1 /*
2 * net/tipc/link.c: TIPC link code
3 *
4 * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5 * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
19 *
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
35 */
36
37 #include "core.h"
38 #include "subscr.h"
39 #include "link.h"
40 #include "bcast.h"
41 #include "socket.h"
42 #include "name_distr.h"
43 #include "discover.h"
44 #include "netlink.h"
45 #include "monitor.h"
46 #include "trace.h"
47
48 #include <linux/pkt_sched.h>
49
50 struct tipc_stats {
51 u32 sent_pkts;
52 u32 recv_pkts;
53 u32 sent_states;
54 u32 recv_states;
55 u32 sent_probes;
56 u32 recv_probes;
57 u32 sent_nacks;
58 u32 recv_nacks;
59 u32 sent_acks;
60 u32 sent_bundled;
61 u32 sent_bundles;
62 u32 recv_bundled;
63 u32 recv_bundles;
64 u32 retransmitted;
65 u32 sent_fragmented;
66 u32 sent_fragments;
67 u32 recv_fragmented;
68 u32 recv_fragments;
69 u32 link_congs; /* # port sends blocked by congestion */
70 u32 deferred_recv;
71 u32 duplicates;
72 u32 max_queue_sz; /* send queue size high water mark */
73 u32 accu_queue_sz; /* used for send queue size profiling */
74 u32 queue_sz_counts; /* used for send queue size profiling */
75 u32 msg_length_counts; /* used for message length profiling */
76 u32 msg_lengths_total; /* used for message length profiling */
77 u32 msg_length_profile[7]; /* used for msg. length profiling */
78 };
79
80 /**
81 * struct tipc_link - TIPC link data structure
82 * @addr: network address of link's peer node
83 * @name: link name character string
84 * @media_addr: media address to use when sending messages over link
85 * @timer: link timer
86 * @net: pointer to namespace struct
87 * @refcnt: reference counter for permanent references (owner node & timer)
88 * @peer_session: link session # being used by peer end of link
89 * @peer_bearer_id: bearer id used by link's peer endpoint
90 * @bearer_id: local bearer id used by link
91 * @tolerance: minimum link continuity loss needed to reset link [in ms]
92 * @abort_limit: # of unacknowledged continuity probes needed to reset link
93 * @state: current state of link FSM
94 * @peer_caps: bitmap describing capabilities of peer node
95 * @silent_intv_cnt: # of timer intervals without any reception from peer
96 * @proto_msg: template for control messages generated by link
97 * @pmsg: convenience pointer to "proto_msg" field
98 * @priority: current link priority
99 * @net_plane: current link network plane ('A' through 'H')
100 * @mon_state: cookie with information needed by link monitor
101 * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
102 * @exp_msg_count: # of tunnelled messages expected during link changeover
103 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
104 * @mtu: current maximum packet size for this link
105 * @advertised_mtu: advertised own mtu when link is being established
106 * @transmitq: queue for sent, non-acked messages
107 * @backlogq: queue for messages waiting to be sent
108 * @snt_nxt: next sequence number to use for outbound messages
109 * @ackers: # of peers that needs to ack each packet before it can be released
110 * @acked: # last packet acked by a certain peer. Used for broadcast.
111 * @rcv_nxt: next sequence number to expect for inbound messages
112 * @deferred_queue: deferred queue saved OOS b'cast message received from node
113 * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
114 * @inputq: buffer queue for messages to be delivered upwards
115 * @namedq: buffer queue for name table messages to be delivered upwards
116 * @next_out: ptr to first unsent outbound message in queue
117 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
118 * @long_msg_seq_no: next identifier to use for outbound fragmented messages
119 * @reasm_buf: head of partially reassembled inbound message fragments
120 * @bc_rcvr: marks that this is a broadcast receiver link
121 * @stats: collects statistics regarding link activity
122 */
123 struct tipc_link {
124 u32 addr;
125 char name[TIPC_MAX_LINK_NAME];
126 struct net *net;
127
128 /* Management and link supervision data */
129 u16 peer_session;
130 u16 session;
131 u16 snd_nxt_state;
132 u16 rcv_nxt_state;
133 u32 peer_bearer_id;
134 u32 bearer_id;
135 u32 tolerance;
136 u32 abort_limit;
137 u32 state;
138 u16 peer_caps;
139 bool in_session;
140 bool active;
141 u32 silent_intv_cnt;
142 char if_name[TIPC_MAX_IF_NAME];
143 u32 priority;
144 char net_plane;
145 struct tipc_mon_state mon_state;
146 u16 rst_cnt;
147
148 /* Failover/synch */
149 u16 drop_point;
150 struct sk_buff *failover_reasm_skb;
151 struct sk_buff_head failover_deferdq;
152
153 /* Max packet negotiation */
154 u16 mtu;
155 u16 advertised_mtu;
156
157 /* Sending */
158 struct sk_buff_head transmq;
159 struct sk_buff_head backlogq;
160 struct {
161 u16 len;
162 u16 limit;
163 struct sk_buff *target_bskb;
164 } backlog[5];
165 u16 snd_nxt;
166 u16 window;
167
168 /* Reception */
169 u16 rcv_nxt;
170 u32 rcv_unacked;
171 struct sk_buff_head deferdq;
172 struct sk_buff_head *inputq;
173 struct sk_buff_head *namedq;
174
175 /* Congestion handling */
176 struct sk_buff_head wakeupq;
177
178 /* Fragmentation/reassembly */
179 struct sk_buff *reasm_buf;
180 struct sk_buff *reasm_tnlmsg;
181
182 /* Broadcast */
183 u16 ackers;
184 u16 acked;
185 struct tipc_link *bc_rcvlink;
186 struct tipc_link *bc_sndlink;
187 u8 nack_state;
188 bool bc_peer_is_up;
189
190 /* Statistics */
191 struct tipc_stats stats;
192 };
193
194 /*
195 * Error message prefixes
196 */
197 static const char *link_co_err = "Link tunneling error, ";
198 static const char *link_rst_msg = "Resetting link ";
199
200 /* Send states for broadcast NACKs
201 */
202 enum {
203 BC_NACK_SND_CONDITIONAL,
204 BC_NACK_SND_UNCONDITIONAL,
205 BC_NACK_SND_SUPPRESS,
206 };
207
208 #define TIPC_BC_RETR_LIM (jiffies + msecs_to_jiffies(10))
209 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
210
211 /*
212 * Interval between NACKs when packets arrive out of order
213 */
214 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
215
216 /* Link FSM states:
217 */
218 enum {
219 LINK_ESTABLISHED = 0xe,
220 LINK_ESTABLISHING = 0xe << 4,
221 LINK_RESET = 0x1 << 8,
222 LINK_RESETTING = 0x2 << 12,
223 LINK_PEER_RESET = 0xd << 16,
224 LINK_FAILINGOVER = 0xf << 20,
225 LINK_SYNCHING = 0xc << 24
226 };
227
228 /* Link FSM state checking routines
229 */
link_is_up(struct tipc_link * l)230 static int link_is_up(struct tipc_link *l)
231 {
232 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
233 }
234
235 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
236 struct sk_buff_head *xmitq);
237 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
238 bool probe_reply, u16 rcvgap,
239 int tolerance, int priority,
240 struct sk_buff_head *xmitq);
241 static void link_print(struct tipc_link *l, const char *str);
242 static int tipc_link_build_nack_msg(struct tipc_link *l,
243 struct sk_buff_head *xmitq);
244 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
245 struct sk_buff_head *xmitq);
246 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
247 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data);
248 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
249 struct tipc_gap_ack_blks *ga,
250 struct sk_buff_head *xmitq);
251
252 /*
253 * Simple non-static link routines (i.e. referenced outside this file)
254 */
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_set_active(struct tipc_link * l,bool active)300 void tipc_link_set_active(struct tipc_link *l, bool active)
301 {
302 l->active = active;
303 }
304
tipc_link_id(struct tipc_link * l)305 u32 tipc_link_id(struct tipc_link *l)
306 {
307 return l->peer_bearer_id << 16 | l->bearer_id;
308 }
309
tipc_link_window(struct tipc_link * l)310 int tipc_link_window(struct tipc_link *l)
311 {
312 return l->window;
313 }
314
tipc_link_prio(struct tipc_link * l)315 int tipc_link_prio(struct tipc_link *l)
316 {
317 return l->priority;
318 }
319
tipc_link_tolerance(struct tipc_link * l)320 unsigned long tipc_link_tolerance(struct tipc_link *l)
321 {
322 return l->tolerance;
323 }
324
tipc_link_inputq(struct tipc_link * l)325 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
326 {
327 return l->inputq;
328 }
329
tipc_link_plane(struct tipc_link * l)330 char tipc_link_plane(struct tipc_link *l)
331 {
332 return l->net_plane;
333 }
334
tipc_link_update_caps(struct tipc_link * l,u16 capabilities)335 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
336 {
337 l->peer_caps = capabilities;
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 trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
363 tipc_link_reset(rcv_l);
364 rcv_l->state = LINK_RESET;
365 if (!snd_l->ackers) {
366 trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
367 tipc_link_reset(snd_l);
368 snd_l->state = LINK_RESET;
369 __skb_queue_purge(xmitq);
370 }
371 }
372
tipc_link_bc_peers(struct tipc_link * l)373 int tipc_link_bc_peers(struct tipc_link *l)
374 {
375 return l->ackers;
376 }
377
link_bc_rcv_gap(struct tipc_link * l)378 static u16 link_bc_rcv_gap(struct tipc_link *l)
379 {
380 struct sk_buff *skb = skb_peek(&l->deferdq);
381 u16 gap = 0;
382
383 if (more(l->snd_nxt, l->rcv_nxt))
384 gap = l->snd_nxt - l->rcv_nxt;
385 if (skb)
386 gap = buf_seqno(skb) - l->rcv_nxt;
387 return gap;
388 }
389
tipc_link_set_mtu(struct tipc_link * l,int mtu)390 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
391 {
392 l->mtu = mtu;
393 }
394
tipc_link_mtu(struct tipc_link * l)395 int tipc_link_mtu(struct tipc_link *l)
396 {
397 return l->mtu;
398 }
399
tipc_link_rcv_nxt(struct tipc_link * l)400 u16 tipc_link_rcv_nxt(struct tipc_link *l)
401 {
402 return l->rcv_nxt;
403 }
404
tipc_link_acked(struct tipc_link * l)405 u16 tipc_link_acked(struct tipc_link *l)
406 {
407 return l->acked;
408 }
409
tipc_link_name(struct tipc_link * l)410 char *tipc_link_name(struct tipc_link *l)
411 {
412 return l->name;
413 }
414
tipc_link_state(struct tipc_link * l)415 u32 tipc_link_state(struct tipc_link *l)
416 {
417 return l->state;
418 }
419
420 /**
421 * tipc_link_create - create a new link
422 * @n: pointer to associated node
423 * @if_name: associated interface name
424 * @bearer_id: id (index) of associated bearer
425 * @tolerance: link tolerance to be used by link
426 * @net_plane: network plane (A,B,c..) this link belongs to
427 * @mtu: mtu to be advertised by link
428 * @priority: priority to be used by link
429 * @window: send window to be used by link
430 * @session: session to be used by link
431 * @ownnode: identity of own node
432 * @peer: node id of peer node
433 * @peer_caps: bitmap describing peer node capabilities
434 * @bc_sndlink: the namespace global link used for broadcast sending
435 * @bc_rcvlink: the peer specific link used for broadcast reception
436 * @inputq: queue to put messages ready for delivery
437 * @namedq: queue to put binding table update messages ready for delivery
438 * @link: return value, pointer to put the created link
439 *
440 * Returns true if link was created, otherwise false
441 */
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 self,u32 peer,u8 * peer_id,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)442 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
443 int tolerance, char net_plane, u32 mtu, int priority,
444 int window, u32 session, u32 self,
445 u32 peer, u8 *peer_id, u16 peer_caps,
446 struct tipc_link *bc_sndlink,
447 struct tipc_link *bc_rcvlink,
448 struct sk_buff_head *inputq,
449 struct sk_buff_head *namedq,
450 struct tipc_link **link)
451 {
452 char peer_str[NODE_ID_STR_LEN] = {0,};
453 char self_str[NODE_ID_STR_LEN] = {0,};
454 struct tipc_link *l;
455
456 l = kzalloc(sizeof(*l), GFP_ATOMIC);
457 if (!l)
458 return false;
459 *link = l;
460 l->session = session;
461
462 /* Set link name for unicast links only */
463 if (peer_id) {
464 tipc_nodeid2string(self_str, tipc_own_id(net));
465 if (strlen(self_str) > 16)
466 sprintf(self_str, "%x", self);
467 tipc_nodeid2string(peer_str, peer_id);
468 if (strlen(peer_str) > 16)
469 sprintf(peer_str, "%x", peer);
470 }
471 /* Peer i/f name will be completed by reset/activate message */
472 snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
473 self_str, if_name, peer_str);
474
475 strcpy(l->if_name, if_name);
476 l->addr = peer;
477 l->peer_caps = peer_caps;
478 l->net = net;
479 l->in_session = false;
480 l->bearer_id = bearer_id;
481 l->tolerance = tolerance;
482 if (bc_rcvlink)
483 bc_rcvlink->tolerance = tolerance;
484 l->net_plane = net_plane;
485 l->advertised_mtu = mtu;
486 l->mtu = mtu;
487 l->priority = priority;
488 tipc_link_set_queue_limits(l, window);
489 l->ackers = 1;
490 l->bc_sndlink = bc_sndlink;
491 l->bc_rcvlink = bc_rcvlink;
492 l->inputq = inputq;
493 l->namedq = namedq;
494 l->state = LINK_RESETTING;
495 __skb_queue_head_init(&l->transmq);
496 __skb_queue_head_init(&l->backlogq);
497 __skb_queue_head_init(&l->deferdq);
498 __skb_queue_head_init(&l->failover_deferdq);
499 skb_queue_head_init(&l->wakeupq);
500 skb_queue_head_init(l->inputq);
501 return true;
502 }
503
504 /**
505 * tipc_link_bc_create - create new link to be used for broadcast
506 * @n: pointer to associated node
507 * @mtu: mtu to be used initially if no peers
508 * @window: send window to be used
509 * @inputq: queue to put messages ready for delivery
510 * @namedq: queue to put binding table update messages ready for delivery
511 * @link: return value, pointer to put the created link
512 *
513 * Returns true if link was created, otherwise false
514 */
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)515 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
516 int mtu, int window, u16 peer_caps,
517 struct sk_buff_head *inputq,
518 struct sk_buff_head *namedq,
519 struct tipc_link *bc_sndlink,
520 struct tipc_link **link)
521 {
522 struct tipc_link *l;
523
524 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
525 0, ownnode, peer, NULL, peer_caps, bc_sndlink,
526 NULL, inputq, namedq, link))
527 return false;
528
529 l = *link;
530 strcpy(l->name, tipc_bclink_name);
531 trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
532 tipc_link_reset(l);
533 l->state = LINK_RESET;
534 l->ackers = 0;
535 l->bc_rcvlink = l;
536
537 /* Broadcast send link is always up */
538 if (link_is_bc_sndlink(l))
539 l->state = LINK_ESTABLISHED;
540
541 /* Disable replicast if even a single peer doesn't support it */
542 if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
543 tipc_bcast_disable_rcast(net);
544
545 return true;
546 }
547
548 /**
549 * tipc_link_fsm_evt - link finite state machine
550 * @l: pointer to link
551 * @evt: state machine event to be processed
552 */
tipc_link_fsm_evt(struct tipc_link * l,int evt)553 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
554 {
555 int rc = 0;
556 int old_state = l->state;
557
558 switch (l->state) {
559 case LINK_RESETTING:
560 switch (evt) {
561 case LINK_PEER_RESET_EVT:
562 l->state = LINK_PEER_RESET;
563 break;
564 case LINK_RESET_EVT:
565 l->state = LINK_RESET;
566 break;
567 case LINK_FAILURE_EVT:
568 case LINK_FAILOVER_BEGIN_EVT:
569 case LINK_ESTABLISH_EVT:
570 case LINK_FAILOVER_END_EVT:
571 case LINK_SYNCH_BEGIN_EVT:
572 case LINK_SYNCH_END_EVT:
573 default:
574 goto illegal_evt;
575 }
576 break;
577 case LINK_RESET:
578 switch (evt) {
579 case LINK_PEER_RESET_EVT:
580 l->state = LINK_ESTABLISHING;
581 break;
582 case LINK_FAILOVER_BEGIN_EVT:
583 l->state = LINK_FAILINGOVER;
584 case LINK_FAILURE_EVT:
585 case LINK_RESET_EVT:
586 case LINK_ESTABLISH_EVT:
587 case LINK_FAILOVER_END_EVT:
588 break;
589 case LINK_SYNCH_BEGIN_EVT:
590 case LINK_SYNCH_END_EVT:
591 default:
592 goto illegal_evt;
593 }
594 break;
595 case LINK_PEER_RESET:
596 switch (evt) {
597 case LINK_RESET_EVT:
598 l->state = LINK_ESTABLISHING;
599 break;
600 case LINK_PEER_RESET_EVT:
601 case LINK_ESTABLISH_EVT:
602 case LINK_FAILURE_EVT:
603 break;
604 case LINK_SYNCH_BEGIN_EVT:
605 case LINK_SYNCH_END_EVT:
606 case LINK_FAILOVER_BEGIN_EVT:
607 case LINK_FAILOVER_END_EVT:
608 default:
609 goto illegal_evt;
610 }
611 break;
612 case LINK_FAILINGOVER:
613 switch (evt) {
614 case LINK_FAILOVER_END_EVT:
615 l->state = LINK_RESET;
616 break;
617 case LINK_PEER_RESET_EVT:
618 case LINK_RESET_EVT:
619 case LINK_ESTABLISH_EVT:
620 case LINK_FAILURE_EVT:
621 break;
622 case LINK_FAILOVER_BEGIN_EVT:
623 case LINK_SYNCH_BEGIN_EVT:
624 case LINK_SYNCH_END_EVT:
625 default:
626 goto illegal_evt;
627 }
628 break;
629 case LINK_ESTABLISHING:
630 switch (evt) {
631 case LINK_ESTABLISH_EVT:
632 l->state = LINK_ESTABLISHED;
633 break;
634 case LINK_FAILOVER_BEGIN_EVT:
635 l->state = LINK_FAILINGOVER;
636 break;
637 case LINK_RESET_EVT:
638 l->state = LINK_RESET;
639 break;
640 case LINK_FAILURE_EVT:
641 case LINK_PEER_RESET_EVT:
642 case LINK_SYNCH_BEGIN_EVT:
643 case LINK_FAILOVER_END_EVT:
644 break;
645 case LINK_SYNCH_END_EVT:
646 default:
647 goto illegal_evt;
648 }
649 break;
650 case LINK_ESTABLISHED:
651 switch (evt) {
652 case LINK_PEER_RESET_EVT:
653 l->state = LINK_PEER_RESET;
654 rc |= TIPC_LINK_DOWN_EVT;
655 break;
656 case LINK_FAILURE_EVT:
657 l->state = LINK_RESETTING;
658 rc |= TIPC_LINK_DOWN_EVT;
659 break;
660 case LINK_RESET_EVT:
661 l->state = LINK_RESET;
662 break;
663 case LINK_ESTABLISH_EVT:
664 case LINK_SYNCH_END_EVT:
665 break;
666 case LINK_SYNCH_BEGIN_EVT:
667 l->state = LINK_SYNCHING;
668 break;
669 case LINK_FAILOVER_BEGIN_EVT:
670 case LINK_FAILOVER_END_EVT:
671 default:
672 goto illegal_evt;
673 }
674 break;
675 case LINK_SYNCHING:
676 switch (evt) {
677 case LINK_PEER_RESET_EVT:
678 l->state = LINK_PEER_RESET;
679 rc |= TIPC_LINK_DOWN_EVT;
680 break;
681 case LINK_FAILURE_EVT:
682 l->state = LINK_RESETTING;
683 rc |= TIPC_LINK_DOWN_EVT;
684 break;
685 case LINK_RESET_EVT:
686 l->state = LINK_RESET;
687 break;
688 case LINK_ESTABLISH_EVT:
689 case LINK_SYNCH_BEGIN_EVT:
690 break;
691 case LINK_SYNCH_END_EVT:
692 l->state = LINK_ESTABLISHED;
693 break;
694 case LINK_FAILOVER_BEGIN_EVT:
695 case LINK_FAILOVER_END_EVT:
696 default:
697 goto illegal_evt;
698 }
699 break;
700 default:
701 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
702 }
703 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
704 return rc;
705 illegal_evt:
706 pr_err("Illegal FSM event %x in state %x on link %s\n",
707 evt, l->state, l->name);
708 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
709 return rc;
710 }
711
712 /* link_profile_stats - update statistical profiling of traffic
713 */
link_profile_stats(struct tipc_link * l)714 static void link_profile_stats(struct tipc_link *l)
715 {
716 struct sk_buff *skb;
717 struct tipc_msg *msg;
718 int length;
719
720 /* Update counters used in statistical profiling of send traffic */
721 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
722 l->stats.queue_sz_counts++;
723
724 skb = skb_peek(&l->transmq);
725 if (!skb)
726 return;
727 msg = buf_msg(skb);
728 length = msg_size(msg);
729
730 if (msg_user(msg) == MSG_FRAGMENTER) {
731 if (msg_type(msg) != FIRST_FRAGMENT)
732 return;
733 length = msg_size(msg_inner_hdr(msg));
734 }
735 l->stats.msg_lengths_total += length;
736 l->stats.msg_length_counts++;
737 if (length <= 64)
738 l->stats.msg_length_profile[0]++;
739 else if (length <= 256)
740 l->stats.msg_length_profile[1]++;
741 else if (length <= 1024)
742 l->stats.msg_length_profile[2]++;
743 else if (length <= 4096)
744 l->stats.msg_length_profile[3]++;
745 else if (length <= 16384)
746 l->stats.msg_length_profile[4]++;
747 else if (length <= 32768)
748 l->stats.msg_length_profile[5]++;
749 else
750 l->stats.msg_length_profile[6]++;
751 }
752
753 /**
754 * tipc_link_too_silent - check if link is "too silent"
755 * @l: tipc link to be checked
756 *
757 * Returns true if the link 'silent_intv_cnt' is about to reach the
758 * 'abort_limit' value, otherwise false
759 */
tipc_link_too_silent(struct tipc_link * l)760 bool tipc_link_too_silent(struct tipc_link *l)
761 {
762 return (l->silent_intv_cnt + 2 > l->abort_limit);
763 }
764
765 /* tipc_link_timeout - perform periodic task as instructed from node timeout
766 */
tipc_link_timeout(struct tipc_link * l,struct sk_buff_head * xmitq)767 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
768 {
769 int mtyp = 0;
770 int rc = 0;
771 bool state = false;
772 bool probe = false;
773 bool setup = false;
774 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
775 u16 bc_acked = l->bc_rcvlink->acked;
776 struct tipc_mon_state *mstate = &l->mon_state;
777
778 trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
779 trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
780 switch (l->state) {
781 case LINK_ESTABLISHED:
782 case LINK_SYNCHING:
783 mtyp = STATE_MSG;
784 link_profile_stats(l);
785 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
786 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
787 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
788 state = bc_acked != bc_snt;
789 state |= l->bc_rcvlink->rcv_unacked;
790 state |= l->rcv_unacked;
791 state |= !skb_queue_empty(&l->transmq);
792 state |= !skb_queue_empty(&l->deferdq);
793 probe = mstate->probing;
794 probe |= l->silent_intv_cnt;
795 if (probe || mstate->monitoring)
796 l->silent_intv_cnt++;
797 break;
798 case LINK_RESET:
799 setup = l->rst_cnt++ <= 4;
800 setup |= !(l->rst_cnt % 16);
801 mtyp = RESET_MSG;
802 break;
803 case LINK_ESTABLISHING:
804 setup = true;
805 mtyp = ACTIVATE_MSG;
806 break;
807 case LINK_PEER_RESET:
808 case LINK_RESETTING:
809 case LINK_FAILINGOVER:
810 break;
811 default:
812 break;
813 }
814
815 if (state || probe || setup)
816 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
817
818 return rc;
819 }
820
821 /**
822 * link_schedule_user - schedule a message sender for wakeup after congestion
823 * @l: congested link
824 * @hdr: header of message that is being sent
825 * Create pseudo msg to send back to user when congestion abates
826 */
link_schedule_user(struct tipc_link * l,struct tipc_msg * hdr)827 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
828 {
829 u32 dnode = tipc_own_addr(l->net);
830 u32 dport = msg_origport(hdr);
831 struct sk_buff *skb;
832
833 /* Create and schedule wakeup pseudo message */
834 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
835 dnode, l->addr, dport, 0, 0);
836 if (!skb)
837 return -ENOBUFS;
838 msg_set_dest_droppable(buf_msg(skb), true);
839 TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
840 skb_queue_tail(&l->wakeupq, skb);
841 l->stats.link_congs++;
842 trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
843 return -ELINKCONG;
844 }
845
846 /**
847 * link_prepare_wakeup - prepare users for wakeup after congestion
848 * @l: congested link
849 * Wake up a number of waiting users, as permitted by available space
850 * in the send queue
851 */
link_prepare_wakeup(struct tipc_link * l)852 static void link_prepare_wakeup(struct tipc_link *l)
853 {
854 struct sk_buff_head *wakeupq = &l->wakeupq;
855 struct sk_buff_head *inputq = l->inputq;
856 struct sk_buff *skb, *tmp;
857 struct sk_buff_head tmpq;
858 int avail[5] = {0,};
859 int imp = 0;
860
861 __skb_queue_head_init(&tmpq);
862
863 for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
864 avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
865
866 skb_queue_walk_safe(wakeupq, skb, tmp) {
867 imp = TIPC_SKB_CB(skb)->chain_imp;
868 if (avail[imp] <= 0)
869 continue;
870 avail[imp]--;
871 __skb_unlink(skb, wakeupq);
872 __skb_queue_tail(&tmpq, skb);
873 }
874
875 spin_lock_bh(&inputq->lock);
876 skb_queue_splice_tail(&tmpq, inputq);
877 spin_unlock_bh(&inputq->lock);
878
879 }
880
tipc_link_reset(struct tipc_link * l)881 void tipc_link_reset(struct tipc_link *l)
882 {
883 struct sk_buff_head list;
884 u32 imp;
885
886 __skb_queue_head_init(&list);
887
888 l->in_session = false;
889 /* Force re-synch of peer session number before establishing */
890 l->peer_session--;
891 l->session++;
892 l->mtu = l->advertised_mtu;
893
894 spin_lock_bh(&l->wakeupq.lock);
895 skb_queue_splice_init(&l->wakeupq, &list);
896 spin_unlock_bh(&l->wakeupq.lock);
897
898 spin_lock_bh(&l->inputq->lock);
899 skb_queue_splice_init(&list, l->inputq);
900 spin_unlock_bh(&l->inputq->lock);
901
902 __skb_queue_purge(&l->transmq);
903 __skb_queue_purge(&l->deferdq);
904 __skb_queue_purge(&l->backlogq);
905 __skb_queue_purge(&l->failover_deferdq);
906 for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
907 l->backlog[imp].len = 0;
908 l->backlog[imp].target_bskb = NULL;
909 }
910 kfree_skb(l->reasm_buf);
911 kfree_skb(l->reasm_tnlmsg);
912 kfree_skb(l->failover_reasm_skb);
913 l->reasm_buf = NULL;
914 l->reasm_tnlmsg = NULL;
915 l->failover_reasm_skb = NULL;
916 l->rcv_unacked = 0;
917 l->snd_nxt = 1;
918 l->rcv_nxt = 1;
919 l->snd_nxt_state = 1;
920 l->rcv_nxt_state = 1;
921 l->acked = 0;
922 l->silent_intv_cnt = 0;
923 l->rst_cnt = 0;
924 l->bc_peer_is_up = false;
925 memset(&l->mon_state, 0, sizeof(l->mon_state));
926 tipc_link_reset_stats(l);
927 }
928
929 /**
930 * tipc_link_xmit(): enqueue buffer list according to queue situation
931 * @link: link to use
932 * @list: chain of buffers containing message
933 * @xmitq: returned list of packets to be sent by caller
934 *
935 * Consumes the buffer chain.
936 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
937 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
938 */
tipc_link_xmit(struct tipc_link * l,struct sk_buff_head * list,struct sk_buff_head * xmitq)939 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
940 struct sk_buff_head *xmitq)
941 {
942 unsigned int maxwin = l->window;
943 unsigned int mtu = l->mtu;
944 u16 ack = l->rcv_nxt - 1;
945 u16 seqno = l->snd_nxt;
946 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
947 struct sk_buff_head *transmq = &l->transmq;
948 struct sk_buff_head *backlogq = &l->backlogq;
949 struct sk_buff *skb, *_skb, **tskb;
950 int pkt_cnt = skb_queue_len(list);
951 struct tipc_msg *hdr;
952 int rc = 0;
953 int imp;
954
955 if (pkt_cnt <= 0)
956 return 0;
957
958 hdr = buf_msg(skb_peek(list));
959 if (unlikely(msg_size(hdr) > mtu)) {
960 pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
961 skb_queue_len(list), msg_user(hdr),
962 msg_type(hdr), msg_size(hdr), mtu);
963 __skb_queue_purge(list);
964 return -EMSGSIZE;
965 }
966
967 imp = msg_importance(hdr);
968 /* Allow oversubscription of one data msg per source at congestion */
969 if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
970 if (imp == TIPC_SYSTEM_IMPORTANCE) {
971 pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
972 return -ENOBUFS;
973 }
974 rc = link_schedule_user(l, hdr);
975 }
976
977 if (pkt_cnt > 1) {
978 l->stats.sent_fragmented++;
979 l->stats.sent_fragments += pkt_cnt;
980 }
981
982 /* Prepare each packet for sending, and add to relevant queue: */
983 while (skb_queue_len(list)) {
984 skb = skb_peek(list);
985 hdr = buf_msg(skb);
986 msg_set_seqno(hdr, seqno);
987 msg_set_ack(hdr, ack);
988 msg_set_bcast_ack(hdr, bc_ack);
989
990 if (likely(skb_queue_len(transmq) < maxwin)) {
991 _skb = skb_clone(skb, GFP_ATOMIC);
992 if (!_skb) {
993 __skb_queue_purge(list);
994 return -ENOBUFS;
995 }
996 __skb_dequeue(list);
997 __skb_queue_tail(transmq, skb);
998 /* next retransmit attempt */
999 if (link_is_bc_sndlink(l))
1000 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1001 __skb_queue_tail(xmitq, _skb);
1002 TIPC_SKB_CB(skb)->ackers = l->ackers;
1003 l->rcv_unacked = 0;
1004 l->stats.sent_pkts++;
1005 seqno++;
1006 continue;
1007 }
1008 tskb = &l->backlog[imp].target_bskb;
1009 if (tipc_msg_bundle(*tskb, hdr, mtu)) {
1010 kfree_skb(__skb_dequeue(list));
1011 l->stats.sent_bundled++;
1012 continue;
1013 }
1014 if (tipc_msg_make_bundle(tskb, hdr, mtu, l->addr)) {
1015 kfree_skb(__skb_dequeue(list));
1016 __skb_queue_tail(backlogq, *tskb);
1017 l->backlog[imp].len++;
1018 l->stats.sent_bundled++;
1019 l->stats.sent_bundles++;
1020 continue;
1021 }
1022 l->backlog[imp].target_bskb = NULL;
1023 l->backlog[imp].len += skb_queue_len(list);
1024 skb_queue_splice_tail_init(list, backlogq);
1025 }
1026 l->snd_nxt = seqno;
1027 return rc;
1028 }
1029
tipc_link_advance_backlog(struct tipc_link * l,struct sk_buff_head * xmitq)1030 static void tipc_link_advance_backlog(struct tipc_link *l,
1031 struct sk_buff_head *xmitq)
1032 {
1033 struct sk_buff *skb, *_skb;
1034 struct tipc_msg *hdr;
1035 u16 seqno = l->snd_nxt;
1036 u16 ack = l->rcv_nxt - 1;
1037 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1038 u32 imp;
1039
1040 while (skb_queue_len(&l->transmq) < l->window) {
1041 skb = skb_peek(&l->backlogq);
1042 if (!skb)
1043 break;
1044 _skb = skb_clone(skb, GFP_ATOMIC);
1045 if (!_skb)
1046 break;
1047 __skb_dequeue(&l->backlogq);
1048 hdr = buf_msg(skb);
1049 imp = msg_importance(hdr);
1050 l->backlog[imp].len--;
1051 if (unlikely(skb == l->backlog[imp].target_bskb))
1052 l->backlog[imp].target_bskb = NULL;
1053 __skb_queue_tail(&l->transmq, skb);
1054 /* next retransmit attempt */
1055 if (link_is_bc_sndlink(l))
1056 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1057
1058 __skb_queue_tail(xmitq, _skb);
1059 TIPC_SKB_CB(skb)->ackers = l->ackers;
1060 msg_set_seqno(hdr, seqno);
1061 msg_set_ack(hdr, ack);
1062 msg_set_bcast_ack(hdr, bc_ack);
1063 l->rcv_unacked = 0;
1064 l->stats.sent_pkts++;
1065 seqno++;
1066 }
1067 l->snd_nxt = seqno;
1068 }
1069
1070 /**
1071 * link_retransmit_failure() - Detect repeated retransmit failures
1072 * @l: tipc link sender
1073 * @r: tipc link receiver (= l in case of unicast)
1074 * @rc: returned code
1075 *
1076 * Return: true if the repeated retransmit failures happens, otherwise
1077 * false
1078 */
link_retransmit_failure(struct tipc_link * l,struct tipc_link * r,int * rc)1079 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1080 int *rc)
1081 {
1082 struct sk_buff *skb = skb_peek(&l->transmq);
1083 struct tipc_msg *hdr;
1084
1085 if (!skb)
1086 return false;
1087
1088 if (!TIPC_SKB_CB(skb)->retr_cnt)
1089 return false;
1090
1091 if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1092 msecs_to_jiffies(r->tolerance * 10)))
1093 return false;
1094
1095 hdr = buf_msg(skb);
1096 if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1097 return false;
1098
1099 pr_warn("Retransmission failure on link <%s>\n", l->name);
1100 link_print(l, "State of link ");
1101 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1102 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1103 pr_info("sqno %u, prev: %x, dest: %x\n",
1104 msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1105 pr_info("retr_stamp %d, retr_cnt %d\n",
1106 jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1107 TIPC_SKB_CB(skb)->retr_cnt);
1108
1109 trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1110 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1111 trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1112
1113 if (link_is_bc_sndlink(l)) {
1114 r->state = LINK_RESET;
1115 *rc = TIPC_LINK_DOWN_EVT;
1116 } else {
1117 *rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1118 }
1119
1120 return true;
1121 }
1122
1123 /* tipc_link_bc_retrans() - retransmit zero or more packets
1124 * @l: the link to transmit on
1125 * @r: the receiving link ordering the retransmit. Same as l if unicast
1126 * @from: retransmit from (inclusive) this sequence number
1127 * @to: retransmit to (inclusive) this sequence number
1128 * xmitq: queue for accumulating the retransmitted packets
1129 */
tipc_link_bc_retrans(struct tipc_link * l,struct tipc_link * r,u16 from,u16 to,struct sk_buff_head * xmitq)1130 static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r,
1131 u16 from, u16 to, struct sk_buff_head *xmitq)
1132 {
1133 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1134 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1135 u16 ack = l->rcv_nxt - 1;
1136 struct tipc_msg *hdr;
1137 int rc = 0;
1138
1139 if (!skb)
1140 return 0;
1141 if (less(to, from))
1142 return 0;
1143
1144 trace_tipc_link_retrans(r, from, to, &l->transmq);
1145
1146 if (link_retransmit_failure(l, r, &rc))
1147 return rc;
1148
1149 skb_queue_walk(&l->transmq, skb) {
1150 hdr = buf_msg(skb);
1151 if (less(msg_seqno(hdr), from))
1152 continue;
1153 if (more(msg_seqno(hdr), to))
1154 break;
1155
1156 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1157 continue;
1158 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1159 _skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE, GFP_ATOMIC);
1160 if (!_skb)
1161 return 0;
1162 hdr = buf_msg(_skb);
1163 msg_set_ack(hdr, ack);
1164 msg_set_bcast_ack(hdr, bc_ack);
1165 _skb->priority = TC_PRIO_CONTROL;
1166 __skb_queue_tail(xmitq, _skb);
1167 l->stats.retransmitted++;
1168
1169 /* Increase actual retrans counter & mark first time */
1170 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1171 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1172 }
1173 return 0;
1174 }
1175
1176 /* tipc_data_input - deliver data and name distr msgs to upper layer
1177 *
1178 * Consumes buffer if message is of right type
1179 * Node lock must be held
1180 */
tipc_data_input(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq)1181 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1182 struct sk_buff_head *inputq)
1183 {
1184 struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1185 struct tipc_msg *hdr = buf_msg(skb);
1186
1187 switch (msg_user(hdr)) {
1188 case TIPC_LOW_IMPORTANCE:
1189 case TIPC_MEDIUM_IMPORTANCE:
1190 case TIPC_HIGH_IMPORTANCE:
1191 case TIPC_CRITICAL_IMPORTANCE:
1192 if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1193 skb_queue_tail(mc_inputq, skb);
1194 return true;
1195 }
1196 /* fall through */
1197 case CONN_MANAGER:
1198 skb_queue_tail(inputq, skb);
1199 return true;
1200 case GROUP_PROTOCOL:
1201 skb_queue_tail(mc_inputq, skb);
1202 return true;
1203 case NAME_DISTRIBUTOR:
1204 l->bc_rcvlink->state = LINK_ESTABLISHED;
1205 skb_queue_tail(l->namedq, skb);
1206 return true;
1207 case MSG_BUNDLER:
1208 case TUNNEL_PROTOCOL:
1209 case MSG_FRAGMENTER:
1210 case BCAST_PROTOCOL:
1211 return false;
1212 default:
1213 pr_warn("Dropping received illegal msg type\n");
1214 kfree_skb(skb);
1215 return true;
1216 };
1217 }
1218
1219 /* tipc_link_input - process packet that has passed link protocol check
1220 *
1221 * Consumes buffer
1222 */
tipc_link_input(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq,struct sk_buff ** reasm_skb)1223 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1224 struct sk_buff_head *inputq,
1225 struct sk_buff **reasm_skb)
1226 {
1227 struct tipc_msg *hdr = buf_msg(skb);
1228 struct sk_buff *iskb;
1229 struct sk_buff_head tmpq;
1230 int usr = msg_user(hdr);
1231 int pos = 0;
1232
1233 if (usr == MSG_BUNDLER) {
1234 skb_queue_head_init(&tmpq);
1235 l->stats.recv_bundles++;
1236 l->stats.recv_bundled += msg_msgcnt(hdr);
1237 while (tipc_msg_extract(skb, &iskb, &pos))
1238 tipc_data_input(l, iskb, &tmpq);
1239 tipc_skb_queue_splice_tail(&tmpq, inputq);
1240 return 0;
1241 } else if (usr == MSG_FRAGMENTER) {
1242 l->stats.recv_fragments++;
1243 if (tipc_buf_append(reasm_skb, &skb)) {
1244 l->stats.recv_fragmented++;
1245 tipc_data_input(l, skb, inputq);
1246 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1247 pr_warn_ratelimited("Unable to build fragment list\n");
1248 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1249 }
1250 return 0;
1251 } else if (usr == BCAST_PROTOCOL) {
1252 tipc_bcast_lock(l->net);
1253 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1254 tipc_bcast_unlock(l->net);
1255 }
1256
1257 kfree_skb(skb);
1258 return 0;
1259 }
1260
1261 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1262 * inner message along with the ones in the old link's
1263 * deferdq
1264 * @l: tunnel link
1265 * @skb: TUNNEL_PROTOCOL message
1266 * @inputq: queue to put messages ready for delivery
1267 */
tipc_link_tnl_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq)1268 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1269 struct sk_buff_head *inputq)
1270 {
1271 struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1272 struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1273 struct sk_buff_head *fdefq = &l->failover_deferdq;
1274 struct tipc_msg *hdr = buf_msg(skb);
1275 struct sk_buff *iskb;
1276 int ipos = 0;
1277 int rc = 0;
1278 u16 seqno;
1279
1280 if (msg_type(hdr) == SYNCH_MSG) {
1281 kfree_skb(skb);
1282 return 0;
1283 }
1284
1285 /* Not a fragment? */
1286 if (likely(!msg_nof_fragms(hdr))) {
1287 if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1288 pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1289 skb_queue_len(fdefq));
1290 return 0;
1291 }
1292 kfree_skb(skb);
1293 } else {
1294 /* Set fragment type for buf_append */
1295 if (msg_fragm_no(hdr) == 1)
1296 msg_set_type(hdr, FIRST_FRAGMENT);
1297 else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1298 msg_set_type(hdr, FRAGMENT);
1299 else
1300 msg_set_type(hdr, LAST_FRAGMENT);
1301
1302 if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1303 /* Successful but non-complete reassembly? */
1304 if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1305 return 0;
1306 pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1307 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1308 }
1309 iskb = skb;
1310 }
1311
1312 do {
1313 seqno = buf_seqno(iskb);
1314 if (unlikely(less(seqno, l->drop_point))) {
1315 kfree_skb(iskb);
1316 continue;
1317 }
1318 if (unlikely(seqno != l->drop_point)) {
1319 __tipc_skb_queue_sorted(fdefq, seqno, iskb);
1320 continue;
1321 }
1322
1323 l->drop_point++;
1324 if (!tipc_data_input(l, iskb, inputq))
1325 rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1326 if (unlikely(rc))
1327 break;
1328 } while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1329
1330 return rc;
1331 }
1332
tipc_link_release_pkts(struct tipc_link * l,u16 acked)1333 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1334 {
1335 bool released = false;
1336 struct sk_buff *skb, *tmp;
1337
1338 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1339 if (more(buf_seqno(skb), acked))
1340 break;
1341 __skb_unlink(skb, &l->transmq);
1342 kfree_skb(skb);
1343 released = true;
1344 }
1345 return released;
1346 }
1347
1348 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1349 * @l: tipc link that data have come with gaps in sequence if any
1350 * @data: data buffer to store the Gap ACK blocks after built
1351 *
1352 * returns the actual allocated memory size
1353 */
tipc_build_gap_ack_blks(struct tipc_link * l,void * data)1354 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data)
1355 {
1356 struct sk_buff *skb = skb_peek(&l->deferdq);
1357 struct tipc_gap_ack_blks *ga = data;
1358 u16 len, expect, seqno = 0;
1359 u8 n = 0;
1360
1361 if (!skb)
1362 goto exit;
1363
1364 expect = buf_seqno(skb);
1365 skb_queue_walk(&l->deferdq, skb) {
1366 seqno = buf_seqno(skb);
1367 if (unlikely(more(seqno, expect))) {
1368 ga->gacks[n].ack = htons(expect - 1);
1369 ga->gacks[n].gap = htons(seqno - expect);
1370 if (++n >= MAX_GAP_ACK_BLKS) {
1371 pr_info_ratelimited("Too few Gap ACK blocks!\n");
1372 goto exit;
1373 }
1374 } else if (unlikely(less(seqno, expect))) {
1375 pr_warn("Unexpected skb in deferdq!\n");
1376 continue;
1377 }
1378 expect = seqno + 1;
1379 }
1380
1381 /* last block */
1382 ga->gacks[n].ack = htons(seqno);
1383 ga->gacks[n].gap = 0;
1384 n++;
1385
1386 exit:
1387 len = tipc_gap_ack_blks_sz(n);
1388 ga->len = htons(len);
1389 ga->gack_cnt = n;
1390 return len;
1391 }
1392
1393 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1394 * acked packets, also doing retransmissions if
1395 * gaps found
1396 * @l: tipc link with transmq queue to be advanced
1397 * @acked: seqno of last packet acked by peer without any gaps before
1398 * @gap: # of gap packets
1399 * @ga: buffer pointer to Gap ACK blocks from peer
1400 * @xmitq: queue for accumulating the retransmitted packets if any
1401 *
1402 * In case of a repeated retransmit failures, the call will return shortly
1403 * with a returned code (e.g. TIPC_LINK_DOWN_EVT)
1404 */
tipc_link_advance_transmq(struct tipc_link * l,u16 acked,u16 gap,struct tipc_gap_ack_blks * ga,struct sk_buff_head * xmitq)1405 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
1406 struct tipc_gap_ack_blks *ga,
1407 struct sk_buff_head *xmitq)
1408 {
1409 struct sk_buff *skb, *_skb, *tmp;
1410 struct tipc_msg *hdr;
1411 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1412 u16 ack = l->rcv_nxt - 1;
1413 bool passed = false;
1414 u16 seqno, n = 0;
1415 int rc = 0;
1416
1417 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1418 seqno = buf_seqno(skb);
1419
1420 next_gap_ack:
1421 if (less_eq(seqno, acked)) {
1422 /* release skb */
1423 __skb_unlink(skb, &l->transmq);
1424 kfree_skb(skb);
1425 } else if (less_eq(seqno, acked + gap)) {
1426 /* First, check if repeated retrans failures occurs? */
1427 if (!passed && link_retransmit_failure(l, l, &rc))
1428 return rc;
1429 passed = true;
1430
1431 /* retransmit skb if unrestricted*/
1432 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1433 continue;
1434 TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
1435 _skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE,
1436 GFP_ATOMIC);
1437 if (!_skb)
1438 continue;
1439 hdr = buf_msg(_skb);
1440 msg_set_ack(hdr, ack);
1441 msg_set_bcast_ack(hdr, bc_ack);
1442 _skb->priority = TC_PRIO_CONTROL;
1443 __skb_queue_tail(xmitq, _skb);
1444 l->stats.retransmitted++;
1445
1446 /* Increase actual retrans counter & mark first time */
1447 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1448 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1449 } else {
1450 /* retry with Gap ACK blocks if any */
1451 if (!ga || n >= ga->gack_cnt)
1452 break;
1453 acked = ntohs(ga->gacks[n].ack);
1454 gap = ntohs(ga->gacks[n].gap);
1455 n++;
1456 goto next_gap_ack;
1457 }
1458 }
1459
1460 return 0;
1461 }
1462
1463 /* tipc_link_build_state_msg: prepare link state message for transmission
1464 *
1465 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1466 * risk of ack storms towards the sender
1467 */
tipc_link_build_state_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1468 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1469 {
1470 if (!l)
1471 return 0;
1472
1473 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1474 if (link_is_bc_rcvlink(l)) {
1475 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1476 return 0;
1477 l->rcv_unacked = 0;
1478
1479 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1480 l->snd_nxt = l->rcv_nxt;
1481 return TIPC_LINK_SND_STATE;
1482 }
1483
1484 /* Unicast ACK */
1485 l->rcv_unacked = 0;
1486 l->stats.sent_acks++;
1487 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1488 return 0;
1489 }
1490
1491 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1492 */
tipc_link_build_reset_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1493 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1494 {
1495 int mtyp = RESET_MSG;
1496 struct sk_buff *skb;
1497
1498 if (l->state == LINK_ESTABLISHING)
1499 mtyp = ACTIVATE_MSG;
1500
1501 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1502
1503 /* Inform peer that this endpoint is going down if applicable */
1504 skb = skb_peek_tail(xmitq);
1505 if (skb && (l->state == LINK_RESET))
1506 msg_set_peer_stopping(buf_msg(skb), 1);
1507 }
1508
1509 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1510 * Note that sending of broadcast NACK is coordinated among nodes, to
1511 * reduce the risk of NACK storms towards the sender
1512 */
tipc_link_build_nack_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1513 static int tipc_link_build_nack_msg(struct tipc_link *l,
1514 struct sk_buff_head *xmitq)
1515 {
1516 u32 def_cnt = ++l->stats.deferred_recv;
1517 u32 defq_len = skb_queue_len(&l->deferdq);
1518 int match1, match2;
1519
1520 if (link_is_bc_rcvlink(l)) {
1521 match1 = def_cnt & 0xf;
1522 match2 = tipc_own_addr(l->net) & 0xf;
1523 if (match1 == match2)
1524 return TIPC_LINK_SND_STATE;
1525 return 0;
1526 }
1527
1528 if (defq_len >= 3 && !((defq_len - 3) % 16))
1529 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1530 return 0;
1531 }
1532
1533 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1534 * @l: the link that should handle the message
1535 * @skb: TIPC packet
1536 * @xmitq: queue to place packets to be sent after this call
1537 */
tipc_link_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)1538 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1539 struct sk_buff_head *xmitq)
1540 {
1541 struct sk_buff_head *defq = &l->deferdq;
1542 struct tipc_msg *hdr = buf_msg(skb);
1543 u16 seqno, rcv_nxt, win_lim;
1544 int rc = 0;
1545
1546 /* Verify and update link state */
1547 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1548 return tipc_link_proto_rcv(l, skb, xmitq);
1549
1550 /* Don't send probe at next timeout expiration */
1551 l->silent_intv_cnt = 0;
1552
1553 do {
1554 hdr = buf_msg(skb);
1555 seqno = msg_seqno(hdr);
1556 rcv_nxt = l->rcv_nxt;
1557 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1558
1559 if (unlikely(!link_is_up(l))) {
1560 if (l->state == LINK_ESTABLISHING)
1561 rc = TIPC_LINK_UP_EVT;
1562 goto drop;
1563 }
1564
1565 /* Drop if outside receive window */
1566 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1567 l->stats.duplicates++;
1568 goto drop;
1569 }
1570
1571 /* Forward queues and wake up waiting users */
1572 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1573 tipc_link_advance_backlog(l, xmitq);
1574 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1575 link_prepare_wakeup(l);
1576 }
1577
1578 /* Defer delivery if sequence gap */
1579 if (unlikely(seqno != rcv_nxt)) {
1580 __tipc_skb_queue_sorted(defq, seqno, skb);
1581 rc |= tipc_link_build_nack_msg(l, xmitq);
1582 break;
1583 }
1584
1585 /* Deliver packet */
1586 l->rcv_nxt++;
1587 l->stats.recv_pkts++;
1588
1589 if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1590 rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1591 else if (!tipc_data_input(l, skb, l->inputq))
1592 rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1593 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1594 rc |= tipc_link_build_state_msg(l, xmitq);
1595 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1596 break;
1597 } while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1598
1599 return rc;
1600 drop:
1601 kfree_skb(skb);
1602 return rc;
1603 }
1604
tipc_link_build_proto_msg(struct tipc_link * l,int mtyp,bool probe,bool probe_reply,u16 rcvgap,int tolerance,int priority,struct sk_buff_head * xmitq)1605 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1606 bool probe_reply, u16 rcvgap,
1607 int tolerance, int priority,
1608 struct sk_buff_head *xmitq)
1609 {
1610 struct tipc_link *bcl = l->bc_rcvlink;
1611 struct sk_buff *skb;
1612 struct tipc_msg *hdr;
1613 struct sk_buff_head *dfq = &l->deferdq;
1614 bool node_up = link_is_up(bcl);
1615 struct tipc_mon_state *mstate = &l->mon_state;
1616 int dlen = 0;
1617 void *data;
1618 u16 glen = 0;
1619
1620 /* Don't send protocol message during reset or link failover */
1621 if (tipc_link_is_blocked(l))
1622 return;
1623
1624 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1625 return;
1626
1627 if (!skb_queue_empty(dfq))
1628 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1629
1630 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1631 tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1632 l->addr, tipc_own_addr(l->net), 0, 0, 0);
1633 if (!skb)
1634 return;
1635
1636 hdr = buf_msg(skb);
1637 data = msg_data(hdr);
1638 msg_set_session(hdr, l->session);
1639 msg_set_bearer_id(hdr, l->bearer_id);
1640 msg_set_net_plane(hdr, l->net_plane);
1641 msg_set_next_sent(hdr, l->snd_nxt);
1642 msg_set_ack(hdr, l->rcv_nxt - 1);
1643 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1644 msg_set_bc_ack_invalid(hdr, !node_up);
1645 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1646 msg_set_link_tolerance(hdr, tolerance);
1647 msg_set_linkprio(hdr, priority);
1648 msg_set_redundant_link(hdr, node_up);
1649 msg_set_seq_gap(hdr, 0);
1650 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1651
1652 if (mtyp == STATE_MSG) {
1653 if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1654 msg_set_seqno(hdr, l->snd_nxt_state++);
1655 msg_set_seq_gap(hdr, rcvgap);
1656 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1657 msg_set_probe(hdr, probe);
1658 msg_set_is_keepalive(hdr, probe || probe_reply);
1659 if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1660 glen = tipc_build_gap_ack_blks(l, data);
1661 tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1662 msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1663 skb_trim(skb, INT_H_SIZE + glen + dlen);
1664 l->stats.sent_states++;
1665 l->rcv_unacked = 0;
1666 } else {
1667 /* RESET_MSG or ACTIVATE_MSG */
1668 if (mtyp == ACTIVATE_MSG) {
1669 msg_set_dest_session_valid(hdr, 1);
1670 msg_set_dest_session(hdr, l->peer_session);
1671 }
1672 msg_set_max_pkt(hdr, l->advertised_mtu);
1673 strcpy(data, l->if_name);
1674 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1675 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1676 }
1677 if (probe)
1678 l->stats.sent_probes++;
1679 if (rcvgap)
1680 l->stats.sent_nacks++;
1681 skb->priority = TC_PRIO_CONTROL;
1682 __skb_queue_tail(xmitq, skb);
1683 trace_tipc_proto_build(skb, false, l->name);
1684 }
1685
tipc_link_create_dummy_tnl_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1686 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1687 struct sk_buff_head *xmitq)
1688 {
1689 u32 onode = tipc_own_addr(l->net);
1690 struct tipc_msg *hdr, *ihdr;
1691 struct sk_buff_head tnlq;
1692 struct sk_buff *skb;
1693 u32 dnode = l->addr;
1694
1695 __skb_queue_head_init(&tnlq);
1696 skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1697 INT_H_SIZE, BASIC_H_SIZE,
1698 dnode, onode, 0, 0, 0);
1699 if (!skb) {
1700 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1701 return;
1702 }
1703
1704 hdr = buf_msg(skb);
1705 msg_set_msgcnt(hdr, 1);
1706 msg_set_bearer_id(hdr, l->peer_bearer_id);
1707
1708 ihdr = (struct tipc_msg *)msg_data(hdr);
1709 tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1710 BASIC_H_SIZE, dnode);
1711 msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1712 __skb_queue_tail(&tnlq, skb);
1713 tipc_link_xmit(l, &tnlq, xmitq);
1714 }
1715
1716 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1717 * with contents of the link's transmit and backlog queues.
1718 */
tipc_link_tnl_prepare(struct tipc_link * l,struct tipc_link * tnl,int mtyp,struct sk_buff_head * xmitq)1719 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1720 int mtyp, struct sk_buff_head *xmitq)
1721 {
1722 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1723 struct sk_buff *skb, *tnlskb;
1724 struct tipc_msg *hdr, tnlhdr;
1725 struct sk_buff_head *queue = &l->transmq;
1726 struct sk_buff_head tmpxq, tnlq, frags;
1727 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1728 bool pktcnt_need_update = false;
1729 u16 syncpt;
1730 int rc;
1731
1732 if (!tnl)
1733 return;
1734
1735 __skb_queue_head_init(&tnlq);
1736 __skb_queue_head_init(&tmpxq);
1737 __skb_queue_head_init(&frags);
1738
1739 /* At least one packet required for safe algorithm => add dummy */
1740 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1741 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1742 0, 0, TIPC_ERR_NO_PORT);
1743 if (!skb) {
1744 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1745 return;
1746 }
1747 __skb_queue_tail(&tnlq, skb);
1748 tipc_link_xmit(l, &tnlq, &tmpxq);
1749 __skb_queue_purge(&tmpxq);
1750
1751 /* Link Synching:
1752 * From now on, send only one single ("dummy") SYNCH message
1753 * to peer. The SYNCH message does not contain any data, just
1754 * a header conveying the synch point to the peer.
1755 */
1756 if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1757 tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1758 INT_H_SIZE, 0, l->addr,
1759 tipc_own_addr(l->net),
1760 0, 0, 0);
1761 if (!tnlskb) {
1762 pr_warn("%sunable to create dummy SYNCH_MSG\n",
1763 link_co_err);
1764 return;
1765 }
1766
1767 hdr = buf_msg(tnlskb);
1768 syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
1769 msg_set_syncpt(hdr, syncpt);
1770 msg_set_bearer_id(hdr, l->peer_bearer_id);
1771 __skb_queue_tail(&tnlq, tnlskb);
1772 tipc_link_xmit(tnl, &tnlq, xmitq);
1773 return;
1774 }
1775
1776 /* Initialize reusable tunnel packet header */
1777 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1778 mtyp, INT_H_SIZE, l->addr);
1779 if (mtyp == SYNCH_MSG)
1780 pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
1781 else
1782 pktcnt = skb_queue_len(&l->transmq);
1783 pktcnt += skb_queue_len(&l->backlogq);
1784 msg_set_msgcnt(&tnlhdr, pktcnt);
1785 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1786 tnl:
1787 /* Wrap each packet into a tunnel packet */
1788 skb_queue_walk(queue, skb) {
1789 hdr = buf_msg(skb);
1790 if (queue == &l->backlogq)
1791 msg_set_seqno(hdr, seqno++);
1792 pktlen = msg_size(hdr);
1793
1794 /* Tunnel link MTU is not large enough? This could be
1795 * due to:
1796 * 1) Link MTU has just changed or set differently;
1797 * 2) Or FAILOVER on the top of a SYNCH message
1798 *
1799 * The 2nd case should not happen if peer supports
1800 * TIPC_TUNNEL_ENHANCED
1801 */
1802 if (pktlen > tnl->mtu - INT_H_SIZE) {
1803 if (mtyp == FAILOVER_MSG &&
1804 (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1805 rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
1806 &frags);
1807 if (rc) {
1808 pr_warn("%sunable to frag msg: rc %d\n",
1809 link_co_err, rc);
1810 return;
1811 }
1812 pktcnt += skb_queue_len(&frags) - 1;
1813 pktcnt_need_update = true;
1814 skb_queue_splice_tail_init(&frags, &tnlq);
1815 continue;
1816 }
1817 /* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
1818 * => Just warn it and return!
1819 */
1820 pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
1821 link_co_err, msg_user(hdr),
1822 msg_type(hdr), msg_size(hdr));
1823 return;
1824 }
1825
1826 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1827 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1828 if (!tnlskb) {
1829 pr_warn("%sunable to send packet\n", link_co_err);
1830 return;
1831 }
1832 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1833 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1834 __skb_queue_tail(&tnlq, tnlskb);
1835 }
1836 if (queue != &l->backlogq) {
1837 queue = &l->backlogq;
1838 goto tnl;
1839 }
1840
1841 if (pktcnt_need_update)
1842 skb_queue_walk(&tnlq, skb) {
1843 hdr = buf_msg(skb);
1844 msg_set_msgcnt(hdr, pktcnt);
1845 }
1846
1847 tipc_link_xmit(tnl, &tnlq, xmitq);
1848
1849 if (mtyp == FAILOVER_MSG) {
1850 tnl->drop_point = l->rcv_nxt;
1851 tnl->failover_reasm_skb = l->reasm_buf;
1852 l->reasm_buf = NULL;
1853
1854 /* Failover the link's deferdq */
1855 if (unlikely(!skb_queue_empty(fdefq))) {
1856 pr_warn("Link failover deferdq not empty: %d!\n",
1857 skb_queue_len(fdefq));
1858 __skb_queue_purge(fdefq);
1859 }
1860 skb_queue_splice_init(&l->deferdq, fdefq);
1861 }
1862 }
1863
1864 /**
1865 * tipc_link_failover_prepare() - prepare tnl for link failover
1866 *
1867 * This is a special version of the precursor - tipc_link_tnl_prepare(),
1868 * see the tipc_node_link_failover() for details
1869 *
1870 * @l: failover link
1871 * @tnl: tunnel link
1872 * @xmitq: queue for messages to be xmited
1873 */
tipc_link_failover_prepare(struct tipc_link * l,struct tipc_link * tnl,struct sk_buff_head * xmitq)1874 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
1875 struct sk_buff_head *xmitq)
1876 {
1877 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1878
1879 tipc_link_create_dummy_tnl_msg(tnl, xmitq);
1880
1881 /* This failover link enpoint was never established before,
1882 * so it has not received anything from peer.
1883 * Otherwise, it must be a normal failover situation or the
1884 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
1885 * would have to start over from scratch instead.
1886 */
1887 tnl->drop_point = 1;
1888 tnl->failover_reasm_skb = NULL;
1889
1890 /* Initiate the link's failover deferdq */
1891 if (unlikely(!skb_queue_empty(fdefq))) {
1892 pr_warn("Link failover deferdq not empty: %d!\n",
1893 skb_queue_len(fdefq));
1894 __skb_queue_purge(fdefq);
1895 }
1896 }
1897
1898 /* tipc_link_validate_msg(): validate message against current link state
1899 * Returns true if message should be accepted, otherwise false
1900 */
tipc_link_validate_msg(struct tipc_link * l,struct tipc_msg * hdr)1901 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
1902 {
1903 u16 curr_session = l->peer_session;
1904 u16 session = msg_session(hdr);
1905 int mtyp = msg_type(hdr);
1906
1907 if (msg_user(hdr) != LINK_PROTOCOL)
1908 return true;
1909
1910 switch (mtyp) {
1911 case RESET_MSG:
1912 if (!l->in_session)
1913 return true;
1914 /* Accept only RESET with new session number */
1915 return more(session, curr_session);
1916 case ACTIVATE_MSG:
1917 if (!l->in_session)
1918 return true;
1919 /* Accept only ACTIVATE with new or current session number */
1920 return !less(session, curr_session);
1921 case STATE_MSG:
1922 /* Accept only STATE with current session number */
1923 if (!l->in_session)
1924 return false;
1925 if (session != curr_session)
1926 return false;
1927 /* Extra sanity check */
1928 if (!link_is_up(l) && msg_ack(hdr))
1929 return false;
1930 if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
1931 return true;
1932 /* Accept only STATE with new sequence number */
1933 return !less(msg_seqno(hdr), l->rcv_nxt_state);
1934 default:
1935 return false;
1936 }
1937 }
1938
1939 /* tipc_link_proto_rcv(): receive link level protocol message :
1940 * Note that network plane id propagates through the network, and may
1941 * change at any time. The node with lowest numerical id determines
1942 * network plane
1943 */
tipc_link_proto_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)1944 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1945 struct sk_buff_head *xmitq)
1946 {
1947 struct tipc_msg *hdr = buf_msg(skb);
1948 struct tipc_gap_ack_blks *ga = NULL;
1949 u16 rcvgap = 0;
1950 u16 ack = msg_ack(hdr);
1951 u16 gap = msg_seq_gap(hdr);
1952 u16 peers_snd_nxt = msg_next_sent(hdr);
1953 u16 peers_tol = msg_link_tolerance(hdr);
1954 u16 peers_prio = msg_linkprio(hdr);
1955 u16 rcv_nxt = l->rcv_nxt;
1956 u32 dlen = msg_data_sz(hdr), glen = 0;
1957 int mtyp = msg_type(hdr);
1958 bool reply = msg_probe(hdr);
1959 void *data;
1960 char *if_name;
1961 int rc = 0;
1962
1963 trace_tipc_proto_rcv(skb, false, l->name);
1964
1965 if (dlen > U16_MAX)
1966 goto exit;
1967
1968 if (tipc_link_is_blocked(l) || !xmitq)
1969 goto exit;
1970
1971 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1972 l->net_plane = msg_net_plane(hdr);
1973
1974 if (skb_linearize(skb))
1975 goto exit;
1976
1977 hdr = buf_msg(skb);
1978 data = msg_data(hdr);
1979
1980 if (!tipc_link_validate_msg(l, hdr)) {
1981 trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
1982 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
1983 goto exit;
1984 }
1985
1986 switch (mtyp) {
1987 case RESET_MSG:
1988 case ACTIVATE_MSG:
1989 /* Complete own link name with peer's interface name */
1990 if_name = strrchr(l->name, ':') + 1;
1991 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1992 break;
1993 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1994 break;
1995 strncpy(if_name, data, TIPC_MAX_IF_NAME);
1996
1997 /* Update own tolerance if peer indicates a non-zero value */
1998 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1999 l->tolerance = peers_tol;
2000 l->bc_rcvlink->tolerance = peers_tol;
2001 }
2002 /* Update own priority if peer's priority is higher */
2003 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
2004 l->priority = peers_prio;
2005
2006 /* If peer is going down we want full re-establish cycle */
2007 if (msg_peer_stopping(hdr)) {
2008 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2009 break;
2010 }
2011
2012 /* If this endpoint was re-created while peer was ESTABLISHING
2013 * it doesn't know current session number. Force re-synch.
2014 */
2015 if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2016 l->session != msg_dest_session(hdr)) {
2017 if (less(l->session, msg_dest_session(hdr)))
2018 l->session = msg_dest_session(hdr) + 1;
2019 break;
2020 }
2021
2022 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2023 if (mtyp == RESET_MSG || !link_is_up(l))
2024 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2025
2026 /* ACTIVATE_MSG takes up link if it was already locally reset */
2027 if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2028 rc = TIPC_LINK_UP_EVT;
2029
2030 l->peer_session = msg_session(hdr);
2031 l->in_session = true;
2032 l->peer_bearer_id = msg_bearer_id(hdr);
2033 if (l->mtu > msg_max_pkt(hdr))
2034 l->mtu = msg_max_pkt(hdr);
2035 break;
2036
2037 case STATE_MSG:
2038 /* Receive Gap ACK blocks from peer if any */
2039 if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
2040 ga = (struct tipc_gap_ack_blks *)data;
2041 glen = ntohs(ga->len);
2042 /* sanity check: if failed, ignore Gap ACK blocks */
2043 if (glen != tipc_gap_ack_blks_sz(ga->gack_cnt))
2044 ga = NULL;
2045 }
2046 if(glen > dlen)
2047 break;
2048
2049 l->rcv_nxt_state = msg_seqno(hdr) + 1;
2050
2051 /* Update own tolerance if peer indicates a non-zero value */
2052 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2053 l->tolerance = peers_tol;
2054 l->bc_rcvlink->tolerance = peers_tol;
2055 }
2056 /* Update own prio if peer indicates a different value */
2057 if ((peers_prio != l->priority) &&
2058 in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2059 l->priority = peers_prio;
2060 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2061 }
2062
2063 l->silent_intv_cnt = 0;
2064 l->stats.recv_states++;
2065 if (msg_probe(hdr))
2066 l->stats.recv_probes++;
2067
2068 if (!link_is_up(l)) {
2069 if (l->state == LINK_ESTABLISHING)
2070 rc = TIPC_LINK_UP_EVT;
2071 break;
2072 }
2073
2074 tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2075 &l->mon_state, l->bearer_id);
2076
2077 /* Send NACK if peer has sent pkts we haven't received yet */
2078 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
2079 rcvgap = peers_snd_nxt - l->rcv_nxt;
2080 if (rcvgap || reply)
2081 tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2082 rcvgap, 0, 0, xmitq);
2083
2084 rc |= tipc_link_advance_transmq(l, ack, gap, ga, xmitq);
2085
2086 /* If NACK, retransmit will now start at right position */
2087 if (gap)
2088 l->stats.recv_nacks++;
2089
2090 tipc_link_advance_backlog(l, xmitq);
2091 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2092 link_prepare_wakeup(l);
2093 }
2094 exit:
2095 kfree_skb(skb);
2096 return rc;
2097 }
2098
2099 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2100 */
tipc_link_build_bc_proto_msg(struct tipc_link * l,bool bcast,u16 peers_snd_nxt,struct sk_buff_head * xmitq)2101 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2102 u16 peers_snd_nxt,
2103 struct sk_buff_head *xmitq)
2104 {
2105 struct sk_buff *skb;
2106 struct tipc_msg *hdr;
2107 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2108 u16 ack = l->rcv_nxt - 1;
2109 u16 gap_to = peers_snd_nxt - 1;
2110
2111 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2112 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2113 if (!skb)
2114 return false;
2115 hdr = buf_msg(skb);
2116 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2117 msg_set_bcast_ack(hdr, ack);
2118 msg_set_bcgap_after(hdr, ack);
2119 if (dfrd_skb)
2120 gap_to = buf_seqno(dfrd_skb) - 1;
2121 msg_set_bcgap_to(hdr, gap_to);
2122 msg_set_non_seq(hdr, bcast);
2123 __skb_queue_tail(xmitq, skb);
2124 return true;
2125 }
2126
2127 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2128 *
2129 * Give a newly added peer node the sequence number where it should
2130 * start receiving and acking broadcast packets.
2131 */
tipc_link_build_bc_init_msg(struct tipc_link * l,struct sk_buff_head * xmitq)2132 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2133 struct sk_buff_head *xmitq)
2134 {
2135 struct sk_buff_head list;
2136
2137 __skb_queue_head_init(&list);
2138 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2139 return;
2140 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2141 tipc_link_xmit(l, &list, xmitq);
2142 }
2143
2144 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2145 */
tipc_link_bc_init_rcv(struct tipc_link * l,struct tipc_msg * hdr)2146 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2147 {
2148 int mtyp = msg_type(hdr);
2149 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2150
2151 if (link_is_up(l))
2152 return;
2153
2154 if (msg_user(hdr) == BCAST_PROTOCOL) {
2155 l->rcv_nxt = peers_snd_nxt;
2156 l->state = LINK_ESTABLISHED;
2157 return;
2158 }
2159
2160 if (l->peer_caps & TIPC_BCAST_SYNCH)
2161 return;
2162
2163 if (msg_peer_node_is_up(hdr))
2164 return;
2165
2166 /* Compatibility: accept older, less safe initial synch data */
2167 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2168 l->rcv_nxt = peers_snd_nxt;
2169 }
2170
2171 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2172 */
tipc_link_bc_sync_rcv(struct tipc_link * l,struct tipc_msg * hdr,struct sk_buff_head * xmitq)2173 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2174 struct sk_buff_head *xmitq)
2175 {
2176 struct tipc_link *snd_l = l->bc_sndlink;
2177 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2178 u16 from = msg_bcast_ack(hdr) + 1;
2179 u16 to = from + msg_bc_gap(hdr) - 1;
2180 int rc = 0;
2181
2182 if (!link_is_up(l))
2183 return rc;
2184
2185 if (!msg_peer_node_is_up(hdr))
2186 return rc;
2187
2188 /* Open when peer ackowledges our bcast init msg (pkt #1) */
2189 if (msg_ack(hdr))
2190 l->bc_peer_is_up = true;
2191
2192 if (!l->bc_peer_is_up)
2193 return rc;
2194
2195 l->stats.recv_nacks++;
2196
2197 /* Ignore if peers_snd_nxt goes beyond receive window */
2198 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2199 return rc;
2200
2201 rc = tipc_link_bc_retrans(snd_l, l, from, to, xmitq);
2202
2203 l->snd_nxt = peers_snd_nxt;
2204 if (link_bc_rcv_gap(l))
2205 rc |= TIPC_LINK_SND_STATE;
2206
2207 /* Return now if sender supports nack via STATE messages */
2208 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2209 return rc;
2210
2211 /* Otherwise, be backwards compatible */
2212
2213 if (!more(peers_snd_nxt, l->rcv_nxt)) {
2214 l->nack_state = BC_NACK_SND_CONDITIONAL;
2215 return 0;
2216 }
2217
2218 /* Don't NACK if one was recently sent or peeked */
2219 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2220 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2221 return 0;
2222 }
2223
2224 /* Conditionally delay NACK sending until next synch rcv */
2225 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2226 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2227 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2228 return 0;
2229 }
2230
2231 /* Send NACK now but suppress next one */
2232 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2233 l->nack_state = BC_NACK_SND_SUPPRESS;
2234 return 0;
2235 }
2236
tipc_link_bc_ack_rcv(struct tipc_link * l,u16 acked,struct sk_buff_head * xmitq)2237 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
2238 struct sk_buff_head *xmitq)
2239 {
2240 struct sk_buff *skb, *tmp;
2241 struct tipc_link *snd_l = l->bc_sndlink;
2242
2243 if (!link_is_up(l) || !l->bc_peer_is_up)
2244 return;
2245
2246 if (!more(acked, l->acked))
2247 return;
2248
2249 trace_tipc_link_bc_ack(l, l->acked, acked, &snd_l->transmq);
2250 /* Skip over packets peer has already acked */
2251 skb_queue_walk(&snd_l->transmq, skb) {
2252 if (more(buf_seqno(skb), l->acked))
2253 break;
2254 }
2255
2256 /* Update/release the packets peer is acking now */
2257 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
2258 if (more(buf_seqno(skb), acked))
2259 break;
2260 if (!--TIPC_SKB_CB(skb)->ackers) {
2261 __skb_unlink(skb, &snd_l->transmq);
2262 kfree_skb(skb);
2263 }
2264 }
2265 l->acked = acked;
2266 tipc_link_advance_backlog(snd_l, xmitq);
2267 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
2268 link_prepare_wakeup(snd_l);
2269 }
2270
2271 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2272 * This function is here for backwards compatibility, since
2273 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2274 */
tipc_link_bc_nack_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)2275 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2276 struct sk_buff_head *xmitq)
2277 {
2278 struct tipc_msg *hdr = buf_msg(skb);
2279 u32 dnode = msg_destnode(hdr);
2280 int mtyp = msg_type(hdr);
2281 u16 acked = msg_bcast_ack(hdr);
2282 u16 from = acked + 1;
2283 u16 to = msg_bcgap_to(hdr);
2284 u16 peers_snd_nxt = to + 1;
2285 int rc = 0;
2286
2287 kfree_skb(skb);
2288
2289 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2290 return 0;
2291
2292 if (mtyp != STATE_MSG)
2293 return 0;
2294
2295 if (dnode == tipc_own_addr(l->net)) {
2296 tipc_link_bc_ack_rcv(l, acked, xmitq);
2297 rc = tipc_link_bc_retrans(l->bc_sndlink, l, from, to, xmitq);
2298 l->stats.recv_nacks++;
2299 return rc;
2300 }
2301
2302 /* Msg for other node => suppress own NACK at next sync if applicable */
2303 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2304 l->nack_state = BC_NACK_SND_SUPPRESS;
2305
2306 return 0;
2307 }
2308
tipc_link_set_queue_limits(struct tipc_link * l,u32 win)2309 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
2310 {
2311 int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2312
2313 l->window = win;
2314 l->backlog[TIPC_LOW_IMPORTANCE].limit = max_t(u16, 50, win);
2315 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = max_t(u16, 100, win * 2);
2316 l->backlog[TIPC_HIGH_IMPORTANCE].limit = max_t(u16, 150, win * 3);
2317 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
2318 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
2319 }
2320
2321 /**
2322 * link_reset_stats - reset link statistics
2323 * @l: pointer to link
2324 */
tipc_link_reset_stats(struct tipc_link * l)2325 void tipc_link_reset_stats(struct tipc_link *l)
2326 {
2327 memset(&l->stats, 0, sizeof(l->stats));
2328 }
2329
link_print(struct tipc_link * l,const char * str)2330 static void link_print(struct tipc_link *l, const char *str)
2331 {
2332 struct sk_buff *hskb = skb_peek(&l->transmq);
2333 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2334 u16 tail = l->snd_nxt - 1;
2335
2336 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2337 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2338 skb_queue_len(&l->transmq), head, tail,
2339 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2340 }
2341
2342 /* Parse and validate nested (link) properties valid for media, bearer and link
2343 */
tipc_nl_parse_link_prop(struct nlattr * prop,struct nlattr * props[])2344 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2345 {
2346 int err;
2347
2348 err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2349 tipc_nl_prop_policy, NULL);
2350 if (err)
2351 return err;
2352
2353 if (props[TIPC_NLA_PROP_PRIO]) {
2354 u32 prio;
2355
2356 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2357 if (prio > TIPC_MAX_LINK_PRI)
2358 return -EINVAL;
2359 }
2360
2361 if (props[TIPC_NLA_PROP_TOL]) {
2362 u32 tol;
2363
2364 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2365 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2366 return -EINVAL;
2367 }
2368
2369 if (props[TIPC_NLA_PROP_WIN]) {
2370 u32 win;
2371
2372 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2373 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
2374 return -EINVAL;
2375 }
2376
2377 return 0;
2378 }
2379
__tipc_nl_add_stats(struct sk_buff * skb,struct tipc_stats * s)2380 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2381 {
2382 int i;
2383 struct nlattr *stats;
2384
2385 struct nla_map {
2386 u32 key;
2387 u32 val;
2388 };
2389
2390 struct nla_map map[] = {
2391 {TIPC_NLA_STATS_RX_INFO, 0},
2392 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2393 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2394 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2395 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2396 {TIPC_NLA_STATS_TX_INFO, 0},
2397 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2398 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2399 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2400 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2401 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2402 s->msg_length_counts : 1},
2403 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2404 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2405 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2406 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2407 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2408 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2409 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2410 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2411 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2412 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
2413 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2414 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2415 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2416 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
2417 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2418 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2419 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2420 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2421 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2422 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2423 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2424 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2425 (s->accu_queue_sz / s->queue_sz_counts) : 0}
2426 };
2427
2428 stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2429 if (!stats)
2430 return -EMSGSIZE;
2431
2432 for (i = 0; i < ARRAY_SIZE(map); i++)
2433 if (nla_put_u32(skb, map[i].key, map[i].val))
2434 goto msg_full;
2435
2436 nla_nest_end(skb, stats);
2437
2438 return 0;
2439 msg_full:
2440 nla_nest_cancel(skb, stats);
2441
2442 return -EMSGSIZE;
2443 }
2444
2445 /* 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)2446 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2447 struct tipc_link *link, int nlflags)
2448 {
2449 u32 self = tipc_own_addr(net);
2450 struct nlattr *attrs;
2451 struct nlattr *prop;
2452 void *hdr;
2453 int err;
2454
2455 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2456 nlflags, TIPC_NL_LINK_GET);
2457 if (!hdr)
2458 return -EMSGSIZE;
2459
2460 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2461 if (!attrs)
2462 goto msg_full;
2463
2464 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2465 goto attr_msg_full;
2466 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2467 goto attr_msg_full;
2468 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2469 goto attr_msg_full;
2470 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2471 goto attr_msg_full;
2472 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2473 goto attr_msg_full;
2474
2475 if (tipc_link_is_up(link))
2476 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2477 goto attr_msg_full;
2478 if (link->active)
2479 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2480 goto attr_msg_full;
2481
2482 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2483 if (!prop)
2484 goto attr_msg_full;
2485 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2486 goto prop_msg_full;
2487 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2488 goto prop_msg_full;
2489 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2490 link->window))
2491 goto prop_msg_full;
2492 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2493 goto prop_msg_full;
2494 nla_nest_end(msg->skb, prop);
2495
2496 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2497 if (err)
2498 goto attr_msg_full;
2499
2500 nla_nest_end(msg->skb, attrs);
2501 genlmsg_end(msg->skb, hdr);
2502
2503 return 0;
2504
2505 prop_msg_full:
2506 nla_nest_cancel(msg->skb, prop);
2507 attr_msg_full:
2508 nla_nest_cancel(msg->skb, attrs);
2509 msg_full:
2510 genlmsg_cancel(msg->skb, hdr);
2511
2512 return -EMSGSIZE;
2513 }
2514
__tipc_nl_add_bc_link_stat(struct sk_buff * skb,struct tipc_stats * stats)2515 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2516 struct tipc_stats *stats)
2517 {
2518 int i;
2519 struct nlattr *nest;
2520
2521 struct nla_map {
2522 __u32 key;
2523 __u32 val;
2524 };
2525
2526 struct nla_map map[] = {
2527 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2528 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2529 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2530 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2531 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2532 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2533 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2534 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2535 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2536 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2537 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2538 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2539 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2540 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2541 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2542 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2543 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2544 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2545 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2546 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2547 };
2548
2549 nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2550 if (!nest)
2551 return -EMSGSIZE;
2552
2553 for (i = 0; i < ARRAY_SIZE(map); i++)
2554 if (nla_put_u32(skb, map[i].key, map[i].val))
2555 goto msg_full;
2556
2557 nla_nest_end(skb, nest);
2558
2559 return 0;
2560 msg_full:
2561 nla_nest_cancel(skb, nest);
2562
2563 return -EMSGSIZE;
2564 }
2565
tipc_nl_add_bc_link(struct net * net,struct tipc_nl_msg * msg)2566 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2567 {
2568 int err;
2569 void *hdr;
2570 struct nlattr *attrs;
2571 struct nlattr *prop;
2572 struct tipc_net *tn = net_generic(net, tipc_net_id);
2573 u32 bc_mode = tipc_bcast_get_broadcast_mode(net);
2574 u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2575 struct tipc_link *bcl = tn->bcl;
2576
2577 if (!bcl)
2578 return 0;
2579
2580 tipc_bcast_lock(net);
2581
2582 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2583 NLM_F_MULTI, TIPC_NL_LINK_GET);
2584 if (!hdr) {
2585 tipc_bcast_unlock(net);
2586 return -EMSGSIZE;
2587 }
2588
2589 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2590 if (!attrs)
2591 goto msg_full;
2592
2593 /* The broadcast link is always up */
2594 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2595 goto attr_msg_full;
2596
2597 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2598 goto attr_msg_full;
2599 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2600 goto attr_msg_full;
2601 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2602 goto attr_msg_full;
2603 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2604 goto attr_msg_full;
2605
2606 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2607 if (!prop)
2608 goto attr_msg_full;
2609 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2610 goto prop_msg_full;
2611 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2612 goto prop_msg_full;
2613 if (bc_mode & BCLINK_MODE_SEL)
2614 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2615 bc_ratio))
2616 goto prop_msg_full;
2617 nla_nest_end(msg->skb, prop);
2618
2619 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2620 if (err)
2621 goto attr_msg_full;
2622
2623 tipc_bcast_unlock(net);
2624 nla_nest_end(msg->skb, attrs);
2625 genlmsg_end(msg->skb, hdr);
2626
2627 return 0;
2628
2629 prop_msg_full:
2630 nla_nest_cancel(msg->skb, prop);
2631 attr_msg_full:
2632 nla_nest_cancel(msg->skb, attrs);
2633 msg_full:
2634 tipc_bcast_unlock(net);
2635 genlmsg_cancel(msg->skb, hdr);
2636
2637 return -EMSGSIZE;
2638 }
2639
tipc_link_set_tolerance(struct tipc_link * l,u32 tol,struct sk_buff_head * xmitq)2640 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2641 struct sk_buff_head *xmitq)
2642 {
2643 l->tolerance = tol;
2644 if (l->bc_rcvlink)
2645 l->bc_rcvlink->tolerance = tol;
2646 if (link_is_up(l))
2647 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2648 }
2649
tipc_link_set_prio(struct tipc_link * l,u32 prio,struct sk_buff_head * xmitq)2650 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2651 struct sk_buff_head *xmitq)
2652 {
2653 l->priority = prio;
2654 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2655 }
2656
tipc_link_set_abort_limit(struct tipc_link * l,u32 limit)2657 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2658 {
2659 l->abort_limit = limit;
2660 }
2661
tipc_link_name_ext(struct tipc_link * l,char * buf)2662 char *tipc_link_name_ext(struct tipc_link *l, char *buf)
2663 {
2664 if (!l)
2665 scnprintf(buf, TIPC_MAX_LINK_NAME, "null");
2666 else if (link_is_bc_sndlink(l))
2667 scnprintf(buf, TIPC_MAX_LINK_NAME, "broadcast-sender");
2668 else if (link_is_bc_rcvlink(l))
2669 scnprintf(buf, TIPC_MAX_LINK_NAME,
2670 "broadcast-receiver, peer %x", l->addr);
2671 else
2672 memcpy(buf, l->name, TIPC_MAX_LINK_NAME);
2673
2674 return buf;
2675 }
2676
2677 /**
2678 * tipc_link_dump - dump TIPC link data
2679 * @l: tipc link to be dumped
2680 * @dqueues: bitmask to decide if any link queue to be dumped?
2681 * - TIPC_DUMP_NONE: don't dump link queues
2682 * - TIPC_DUMP_TRANSMQ: dump link transmq queue
2683 * - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2684 * - TIPC_DUMP_DEFERDQ: dump link deferd queue
2685 * - TIPC_DUMP_INPUTQ: dump link input queue
2686 * - TIPC_DUMP_WAKEUP: dump link wakeup queue
2687 * - TIPC_DUMP_ALL: dump all the link queues above
2688 * @buf: returned buffer of dump data in format
2689 */
tipc_link_dump(struct tipc_link * l,u16 dqueues,char * buf)2690 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2691 {
2692 int i = 0;
2693 size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2694 struct sk_buff_head *list;
2695 struct sk_buff *hskb, *tskb;
2696 u32 len;
2697
2698 if (!l) {
2699 i += scnprintf(buf, sz, "link data: (null)\n");
2700 return i;
2701 }
2702
2703 i += scnprintf(buf, sz, "link data: %x", l->addr);
2704 i += scnprintf(buf + i, sz - i, " %x", l->state);
2705 i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2706 i += scnprintf(buf + i, sz - i, " %u", l->session);
2707 i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2708 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2709 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2710 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2711 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2712 i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2713 i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2714 i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2715 i += scnprintf(buf + i, sz - i, " %u", 0);
2716 i += scnprintf(buf + i, sz - i, " %u", 0);
2717 i += scnprintf(buf + i, sz - i, " %u", l->acked);
2718
2719 list = &l->transmq;
2720 len = skb_queue_len(list);
2721 hskb = skb_peek(list);
2722 tskb = skb_peek_tail(list);
2723 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2724 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2725 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2726
2727 list = &l->deferdq;
2728 len = skb_queue_len(list);
2729 hskb = skb_peek(list);
2730 tskb = skb_peek_tail(list);
2731 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2732 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2733 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2734
2735 list = &l->backlogq;
2736 len = skb_queue_len(list);
2737 hskb = skb_peek(list);
2738 tskb = skb_peek_tail(list);
2739 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2740 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2741 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2742
2743 list = l->inputq;
2744 len = skb_queue_len(list);
2745 hskb = skb_peek(list);
2746 tskb = skb_peek_tail(list);
2747 i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2748 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2749 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2750
2751 if (dqueues & TIPC_DUMP_TRANSMQ) {
2752 i += scnprintf(buf + i, sz - i, "transmq: ");
2753 i += tipc_list_dump(&l->transmq, false, buf + i);
2754 }
2755 if (dqueues & TIPC_DUMP_BACKLOGQ) {
2756 i += scnprintf(buf + i, sz - i,
2757 "backlogq: <%u %u %u %u %u>, ",
2758 l->backlog[TIPC_LOW_IMPORTANCE].len,
2759 l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2760 l->backlog[TIPC_HIGH_IMPORTANCE].len,
2761 l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2762 l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2763 i += tipc_list_dump(&l->backlogq, false, buf + i);
2764 }
2765 if (dqueues & TIPC_DUMP_DEFERDQ) {
2766 i += scnprintf(buf + i, sz - i, "deferdq: ");
2767 i += tipc_list_dump(&l->deferdq, false, buf + i);
2768 }
2769 if (dqueues & TIPC_DUMP_INPUTQ) {
2770 i += scnprintf(buf + i, sz - i, "inputq: ");
2771 i += tipc_list_dump(l->inputq, false, buf + i);
2772 }
2773 if (dqueues & TIPC_DUMP_WAKEUP) {
2774 i += scnprintf(buf + i, sz - i, "wakeup: ");
2775 i += tipc_list_dump(&l->wakeupq, false, buf + i);
2776 }
2777
2778 return i;
2779 }
2780