 
782 /**
783  * link_schedule_user - schedule a message sender for wakeup after congestion
784  * @l: congested link
785  * @hdr: header of message that is being sent
786  * Create pseudo msg to send back to user when congestion abates
787  */
link_schedule_user(struct tipc_link * l,struct tipc_msg * hdr)788 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
789 {
790 	u32 dnode = tipc_own_addr(l->net);
791 	u32 dport = msg_origport(hdr);
792 	struct sk_buff *skb;
793 
794 	/* Create and schedule wakeup pseudo message */
795 	skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
796 			      dnode, l->addr, dport, 0, 0);
797 	if (!skb)
798 		return -ENOBUFS;
799 	msg_set_dest_droppable(buf_msg(skb), true);
800 	TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
801 	skb_queue_tail(&l->wakeupq, skb);
802 	l->stats.link_congs++;
803 	return -ELINKCONG;
804 }
805 
806 /**
807  * link_prepare_wakeup - prepare users for wakeup after congestion
808  * @l: congested link
809  * Wake up a number of waiting users, as permitted by available space
810  * in the send queue
811  */
link_prepare_wakeup(struct tipc_link * l)812 void link_prepare_wakeup(struct tipc_link *l)
813 {
814 	struct sk_buff_head *wakeupq = &l->wakeupq;
815 	struct sk_buff_head *inputq = l->inputq;
816 	struct sk_buff *skb, *tmp;
817 	struct sk_buff_head tmpq;
818 	int avail[5] = {0,};
819 	int imp = 0;
820 
821 	__skb_queue_head_init(&tmpq);
822 
823 	for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
824 		avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
825 
826 	skb_queue_walk_safe(wakeupq, skb, tmp) {
827 		imp = TIPC_SKB_CB(skb)->chain_imp;
828 		if (avail[imp] <= 0)
829 			continue;
830 		avail[imp]--;
831 		__skb_unlink(skb, wakeupq);
832 		__skb_queue_tail(&tmpq, skb);
833 	}
834 
835 	spin_lock_bh(&inputq->lock);
836 	skb_queue_splice_tail(&tmpq, inputq);
837 	spin_unlock_bh(&inputq->lock);
838 
839 }
840 
tipc_link_reset(struct tipc_link * l)841 void tipc_link_reset(struct tipc_link *l)
842 {
843 	u32 imp;
844 
845 	l->peer_session = ANY_SESSION;
846 	l->session++;
847 	l->mtu = l->advertised_mtu;
848 	__skb_queue_purge(&l->transmq);
849 	__skb_queue_purge(&l->deferdq);
850 	skb_queue_splice_init(&l->wakeupq, l->inputq);
851 	__skb_queue_purge(&l->backlogq);
852 	for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
853 		l->backlog[imp].len = 0;
854 		l->backlog[imp].target_bskb = NULL;
855 	}
856 	kfree_skb(l->reasm_buf);
857 	kfree_skb(l->failover_reasm_skb);
858 	l->reasm_buf = NULL;
859 	l->failover_reasm_skb = NULL;
860 	l->rcv_unacked = 0;
861 	l->snd_nxt = 1;
862 	l->rcv_nxt = 1;
863 	l->acked = 0;
864 	l->silent_intv_cnt = 0;
865 	l->rst_cnt = 0;
866 	l->stale_count = 0;
867 	l->bc_peer_is_up = false;
868 	memset(&l->mon_state, 0, sizeof(l->mon_state));
869 	tipc_link_reset_stats(l);
870 }
871 
872 /**
873  * tipc_link_xmit(): enqueue buffer list according to queue situation
874  * @link: link to use
875  * @list: chain of buffers containing message
876  * @xmitq: returned list of packets to be sent by caller
877  *
878  * Consumes the buffer chain.
879  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
880  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
881  */
tipc_link_xmit(struct tipc_link * l,struct sk_buff_head * list,struct sk_buff_head * xmitq)882 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
883 		   struct sk_buff_head *xmitq)
884 {
885 	struct tipc_msg *hdr = buf_msg(skb_peek(list));
886 	unsigned int maxwin = l->window;
887 	int imp = msg_importance(hdr);
888 	unsigned int mtu = l->mtu;
889 	u16 ack = l->rcv_nxt - 1;
890 	u16 seqno = l->snd_nxt;
891 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
892 	struct sk_buff_head *transmq = &l->transmq;
893 	struct sk_buff_head *backlogq = &l->backlogq;
894 	struct sk_buff *skb, *_skb, **tskb;
895 	int pkt_cnt = skb_queue_len(list);
896 	int rc = 0;
897 
898 	if (unlikely(msg_size(hdr) > mtu)) {
899 		skb_queue_purge(list);
900 		return -EMSGSIZE;
901 	}
902 
903 	/* Allow oversubscription of one data msg per source at congestion */
904 	if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
905 		if (imp == TIPC_SYSTEM_IMPORTANCE) {
906 			pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
907 			return -ENOBUFS;
908 		}
909 		rc = link_schedule_user(l, hdr);
910 	}
911 
912 	if (pkt_cnt > 1) {
913 		l->stats.sent_fragmented++;
914 		l->stats.sent_fragments += pkt_cnt;
915 	}
916 
917 	/* Prepare each packet for sending, and add to relevant queue: */
918 	while (skb_queue_len(list)) {
919 		skb = skb_peek(list);
920 		hdr = buf_msg(skb);
921 		msg_set_seqno(hdr, seqno);
922 		msg_set_ack(hdr, ack);
923 		msg_set_bcast_ack(hdr, bc_ack);
924 
925 		if (likely(skb_queue_len(transmq) < maxwin)) {
926 			_skb = skb_clone(skb, GFP_ATOMIC);
927 			if (!_skb) {
928 				skb_queue_purge(list);
929 				return -ENOBUFS;
930 			}
931 			__skb_dequeue(list);
932 			__skb_queue_tail(transmq, skb);
933 			__skb_queue_tail(xmitq, _skb);
934 			TIPC_SKB_CB(skb)->ackers = l->ackers;
935 			l->rcv_unacked = 0;
936 			l->stats.sent_pkts++;
937 			seqno++;
938 			continue;
939 		}
940 		tskb = &l->backlog[imp].target_bskb;
941 		if (tipc_msg_bundle(*tskb, hdr, mtu)) {
942 			kfree_skb(__skb_dequeue(list));
943 			l->stats.sent_bundled++;
944 			continue;
945 		}
946 		if (tipc_msg_make_bundle(tskb, hdr, mtu, l->addr)) {
947 			kfree_skb(__skb_dequeue(list));
948 			__skb_queue_tail(backlogq, *tskb);
949 			l->backlog[imp].len++;
950 			l->stats.sent_bundled++;
951 			l->stats.sent_bundles++;
952 			continue;
953 		}
954 		l->backlog[imp].target_bskb = NULL;
955 		l->backlog[imp].len += skb_queue_len(list);
956 		skb_queue_splice_tail_init(list, backlogq);
957 	}
958 	l->snd_nxt = seqno;
959 	return rc;
960 }
961 
tipc_link_advance_backlog(struct tipc_link * l,struct sk_buff_head * xmitq)962 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
963 {
964 	struct sk_buff *skb, *_skb;
965 	struct tipc_msg *hdr;
966 	u16 seqno = l->snd_nxt;
967 	u16 ack = l->rcv_nxt - 1;
968 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
969 	u32 imp;
970 
971 	while (skb_queue_len(&l->transmq) < l->window) {
972 		skb = skb_peek(&l->backlogq);
973 		if (!skb)
974 			break;
975 		_skb = skb_clone(skb, GFP_ATOMIC);
976 		if (!_skb)
977 			break;
978 		__skb_dequeue(&l->backlogq);
979 		hdr = buf_msg(skb);
980 		imp = msg_importance(hdr);
981 		l->backlog[imp].len--;
982 		if (unlikely(skb == l->backlog[imp].target_bskb))
983 			l->backlog[imp].target_bskb = NULL;
984 		__skb_queue_tail(&l->transmq, skb);
985 		__skb_queue_tail(xmitq, _skb);
986 		TIPC_SKB_CB(skb)->ackers = l->ackers;
987 		msg_set_seqno(hdr, seqno);
988 		msg_set_ack(hdr, ack);
989 		msg_set_bcast_ack(hdr, bc_ack);
990 		l->rcv_unacked = 0;
991 		l->stats.sent_pkts++;
992 		seqno++;
993 	}
994 	l->snd_nxt = seqno;
995 }
996 
link_retransmit_failure(struct tipc_link * l,struct sk_buff * skb)997 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
998 {
999 	struct tipc_msg *hdr = buf_msg(skb);
1000 
1001 	pr_warn("Retransmission failure on link <%s>\n", l->name);
1002 	link_print(l, "State of link ");
1003 	pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1004 		msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1005 	pr_info("sqno %u, prev: %x, src: %x\n",
1006 		msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
1007 }
1008 
tipc_link_retrans(struct tipc_link * l,struct tipc_link * nacker,u16 from,u16 to,struct sk_buff_head * xmitq)1009 int tipc_link_retrans(struct tipc_link *l, struct tipc_link *nacker,
1010 		      u16 from, u16 to, struct sk_buff_head *xmitq)
1011 {
1012 	struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1013 	struct tipc_msg *hdr;
1014 	u16 ack = l->rcv_nxt - 1;
1015 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1016 
1017 	if (!skb)
1018 		return 0;
1019 
1020 	/* Detect repeated retransmit failures on same packet */
1021 	if (nacker->last_retransm != buf_seqno(skb)) {
1022 		nacker->last_retransm = buf_seqno(skb);
1023 		nacker->stale_count = 1;
1024 	} else if (++nacker->stale_count > 100) {
1025 		link_retransmit_failure(l, skb);
1026 		nacker->stale_count = 0;
1027 		if (link_is_bc_sndlink(l))
1028 			return TIPC_LINK_DOWN_EVT;
1029 		return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1030 	}
1031 
1032 	/* Move forward to where retransmission should start */
1033 	skb_queue_walk(&l->transmq, skb) {
1034 		if (!less(buf_seqno(skb), from))
1035 			break;
1036 	}
1037 
1038 	skb_queue_walk_from(&l->transmq, skb) {
1039 		if (more(buf_seqno(skb), to))
1040 			break;
1041 		hdr = buf_msg(skb);
1042 		_skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1043 		if (!_skb)
1044 			return 0;
1045 		hdr = buf_msg(_skb);
1046 		msg_set_ack(hdr, ack);
1047 		msg_set_bcast_ack(hdr, bc_ack);
1048 		_skb->priority = TC_PRIO_CONTROL;
1049 		__skb_queue_tail(xmitq, _skb);
1050 		l->stats.retransmitted++;
1051 	}
1052 	return 0;
1053 }
1054 
1055 /* tipc_data_input - deliver data and name distr msgs to upper layer
1056  *
1057  * Consumes buffer if message is of right type
1058  * Node lock must be held
1059  */
tipc_data_input(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq)1060 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1061 			    struct sk_buff_head *inputq)
1062 {
1063 	struct tipc_msg *hdr = buf_msg(skb);
1064 
1065 	switch (msg_user(hdr)) {
1066 	case TIPC_LOW_IMPORTANCE:
1067 	case TIPC_MEDIUM_IMPORTANCE:
1068 	case TIPC_HIGH_IMPORTANCE:
1069 	case TIPC_CRITICAL_IMPORTANCE:
1070 		if (unlikely(msg_type(hdr) == TIPC_MCAST_MSG)) {
1071 			skb_queue_tail(l->bc_rcvlink->inputq, skb);
1072 			return true;
1073 		}
1074 	case CONN_MANAGER:
1075 		skb_queue_tail(inputq, skb);
1076 		return true;
1077 	case NAME_DISTRIBUTOR:
1078 		l->bc_rcvlink->state = LINK_ESTABLISHED;
1079 		skb_queue_tail(l->namedq, skb);
1080 		return true;
1081 	case MSG_BUNDLER:
1082 	case TUNNEL_PROTOCOL:
1083 	case MSG_FRAGMENTER:
1084 	case BCAST_PROTOCOL:
1085 		return false;
1086 	default:
1087 		pr_warn("Dropping received illegal msg type\n");
1088 		kfree_skb(skb);
1089 		return true;
1090 	};
1091 }
1092 
1093 /* tipc_link_input - process packet that has passed link protocol check
1094  *
1095  * Consumes buffer
1096  */
tipc_link_input(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq)1097 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1098 			   struct sk_buff_head *inputq)
1099 {
1100 	struct tipc_msg *hdr = buf_msg(skb);
1101 	struct sk_buff **reasm_skb = &l->reasm_buf;
1102 	struct sk_buff *iskb;
1103 	struct sk_buff_head tmpq;
1104 	int usr = msg_user(hdr);
1105 	int rc = 0;
1106 	int pos = 0;
1107 	int ipos = 0;
1108 
1109 	if (unlikely(usr == TUNNEL_PROTOCOL)) {
1110 		if (msg_type(hdr) == SYNCH_MSG) {
1111 			__skb_queue_purge(&l->deferdq);
1112 			goto drop;
1113 		}
1114 		if (!tipc_msg_extract(skb, &iskb, &ipos))
1115 			return rc;
1116 		kfree_skb(skb);
1117 		skb = iskb;
1118 		hdr = buf_msg(skb);
1119 		if (less(msg_seqno(hdr), l->drop_point))
1120 			goto drop;
1121 		if (tipc_data_input(l, skb, inputq))
1122 			return rc;
1123 		usr = msg_user(hdr);
1124 		reasm_skb = &l->failover_reasm_skb;
1125 	}
1126 
1127 	if (usr == MSG_BUNDLER) {
1128 		skb_queue_head_init(&tmpq);
1129 		l->stats.recv_bundles++;
1130 		l->stats.recv_bundled += msg_msgcnt(hdr);
1131 		while (tipc_msg_extract(skb, &iskb, &pos))
1132 			tipc_data_input(l, iskb, &tmpq);
1133 		tipc_skb_queue_splice_tail(&tmpq, inputq);
1134 		return 0;
1135 	} else if (usr == MSG_FRAGMENTER) {
1136 		l->stats.recv_fragments++;
1137 		if (tipc_buf_append(reasm_skb, &skb)) {
1138 			l->stats.recv_fragmented++;
1139 			tipc_data_input(l, skb, inputq);
1140 		} else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1141 			pr_warn_ratelimited("Unable to build fragment list\n");
1142 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1143 		}
1144 		return 0;
1145 	} else if (usr == BCAST_PROTOCOL) {
1146 		tipc_bcast_lock(l->net);
1147 		tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1148 		tipc_bcast_unlock(l->net);
1149 	}
1150 drop:
1151 	kfree_skb(skb);
1152 	return 0;
1153 }
1154 
tipc_link_release_pkts(struct tipc_link * l,u16 acked)1155 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1156 {
1157 	bool released = false;
1158 	struct sk_buff *skb, *tmp;
1159 
1160 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1161 		if (more(buf_seqno(skb), acked))
1162 			break;
1163 		__skb_unlink(skb, &l->transmq);
1164 		kfree_skb(skb);
1165 		released = true;
1166 	}
1167 	return released;
1168 }
1169 
1170 /* tipc_link_build_state_msg: prepare link state message for transmission
1171  *
1172  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1173  * risk of ack storms towards the sender
1174  */
tipc_link_build_state_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1175 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1176 {
1177 	if (!l)
1178 		return 0;
1179 
1180 	/* Broadcast ACK must be sent via a unicast link => defer to caller */
1181 	if (link_is_bc_rcvlink(l)) {
1182 		if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1183 			return 0;
1184 		l->rcv_unacked = 0;
1185 
1186 		/* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1187 		l->snd_nxt = l->rcv_nxt;
1188 		return TIPC_LINK_SND_STATE;
1189 	}
1190 
1191 	/* Unicast ACK */
1192 	l->rcv_unacked = 0;
1193 	l->stats.sent_acks++;
1194 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1195 	return 0;
1196 }
1197 
1198 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1199  */
tipc_link_build_reset_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1200 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1201 {
1202 	int mtyp = RESET_MSG;
1203 	struct sk_buff *skb;
1204 
1205 	if (l->state == LINK_ESTABLISHING)
1206 		mtyp = ACTIVATE_MSG;
1207 
1208 	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1209 
1210 	/* Inform peer that this endpoint is going down if applicable */
1211 	skb = skb_peek_tail(xmitq);
1212 	if (skb && (l->state == LINK_RESET))
1213 		msg_set_peer_stopping(buf_msg(skb), 1);
1214 }
1215 
1216 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1217  * Note that sending of broadcast NACK is coordinated among nodes, to
1218  * reduce the risk of NACK storms towards the sender
1219  */
tipc_link_build_nack_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1220 static int tipc_link_build_nack_msg(struct tipc_link *l,
1221 				    struct sk_buff_head *xmitq)
1222 {
1223 	u32 def_cnt = ++l->stats.deferred_recv;
1224 	int match1, match2;
1225 
1226 	if (link_is_bc_rcvlink(l)) {
1227 		match1 = def_cnt & 0xf;
1228 		match2 = tipc_own_addr(l->net) & 0xf;
1229 		if (match1 == match2)
1230 			return TIPC_LINK_SND_STATE;
1231 		return 0;
1232 	}
1233 
1234 	if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1235 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1236 	return 0;
1237 }
1238 
1239 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1240  * @l: the link that should handle the message
1241  * @skb: TIPC packet
1242  * @xmitq: queue to place packets to be sent after this call
1243  */
tipc_link_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)1244 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1245 		  struct sk_buff_head *xmitq)
1246 {
1247 	struct sk_buff_head *defq = &l->deferdq;
1248 	struct tipc_msg *hdr;
1249 	u16 seqno, rcv_nxt, win_lim;
1250 	int rc = 0;
1251 
1252 	do {
1253 		hdr = buf_msg(skb);
1254 		seqno = msg_seqno(hdr);
1255 		rcv_nxt = l->rcv_nxt;
1256 		win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1257 
1258 		/* Verify and update link state */
1259 		if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1260 			return tipc_link_proto_rcv(l, skb, xmitq);
1261 
1262 		if (unlikely(!link_is_up(l))) {
1263 			if (l->state == LINK_ESTABLISHING)
1264 				rc = TIPC_LINK_UP_EVT;
1265 			goto drop;
1266 		}
1267 
1268 		/* Don't send probe at next timeout expiration */
1269 		l->silent_intv_cnt = 0;
1270 
1271 		/* Drop if outside receive window */
1272 		if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1273 			l->stats.duplicates++;
1274 			goto drop;
1275 		}
1276 
1277 		/* Forward queues and wake up waiting users */
1278 		if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1279 			tipc_link_advance_backlog(l, xmitq);
1280 			if (unlikely(!skb_queue_empty(&l->wakeupq)))
1281 				link_prepare_wakeup(l);
1282 		}
1283 
1284 		/* Defer delivery if sequence gap */
1285 		if (unlikely(seqno != rcv_nxt)) {
1286 			__tipc_skb_queue_sorted(defq, seqno, skb);
1287 			rc |= tipc_link_build_nack_msg(l, xmitq);
1288 			break;
1289 		}
1290 
1291 		/* Deliver packet */
1292 		l->rcv_nxt++;
1293 		l->stats.recv_pkts++;
1294 		if (!tipc_data_input(l, skb, l->inputq))
1295 			rc |= tipc_link_input(l, skb, l->inputq);
1296 		if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1297 			rc |= tipc_link_build_state_msg(l, xmitq);
1298 		if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1299 			break;
1300 	} while ((skb = __skb_dequeue(defq)));
1301 
1302 	return rc;
1303 drop:
1304 	kfree_skb(skb);
1305 	return rc;
1306 }
1307 
tipc_link_build_proto_msg(struct tipc_link * l,int mtyp,bool probe,u16 rcvgap,int tolerance,int priority,struct sk_buff_head * xmitq)1308 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1309 				      u16 rcvgap, int tolerance, int priority,
1310 				      struct sk_buff_head *xmitq)
1311 {
1312 	struct tipc_link *bcl = l->bc_rcvlink;
1313 	struct sk_buff *skb;
1314 	struct tipc_msg *hdr;
1315 	struct sk_buff_head *dfq = &l->deferdq;
1316 	bool node_up = link_is_up(bcl);
1317 	struct tipc_mon_state *mstate = &l->mon_state;
1318 	int dlen = 0;
1319 	void *data;
1320 
1321 	/* Don't send protocol message during reset or link failover */
1322 	if (tipc_link_is_blocked(l))
1323 		return;
1324 
1325 	if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1326 		return;
1327 
1328 	if (!skb_queue_empty(dfq))
1329 		rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1330 
1331 	skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1332 			      tipc_max_domain_size, l->addr,
1333 			      tipc_own_addr(l->net), 0, 0, 0);
1334 	if (!skb)
1335 		return;
1336 
1337 	hdr = buf_msg(skb);
1338 	data = msg_data(hdr);
1339 	msg_set_session(hdr, l->session);
1340 	msg_set_bearer_id(hdr, l->bearer_id);
1341 	msg_set_net_plane(hdr, l->net_plane);
1342 	msg_set_next_sent(hdr, l->snd_nxt);
1343 	msg_set_ack(hdr, l->rcv_nxt - 1);
1344 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1345 	msg_set_bc_ack_invalid(hdr, !node_up);
1346 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1347 	msg_set_link_tolerance(hdr, tolerance);
1348 	msg_set_linkprio(hdr, priority);
1349 	msg_set_redundant_link(hdr, node_up);
1350 	msg_set_seq_gap(hdr, 0);
1351 	msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1352 
1353 	if (mtyp == STATE_MSG) {
1354 		msg_set_seq_gap(hdr, rcvgap);
1355 		msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1356 		msg_set_probe(hdr, probe);
1357 		tipc_mon_prep(l->net, data, &dlen, mstate, l->bearer_id);
1358 		msg_set_size(hdr, INT_H_SIZE + dlen);
1359 		skb_trim(skb, INT_H_SIZE + dlen);
1360 		l->stats.sent_states++;
1361 		l->rcv_unacked = 0;
1362 	} else {
1363 		/* RESET_MSG or ACTIVATE_MSG */
1364 		msg_set_max_pkt(hdr, l->advertised_mtu);
1365 		strcpy(data, l->if_name);
1366 		msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1367 		skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1368 	}
1369 	if (probe)
1370 		l->stats.sent_probes++;
1371 	if (rcvgap)
1372 		l->stats.sent_nacks++;
1373 	skb->priority = TC_PRIO_CONTROL;
1374 	__skb_queue_tail(xmitq, skb);
1375 }
1376 
1377 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1378  * with contents of the link's transmit and backlog queues.
1379  */
tipc_link_tnl_prepare(struct tipc_link * l,struct tipc_link * tnl,int mtyp,struct sk_buff_head * xmitq)1380 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1381 			   int mtyp, struct sk_buff_head *xmitq)
1382 {
1383 	struct sk_buff *skb, *tnlskb;
1384 	struct tipc_msg *hdr, tnlhdr;
1385 	struct sk_buff_head *queue = &l->transmq;
1386 	struct sk_buff_head tmpxq, tnlq;
1387 	u16 pktlen, pktcnt, seqno = l->snd_nxt;
1388 
1389 	if (!tnl)
1390 		return;
1391 
1392 	skb_queue_head_init(&tnlq);
1393 	skb_queue_head_init(&tmpxq);
1394 
1395 	/* At least one packet required for safe algorithm => add dummy */
1396 	skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1397 			      BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1398 			      0, 0, TIPC_ERR_NO_PORT);
1399 	if (!skb) {
1400 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1401 		return;
1402 	}
1403 	skb_queue_tail(&tnlq, skb);
1404 	tipc_link_xmit(l, &tnlq, &tmpxq);
1405 	__skb_queue_purge(&tmpxq);
1406 
1407 	/* Initialize reusable tunnel packet header */
1408 	tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1409 		      mtyp, INT_H_SIZE, l->addr);
1410 	pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1411 	msg_set_msgcnt(&tnlhdr, pktcnt);
1412 	msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1413 tnl:
1414 	/* Wrap each packet into a tunnel packet */
1415 	skb_queue_walk(queue, skb) {
1416 		hdr = buf_msg(skb);
1417 		if (queue == &l->backlogq)
1418 			msg_set_seqno(hdr, seqno++);
1419 		pktlen = msg_size(hdr);
1420 		msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1421 		tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1422 		if (!tnlskb) {
1423 			pr_warn("%sunable to send packet\n", link_co_err);
1424 			return;
1425 		}
1426 		skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1427 		skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1428 		__skb_queue_tail(&tnlq, tnlskb);
1429 	}
1430 	if (queue != &l->backlogq) {
1431 		queue = &l->backlogq;
1432 		goto tnl;
1433 	}
1434 
1435 	tipc_link_xmit(tnl, &tnlq, xmitq);
1436 
1437 	if (mtyp == FAILOVER_MSG) {
1438 		tnl->drop_point = l->rcv_nxt;
1439 		tnl->failover_reasm_skb = l->reasm_buf;
1440 		l->reasm_buf = NULL;
1441 	}
1442 }
1443 
1444 /* tipc_link_proto_rcv(): receive link level protocol message :
1445  * Note that network plane id propagates through the network, and may
1446  * change at any time. The node with lowest numerical id determines
1447  * network plane
1448  */
tipc_link_proto_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)1449 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1450 			       struct sk_buff_head *xmitq)
1451 {
1452 	struct tipc_msg *hdr = buf_msg(skb);
1453 	u16 rcvgap = 0;
1454 	u16 ack = msg_ack(hdr);
1455 	u16 gap = msg_seq_gap(hdr);
1456 	u16 peers_snd_nxt =  msg_next_sent(hdr);
1457 	u16 peers_tol = msg_link_tolerance(hdr);
1458 	u16 peers_prio = msg_linkprio(hdr);
1459 	u16 rcv_nxt = l->rcv_nxt;
1460 	u16 dlen = msg_data_sz(hdr);
1461 	int mtyp = msg_type(hdr);
1462 	void *data;
1463 	char *if_name;
1464 	int rc = 0;
1465 
1466 	if (tipc_link_is_blocked(l) || !xmitq)
1467 		goto exit;
1468 
1469 	if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1470 		l->net_plane = msg_net_plane(hdr);
1471 
1472 	skb_linearize(skb);
1473 	hdr = buf_msg(skb);
1474 	data = msg_data(hdr);
1475 
1476 	switch (mtyp) {
1477 	case RESET_MSG:
1478 
1479 		/* Ignore duplicate RESET with old session number */
1480 		if ((less_eq(msg_session(hdr), l->peer_session)) &&
1481 		    (l->peer_session != ANY_SESSION))
1482 			break;
1483 		/* fall thru' */
1484 
1485 	case ACTIVATE_MSG:
1486 
1487 		/* Complete own link name with peer's interface name */
1488 		if_name =  strrchr(l->name, ':') + 1;
1489 		if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1490 			break;
1491 		if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1492 			break;
1493 		strncpy(if_name, data, TIPC_MAX_IF_NAME);
1494 
1495 		/* Update own tolerance if peer indicates a non-zero value */
1496 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1497 			l->tolerance = peers_tol;
1498 
1499 		/* Update own priority if peer's priority is higher */
1500 		if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1501 			l->priority = peers_prio;
1502 
1503 		/* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1504 		if (msg_peer_stopping(hdr))
1505 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1506 		else if ((mtyp == RESET_MSG) || !link_is_up(l))
1507 			rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1508 
1509 		/* ACTIVATE_MSG takes up link if it was already locally reset */
1510 		if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1511 			rc = TIPC_LINK_UP_EVT;
1512 
1513 		l->peer_session = msg_session(hdr);
1514 		l->peer_bearer_id = msg_bearer_id(hdr);
1515 		if (l->mtu > msg_max_pkt(hdr))
1516 			l->mtu = msg_max_pkt(hdr);
1517 		break;
1518 
1519 	case STATE_MSG:
1520 
1521 		/* Update own tolerance if peer indicates a non-zero value */
1522 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1523 			l->tolerance = peers_tol;
1524 
1525 		/* Update own prio if peer indicates a different value */
1526 		if ((peers_prio != l->priority) &&
1527 		    in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
1528 			l->priority = peers_prio;
1529 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1530 		}
1531 
1532 		l->silent_intv_cnt = 0;
1533 		l->stats.recv_states++;
1534 		if (msg_probe(hdr))
1535 			l->stats.recv_probes++;
1536 
1537 		if (!link_is_up(l)) {
1538 			if (l->state == LINK_ESTABLISHING)
1539 				rc = TIPC_LINK_UP_EVT;
1540 			break;
1541 		}
1542 		tipc_mon_rcv(l->net, data, dlen, l->addr,
1543 			     &l->mon_state, l->bearer_id);
1544 
1545 		/* Send NACK if peer has sent pkts we haven't received yet */
1546 		if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1547 			rcvgap = peers_snd_nxt - l->rcv_nxt;
1548 		if (rcvgap || (msg_probe(hdr)))
1549 			tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1550 						  0, 0, xmitq);
1551 		tipc_link_release_pkts(l, ack);
1552 
1553 		/* If NACK, retransmit will now start at right position */
1554 		if (gap) {
1555 			rc = tipc_link_retrans(l, l, ack + 1, ack + gap, xmitq);
1556 			l->stats.recv_nacks++;
1557 		}
1558 
1559 		tipc_link_advance_backlog(l, xmitq);
1560 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
1561 			link_prepare_wakeup(l);
1562 	}
1563 exit:
1564 	kfree_skb(skb);
1565 	return rc;
1566 }
1567 
1568 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1569  */
tipc_link_build_bc_proto_msg(struct tipc_link * l,bool bcast,u16 peers_snd_nxt,struct sk_buff_head * xmitq)1570 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1571 					 u16 peers_snd_nxt,
1572 					 struct sk_buff_head *xmitq)
1573 {
1574 	struct sk_buff *skb;
1575 	struct tipc_msg *hdr;
1576 	struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1577 	u16 ack = l->rcv_nxt - 1;
1578 	u16 gap_to = peers_snd_nxt - 1;
1579 
1580 	skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1581 			      0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1582 	if (!skb)
1583 		return false;
1584 	hdr = buf_msg(skb);
1585 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1586 	msg_set_bcast_ack(hdr, ack);
1587 	msg_set_bcgap_after(hdr, ack);
1588 	if (dfrd_skb)
1589 		gap_to = buf_seqno(dfrd_skb) - 1;
1590 	msg_set_bcgap_to(hdr, gap_to);
1591 	msg_set_non_seq(hdr, bcast);
1592 	__skb_queue_tail(xmitq, skb);
1593 	return true;
1594 }
1595 
1596 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1597  *
1598  * Give a newly added peer node the sequence number where it should
1599  * start receiving and acking broadcast packets.
1600  */
tipc_link_build_bc_init_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1601 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1602 					struct sk_buff_head *xmitq)
1603 {
1604 	struct sk_buff_head list;
1605 
1606 	__skb_queue_head_init(&list);
1607 	if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1608 		return;
1609 	msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
1610 	tipc_link_xmit(l, &list, xmitq);
1611 }
1612 
1613 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1614  */
tipc_link_bc_init_rcv(struct tipc_link * l,struct tipc_msg * hdr)1615 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1616 {
1617 	int mtyp = msg_type(hdr);
1618 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1619 
1620 	if (link_is_up(l))
1621 		return;
1622 
1623 	if (msg_user(hdr) == BCAST_PROTOCOL) {
1624 		l->rcv_nxt = peers_snd_nxt;
1625 		l->state = LINK_ESTABLISHED;
1626 		return;
1627 	}
1628 
1629 	if (l->peer_caps & TIPC_BCAST_SYNCH)
1630 		return;
1631 
1632 	if (msg_peer_node_is_up(hdr))
1633 		return;
1634 
1635 	/* Compatibility: accept older, less safe initial synch data */
1636 	if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1637 		l->rcv_nxt = peers_snd_nxt;
1638 }
1639 
1640 /* link_bc_retr eval()- check if the indicated range can be retransmitted now
1641  * - Adjust permitted range if there is overlap with previous retransmission
1642  */
link_bc_retr_eval(struct tipc_link * l,u16 * from,u16 * to)1643 static bool link_bc_retr_eval(struct tipc_link *l, u16 *from, u16 *to)
1644 {
1645 	unsigned long elapsed = jiffies_to_msecs(jiffies - l->prev_retr);
1646 
1647 	if (less(*to, *from))
1648 		return false;
1649 
1650 	/* New retransmission request */
1651 	if ((elapsed > TIPC_BC_RETR_LIMIT) ||
1652 	    less(*to, l->prev_from) || more(*from, l->prev_to)) {
1653 		l->prev_from = *from;
1654 		l->prev_to = *to;
1655 		l->prev_retr = jiffies;
1656 		return true;
1657 	}
1658 
1659 	/* Inside range of previous retransmit */
1660 	if (!less(*from, l->prev_from) && !more(*to, l->prev_to))
1661 		return false;
1662 
1663 	/* Fully or partially outside previous range => exclude overlap */
1664 	if (less(*from, l->prev_from)) {
1665 		*to = l->prev_from - 1;
1666 		l->prev_from = *from;
1667 	}
1668 	if (more(*to, l->prev_to)) {
1669 		*from = l->prev_to + 1;
1670 		l->prev_to = *to;
1671 	}
1672 	l->prev_retr = jiffies;
1673 	return true;
1674 }
1675 
1676 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1677  */
tipc_link_bc_sync_rcv(struct tipc_link * l,struct tipc_msg * hdr,struct sk_buff_head * xmitq)1678 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1679 			  struct sk_buff_head *xmitq)
1680 {
1681 	struct tipc_link *snd_l = l->bc_sndlink;
1682 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1683 	u16 from = msg_bcast_ack(hdr) + 1;
1684 	u16 to = from + msg_bc_gap(hdr) - 1;
1685 	int rc = 0;
1686 
1687 	if (!link_is_up(l))
1688 		return rc;
1689 
1690 	if (!msg_peer_node_is_up(hdr))
1691 		return rc;
1692 
1693 	/* Open when peer ackowledges our bcast init msg (pkt #1) */
1694 	if (msg_ack(hdr))
1695 		l->bc_peer_is_up = true;
1696 
1697 	if (!l->bc_peer_is_up)
1698 		return rc;
1699 
1700 	l->stats.recv_nacks++;
1701 
1702 	/* Ignore if peers_snd_nxt goes beyond receive window */
1703 	if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1704 		return rc;
1705 
1706 	if (link_bc_retr_eval(snd_l, &from, &to))
1707 		rc = tipc_link_retrans(snd_l, l, from, to, xmitq);
1708 
1709 	l->snd_nxt = peers_snd_nxt;
1710 	if (link_bc_rcv_gap(l))
1711 		rc |= TIPC_LINK_SND_STATE;
1712 
1713 	/* Return now if sender supports nack via STATE messages */
1714 	if (l->peer_caps & TIPC_BCAST_STATE_NACK)
1715 		return rc;
1716 
1717 	/* Otherwise, be backwards compatible */
1718 
1719 	if (!more(peers_snd_nxt, l->rcv_nxt)) {
1720 		l->nack_state = BC_NACK_SND_CONDITIONAL;
1721 		return 0;
1722 	}
1723 
1724 	/* Don't NACK if one was recently sent or peeked */
1725 	if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1726 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1727 		return 0;
1728 	}
1729 
1730 	/* Conditionally delay NACK sending until next synch rcv */
1731 	if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1732 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1733 		if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1734 			return 0;
1735 	}
1736 
1737 	/* Send NACK now but suppress next one */
1738 	tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1739 	l->nack_state = BC_NACK_SND_SUPPRESS;
1740 	return 0;
1741 }
1742 
tipc_link_bc_ack_rcv(struct tipc_link * l,u16 acked,struct sk_buff_head * xmitq)1743 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1744 			  struct sk_buff_head *xmitq)
1745 {
1746 	struct sk_buff *skb, *tmp;
1747 	struct tipc_link *snd_l = l->bc_sndlink;
1748 
1749 	if (!link_is_up(l) || !l->bc_peer_is_up)
1750 		return;
1751 
1752 	if (!more(acked, l->acked))
1753 		return;
1754 
1755 	/* Skip over packets peer has already acked */
1756 	skb_queue_walk(&snd_l->transmq, skb) {
1757 		if (more(buf_seqno(skb), l->acked))
1758 			break;
1759 	}
1760 
1761 	/* Update/release the packets peer is acking now */
1762 	skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1763 		if (more(buf_seqno(skb), acked))
1764 			break;
1765 		if (!--TIPC_SKB_CB(skb)->ackers) {
1766 			__skb_unlink(skb, &snd_l->transmq);
1767 			kfree_skb(skb);
1768 		}
1769 	}
1770 	l->acked = acked;
1771 	tipc_link_advance_backlog(snd_l, xmitq);
1772 	if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1773 		link_prepare_wakeup(snd_l);
1774 }
1775 
1776 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1777  * This function is here for backwards compatibility, since
1778  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
1779  */
tipc_link_bc_nack_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)1780 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1781 			  struct sk_buff_head *xmitq)
1782 {
1783 	struct tipc_msg *hdr = buf_msg(skb);
1784 	u32 dnode = msg_destnode(hdr);
1785 	int mtyp = msg_type(hdr);
1786 	u16 acked = msg_bcast_ack(hdr);
1787 	u16 from = acked + 1;
1788 	u16 to = msg_bcgap_to(hdr);
1789 	u16 peers_snd_nxt = to + 1;
1790 	int rc = 0;
1791 
1792 	kfree_skb(skb);
1793 
1794 	if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1795 		return 0;
1796 
1797 	if (mtyp != STATE_MSG)
1798 		return 0;
1799 
1800 	if (dnode == tipc_own_addr(l->net)) {
1801 		tipc_link_bc_ack_rcv(l, acked, xmitq);
1802 		rc = tipc_link_retrans(l->bc_sndlink, l, from, to, xmitq);
1803 		l->stats.recv_nacks++;
1804 		return rc;
1805 	}
1806 
1807 	/* Msg for other node => suppress own NACK at next sync if applicable */
1808 	if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1809 		l->nack_state = BC_NACK_SND_SUPPRESS;
1810 
1811 	return 0;
1812 }
1813 
tipc_link_set_queue_limits(struct tipc_link * l,u32 win)1814 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1815 {
1816 	int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1817 
1818 	l->window = win;
1819 	l->backlog[TIPC_LOW_IMPORTANCE].limit      = max_t(u16, 50, win);
1820 	l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = max_t(u16, 100, win * 2);
1821 	l->backlog[TIPC_HIGH_IMPORTANCE].limit     = max_t(u16, 150, win * 3);
1822 	l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
1823 	l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
1824 }
1825 
1826 /**
1827  * link_reset_stats - reset link statistics
1828  * @l: pointer to link
1829  */
tipc_link_reset_stats(struct tipc_link * l)1830 void tipc_link_reset_stats(struct tipc_link *l)
1831 {
1832 	memset(&l->stats, 0, sizeof(l->stats));
1833 }
1834 
link_print(struct tipc_link * l,const char * str)1835 static void link_print(struct tipc_link *l, const char *str)
1836 {
1837 	struct sk_buff *hskb = skb_peek(&l->transmq);
1838 	u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1839 	u16 tail = l->snd_nxt - 1;
1840 
1841 	pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1842 	pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1843 		skb_queue_len(&l->transmq), head, tail,
1844 		skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1845 }
1846 
1847 /* Parse and validate nested (link) properties valid for media, bearer and link
1848  */
tipc_nl_parse_link_prop(struct nlattr * prop,struct nlattr * props[])1849 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1850 {
1851 	int err;
1852 
1853 	err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1854 			       tipc_nl_prop_policy, NULL);
1855 	if (err)
1856 		return err;
1857 
1858 	if (props[TIPC_NLA_PROP_PRIO]) {
1859 		u32 prio;
1860 
1861 		prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1862 		if (prio > TIPC_MAX_LINK_PRI)
1863 			return -EINVAL;
1864 	}
1865 
1866 	if (props[TIPC_NLA_PROP_TOL]) {
1867 		u32 tol;
1868 
1869 		tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1870 		if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1871 			return -EINVAL;
1872 	}
1873 
1874 	if (props[TIPC_NLA_PROP_WIN]) {
1875 		u32 win;
1876 
1877 		win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1878 		if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1879 			return -EINVAL;
1880 	}
1881 
1882 	return 0;
1883 }
1884 
__tipc_nl_add_stats(struct sk_buff * skb,struct tipc_stats * s)1885 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1886 {
1887 	int i;
1888 	struct nlattr *stats;
1889 
1890 	struct nla_map {
1891 		u32 key;
1892 		u32 val;
1893 	};
1894 
1895 	struct nla_map map[] = {
1896 		{TIPC_NLA_STATS_RX_INFO, 0},
1897 		{TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1898 		{TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1899 		{TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1900 		{TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1901 		{TIPC_NLA_STATS_TX_INFO, 0},
1902 		{TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1903 		{TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1904 		{TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1905 		{TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1906 		{TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1907 			s->msg_length_counts : 1},
1908 		{TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1909 		{TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1910 		{TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1911 		{TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1912 		{TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1913 		{TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1914 		{TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1915 		{TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1916 		{TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1917 		{TIPC_NLA_STATS_RX_STATES, s->recv_states},
1918 		{TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1919 		{TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1920 		{TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1921 		{TIPC_NLA_STATS_TX_STATES, s->sent_states},
1922 		{TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1923 		{TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1924 		{TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1925 		{TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1926 		{TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1927 		{TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1928 		{TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1929 		{TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1930 			(s->accu_queue_sz / s->queue_sz_counts) : 0}
1931 	};
1932 
1933 	stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1934 	if (!stats)
1935 		return -EMSGSIZE;
1936 
1937 	for (i = 0; i <  ARRAY_SIZE(map); i++)
1938 		if (nla_put_u32(skb, map[i].key, map[i].val))
1939 			goto msg_full;
1940 
1941 	nla_nest_end(skb, stats);
1942 
1943 	return 0;
1944 msg_full:
1945 	nla_nest_cancel(skb, stats);
1946 
1947 	return -EMSGSIZE;
1948 }
1949 
1950 /* Caller should hold appropriate locks to protect the link */
__tipc_nl_add_link(struct net * net,struct tipc_nl_msg * msg,struct tipc_link * link,int nlflags)1951 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1952 		       struct tipc_link *link, int nlflags)
1953 {
1954 	int err;
1955 	void *hdr;
1956 	struct nlattr *attrs;
1957 	struct nlattr *prop;
1958 	struct tipc_net *tn = net_generic(net, tipc_net_id);
1959 
1960 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1961 			  nlflags, TIPC_NL_LINK_GET);
1962 	if (!hdr)
1963 		return -EMSGSIZE;
1964 
1965 	attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1966 	if (!attrs)
1967 		goto msg_full;
1968 
1969 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1970 		goto attr_msg_full;
1971 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1972 			tipc_cluster_mask(tn->own_addr)))
1973 		goto attr_msg_full;
1974 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1975 		goto attr_msg_full;
1976 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
1977 		goto attr_msg_full;
1978 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
1979 		goto attr_msg_full;
1980 
1981 	if (tipc_link_is_up(link))
1982 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1983 			goto attr_msg_full;
1984 	if (link->active)
1985 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
1986 			goto attr_msg_full;
1987 
1988 	prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1989 	if (!prop)
1990 		goto attr_msg_full;
1991 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1992 		goto prop_msg_full;
1993 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
1994 		goto prop_msg_full;
1995 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
1996 			link->window))
1997 		goto prop_msg_full;
1998 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1999 		goto prop_msg_full;
2000 	nla_nest_end(msg->skb, prop);
2001 
2002 	err = __tipc_nl_add_stats(msg->skb, &link->stats);
2003 	if (err)
2004 		goto attr_msg_full;
2005 
2006 	nla_nest_end(msg->skb, attrs);
2007 	genlmsg_end(msg->skb, hdr);
2008 
2009 	return 0;
2010 
2011 prop_msg_full:
2012 	nla_nest_cancel(msg->skb, prop);
2013 attr_msg_full:
2014 	nla_nest_cancel(msg->skb, attrs);
2015 msg_full:
2016 	genlmsg_cancel(msg->skb, hdr);
2017 
2018 	return -EMSGSIZE;
2019 }
2020 
__tipc_nl_add_bc_link_stat(struct sk_buff * skb,struct tipc_stats * stats)2021 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2022 				      struct tipc_stats *stats)
2023 {
2024 	int i;
2025 	struct nlattr *nest;
2026 
2027 	struct nla_map {
2028 		__u32 key;
2029 		__u32 val;
2030 	};
2031 
2032 	struct nla_map map[] = {
2033 		{TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2034 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2035 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2036 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2037 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2038 		{TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2039 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2040 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2041 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2042 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2043 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2044 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2045 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2046 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2047 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2048 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2049 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2050 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2051 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2052 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2053 	};
2054 
2055 	nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
2056 	if (!nest)
2057 		return -EMSGSIZE;
2058 
2059 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2060 		if (nla_put_u32(skb, map[i].key, map[i].val))
2061 			goto msg_full;
2062 
2063 	nla_nest_end(skb, nest);
2064 
2065 	return 0;
2066 msg_full:
2067 	nla_nest_cancel(skb, nest);
2068 
2069 	return -EMSGSIZE;
2070 }
2071 
tipc_nl_add_bc_link(struct net * net,struct tipc_nl_msg * msg)2072 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2073 {
2074 	int err;
2075 	void *hdr;
2076 	struct nlattr *attrs;
2077 	struct nlattr *prop;
2078 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2079 	struct tipc_link *bcl = tn->bcl;
2080 
2081 	if (!bcl)
2082 		return 0;
2083 
2084 	tipc_bcast_lock(net);
2085 
2086 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2087 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
2088 	if (!hdr) {
2089 		tipc_bcast_unlock(net);
2090 		return -EMSGSIZE;
2091 	}
2092 
2093 	attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
2094 	if (!attrs)
2095 		goto msg_full;
2096 
2097 	/* The broadcast link is always up */
2098 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2099 		goto attr_msg_full;
2100 
2101 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2102 		goto attr_msg_full;
2103 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2104 		goto attr_msg_full;
2105 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2106 		goto attr_msg_full;
2107 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2108 		goto attr_msg_full;
2109 
2110 	prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2111 	if (!prop)
2112 		goto attr_msg_full;
2113 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2114 		goto prop_msg_full;
2115 	nla_nest_end(msg->skb, prop);
2116 
2117 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2118 	if (err)
2119 		goto attr_msg_full;
2120 
2121 	tipc_bcast_unlock(net);
2122 	nla_nest_end(msg->skb, attrs);
2123 	genlmsg_end(msg->skb, hdr);
2124 
2125 	return 0;
2126 
2127 prop_msg_full:
2128 	nla_nest_cancel(msg->skb, prop);
2129 attr_msg_full:
2130 	nla_nest_cancel(msg->skb, attrs);
2131 msg_full:
2132 	tipc_bcast_unlock(net);
2133 	genlmsg_cancel(msg->skb, hdr);
2134 
2135 	return -EMSGSIZE;
2136 }
2137 
tipc_link_set_tolerance(struct tipc_link * l,u32 tol,struct sk_buff_head * xmitq)2138 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2139 			     struct sk_buff_head *xmitq)
2140 {
2141 	l->tolerance = tol;
2142 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, tol, 0, xmitq);
2143 }
2144 
tipc_link_set_prio(struct tipc_link * l,u32 prio,struct sk_buff_head * xmitq)2145 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2146 			struct sk_buff_head *xmitq)
2147 {
2148 	l->priority = prio;
2149 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, prio, xmitq);
2150 }
2151 
tipc_link_set_abort_limit(struct tipc_link * l,u32 limit)2152 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2153 {
2154 	l->abort_limit = limit;
2155 }
2156