projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
tipc: use pseudo message to wake up sockets after link congestion
[deliverable/linux.git] / net / tipc / link.c
1 /*
2 * net/tipc/link.c: TIPC link code
3 *
4 * Copyright (c) 1996-2007, 2012-2014, 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 "link.h"
39 #include "port.h"
40 #include "socket.h"
41 #include "name_distr.h"
42 #include "discover.h"
43 #include "config.h"
44
45 #include <linux/pkt_sched.h>
46
47 /*
48 * Error message prefixes
49 */
50 static const char *link_co_err = "Link changeover error, ";
51 static const char *link_rst_msg = "Resetting link ";
52 static const char *link_unk_evt = "Unknown link event ";
53
54 /*
55 * Out-of-range value for link session numbers
56 */
57 #define INVALID_SESSION 0x10000
58
59 /*
60 * Link state events:
61 */
62 #define STARTING_EVT 856384768 /* link processing trigger */
63 #define TRAFFIC_MSG_EVT 560815u /* rx'd ??? */
64 #define TIMEOUT_EVT 560817u /* link timer expired */
65
66 /*
67 * The following two 'message types' is really just implementation
68 * data conveniently stored in the message header.
69 * They must not be considered part of the protocol
70 */
71 #define OPEN_MSG 0
72 #define CLOSED_MSG 1
73
74 /*
75 * State value stored in 'exp_msg_count'
76 */
77 #define START_CHANGEOVER 100000u
78
79 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
80 struct sk_buff *buf);
81 static void tipc_link_proto_rcv(struct tipc_link *l_ptr, struct sk_buff *buf);
82 static int tipc_link_tunnel_rcv(struct tipc_node *n_ptr,
83 struct sk_buff **buf);
84 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance);
85 static void link_state_event(struct tipc_link *l_ptr, u32 event);
86 static void link_reset_statistics(struct tipc_link *l_ptr);
87 static void link_print(struct tipc_link *l_ptr, const char *str);
88 static void tipc_link_sync_xmit(struct tipc_link *l);
89 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf);
90 static int tipc_link_input(struct tipc_link *l, struct sk_buff *buf);
91 static int tipc_link_prepare_input(struct tipc_link *l, struct sk_buff **buf);
92
93 /*
94 * Simple link routines
95 */
96 static unsigned int align(unsigned int i)
97 {
98 return (i + 3) & ~3u;
99 }
100
101 static void link_init_max_pkt(struct tipc_link *l_ptr)
102 {
103 struct tipc_bearer *b_ptr;
104 u32 max_pkt;
105
106 rcu_read_lock();
107 b_ptr = rcu_dereference_rtnl(bearer_list[l_ptr->bearer_id]);
108 if (!b_ptr) {
109 rcu_read_unlock();
110 return;
111 }
112 max_pkt = (b_ptr->mtu & ~3);
113 rcu_read_unlock();
114
115 if (max_pkt > MAX_MSG_SIZE)
116 max_pkt = MAX_MSG_SIZE;
117
118 l_ptr->max_pkt_target = max_pkt;
119 if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
120 l_ptr->max_pkt = l_ptr->max_pkt_target;
121 else
122 l_ptr->max_pkt = MAX_PKT_DEFAULT;
123
124 l_ptr->max_pkt_probes = 0;
125 }
126
127 static u32 link_next_sent(struct tipc_link *l_ptr)
128 {
129 if (l_ptr->next_out)
130 return buf_seqno(l_ptr->next_out);
131 return mod(l_ptr->next_out_no);
132 }
133
134 static u32 link_last_sent(struct tipc_link *l_ptr)
135 {
136 return mod(link_next_sent(l_ptr) - 1);
137 }
138
139 /*
140 * Simple non-static link routines (i.e. referenced outside this file)
141 */
142 int tipc_link_is_up(struct tipc_link *l_ptr)
143 {
144 if (!l_ptr)
145 return 0;
146 return link_working_working(l_ptr) || link_working_unknown(l_ptr);
147 }
148
149 int tipc_link_is_active(struct tipc_link *l_ptr)
150 {
151 return (l_ptr->owner->active_links[0] == l_ptr) ||
152 (l_ptr->owner->active_links[1] == l_ptr);
153 }
154
155 /**
156 * link_timeout - handle expiration of link timer
157 * @l_ptr: pointer to link
158 */
159 static void link_timeout(struct tipc_link *l_ptr)
160 {
161 tipc_node_lock(l_ptr->owner);
162
163 /* update counters used in statistical profiling of send traffic */
164 l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
165 l_ptr->stats.queue_sz_counts++;
166
167 if (l_ptr->first_out) {
168 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
169 u32 length = msg_size(msg);
170
171 if ((msg_user(msg) == MSG_FRAGMENTER) &&
172 (msg_type(msg) == FIRST_FRAGMENT)) {
173 length = msg_size(msg_get_wrapped(msg));
174 }
175 if (length) {
176 l_ptr->stats.msg_lengths_total += length;
177 l_ptr->stats.msg_length_counts++;
178 if (length <= 64)
179 l_ptr->stats.msg_length_profile[0]++;
180 else if (length <= 256)
181 l_ptr->stats.msg_length_profile[1]++;
182 else if (length <= 1024)
183 l_ptr->stats.msg_length_profile[2]++;
184 else if (length <= 4096)
185 l_ptr->stats.msg_length_profile[3]++;
186 else if (length <= 16384)
187 l_ptr->stats.msg_length_profile[4]++;
188 else if (length <= 32768)
189 l_ptr->stats.msg_length_profile[5]++;
190 else
191 l_ptr->stats.msg_length_profile[6]++;
192 }
193 }
194
195 /* do all other link processing performed on a periodic basis */
196
197 link_state_event(l_ptr, TIMEOUT_EVT);
198
199 if (l_ptr->next_out)
200 tipc_link_push_queue(l_ptr);
201
202 tipc_node_unlock(l_ptr->owner);
203 }
204
205 static void link_set_timer(struct tipc_link *l_ptr, u32 time)
206 {
207 k_start_timer(&l_ptr->timer, time);
208 }
209
210 /**
211 * tipc_link_create - create a new link
212 * @n_ptr: pointer to associated node
213 * @b_ptr: pointer to associated bearer
214 * @media_addr: media address to use when sending messages over link
215 *
216 * Returns pointer to link.
217 */
218 struct tipc_link *tipc_link_create(struct tipc_node *n_ptr,
219 struct tipc_bearer *b_ptr,
220 const struct tipc_media_addr *media_addr)
221 {
222 struct tipc_link *l_ptr;
223 struct tipc_msg *msg;
224 char *if_name;
225 char addr_string[16];
226 u32 peer = n_ptr->addr;
227
228 if (n_ptr->link_cnt >= 2) {
229 tipc_addr_string_fill(addr_string, n_ptr->addr);
230 pr_err("Attempt to establish third link to %s\n", addr_string);
231 return NULL;
232 }
233
234 if (n_ptr->links[b_ptr->identity]) {
235 tipc_addr_string_fill(addr_string, n_ptr->addr);
236 pr_err("Attempt to establish second link on <%s> to %s\n",
237 b_ptr->name, addr_string);
238 return NULL;
239 }
240
241 l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
242 if (!l_ptr) {
243 pr_warn("Link creation failed, no memory\n");
244 return NULL;
245 }
246
247 l_ptr->addr = peer;
248 if_name = strchr(b_ptr->name, ':') + 1;
249 sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
250 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
251 tipc_node(tipc_own_addr),
252 if_name,
253 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
254 /* note: peer i/f name is updated by reset/activate message */
255 memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
256 l_ptr->owner = n_ptr;
257 l_ptr->checkpoint = 1;
258 l_ptr->peer_session = INVALID_SESSION;
259 l_ptr->bearer_id = b_ptr->identity;
260 link_set_supervision_props(l_ptr, b_ptr->tolerance);
261 l_ptr->state = RESET_UNKNOWN;
262
263 l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
264 msg = l_ptr->pmsg;
265 tipc_msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
266 msg_set_size(msg, sizeof(l_ptr->proto_msg));
267 msg_set_session(msg, (tipc_random & 0xffff));
268 msg_set_bearer_id(msg, b_ptr->identity);
269 strcpy((char *)msg_data(msg), if_name);
270
271 l_ptr->priority = b_ptr->priority;
272 tipc_link_set_queue_limits(l_ptr, b_ptr->window);
273
274 l_ptr->net_plane = b_ptr->net_plane;
275 link_init_max_pkt(l_ptr);
276
277 l_ptr->next_out_no = 1;
278 __skb_queue_head_init(&l_ptr->waiting_sks);
279
280 link_reset_statistics(l_ptr);
281
282 tipc_node_attach_link(n_ptr, l_ptr);
283
284 k_init_timer(&l_ptr->timer, (Handler)link_timeout,
285 (unsigned long)l_ptr);
286
287 link_state_event(l_ptr, STARTING_EVT);
288
289 return l_ptr;
290 }
291
292 void tipc_link_delete_list(unsigned int bearer_id, bool shutting_down)
293 {
294 struct tipc_link *l_ptr;
295 struct tipc_node *n_ptr;
296
297 rcu_read_lock();
298 list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
299 tipc_node_lock(n_ptr);
300 l_ptr = n_ptr->links[bearer_id];
301 if (l_ptr) {
302 tipc_link_reset(l_ptr);
303 if (shutting_down || !tipc_node_is_up(n_ptr)) {
304 tipc_node_detach_link(l_ptr->owner, l_ptr);
305 tipc_link_reset_fragments(l_ptr);
306 tipc_node_unlock(n_ptr);
307
308 /* Nobody else can access this link now: */
309 del_timer_sync(&l_ptr->timer);
310 kfree(l_ptr);
311 } else {
312 /* Detach/delete when failover is finished: */
313 l_ptr->flags |= LINK_STOPPED;
314 tipc_node_unlock(n_ptr);
315 del_timer_sync(&l_ptr->timer);
316 }
317 continue;
318 }
319 tipc_node_unlock(n_ptr);
320 }
321 rcu_read_unlock();
322 }
323
324 /**
325 * link_schedule_user - schedule user for wakeup after congestion
326 * @link: congested link
327 * @oport: sending port
328 * @chain_sz: size of buffer chain that was attempted sent
329 * @imp: importance of message attempted sent
330 * Create pseudo msg to send back to user when congestion abates
331 */
332 static bool link_schedule_user(struct tipc_link *link, u32 oport,
333 uint chain_sz, uint imp)
334 {
335 struct sk_buff *buf;
336
337 buf = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0, tipc_own_addr,
338 tipc_own_addr, oport, 0, 0);
339 if (!buf)
340 return false;
341 TIPC_SKB_CB(buf)->chain_sz = chain_sz;
342 TIPC_SKB_CB(buf)->chain_imp = imp;
343 __skb_queue_tail(&link->waiting_sks, buf);
344 link->stats.link_congs++;
345 return true;
346 }
347
348 /**
349 * link_prepare_wakeup - prepare users for wakeup after congestion
350 * @link: congested link
351 * Move a number of waiting users, as permitted by available space in
352 * the send queue, from link wait queue to node wait queue for wakeup
353 */
354 static void link_prepare_wakeup(struct tipc_link *link)
355 {
356 struct sk_buff_head *wq = &link->waiting_sks;
357 struct sk_buff *buf;
358 uint pend_qsz = link->out_queue_size;
359
360 for (buf = skb_peek(wq); buf; buf = skb_peek(wq)) {
361 if (pend_qsz >= link->queue_limit[TIPC_SKB_CB(buf)->chain_imp])
362 break;
363 pend_qsz += TIPC_SKB_CB(buf)->chain_sz;
364 __skb_queue_tail(&link->owner->waiting_sks, __skb_dequeue(wq));
365 }
366 }
367
368 /**
369 * link_release_outqueue - purge link's outbound message queue
370 * @l_ptr: pointer to link
371 */
372 static void link_release_outqueue(struct tipc_link *l_ptr)
373 {
374 kfree_skb_list(l_ptr->first_out);
375 l_ptr->first_out = NULL;
376 l_ptr->out_queue_size = 0;
377 }
378
379 /**
380 * tipc_link_reset_fragments - purge link's inbound message fragments queue
381 * @l_ptr: pointer to link
382 */
383 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
384 {
385 kfree_skb(l_ptr->reasm_buf);
386 l_ptr->reasm_buf = NULL;
387 }
388
389 /**
390 * tipc_link_purge_queues - purge all pkt queues associated with link
391 * @l_ptr: pointer to link
392 */
393 void tipc_link_purge_queues(struct tipc_link *l_ptr)
394 {
395 kfree_skb_list(l_ptr->oldest_deferred_in);
396 kfree_skb_list(l_ptr->first_out);
397 tipc_link_reset_fragments(l_ptr);
398 kfree_skb(l_ptr->proto_msg_queue);
399 l_ptr->proto_msg_queue = NULL;
400 }
401
402 void tipc_link_reset(struct tipc_link *l_ptr)
403 {
404 u32 prev_state = l_ptr->state;
405 u32 checkpoint = l_ptr->next_in_no;
406 int was_active_link = tipc_link_is_active(l_ptr);
407 struct tipc_node *owner = l_ptr->owner;
408
409 msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
410
411 /* Link is down, accept any session */
412 l_ptr->peer_session = INVALID_SESSION;
413
414 /* Prepare for max packet size negotiation */
415 link_init_max_pkt(l_ptr);
416
417 l_ptr->state = RESET_UNKNOWN;
418
419 if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
420 return;
421
422 tipc_node_link_down(l_ptr->owner, l_ptr);
423 tipc_bearer_remove_dest(l_ptr->bearer_id, l_ptr->addr);
424
425 if (was_active_link && tipc_node_active_links(l_ptr->owner)) {
426 l_ptr->reset_checkpoint = checkpoint;
427 l_ptr->exp_msg_count = START_CHANGEOVER;
428 }
429
430 /* Clean up all queues: */
431 link_release_outqueue(l_ptr);
432 kfree_skb(l_ptr->proto_msg_queue);
433 l_ptr->proto_msg_queue = NULL;
434 kfree_skb_list(l_ptr->oldest_deferred_in);
435 if (!skb_queue_empty(&l_ptr->waiting_sks)) {
436 skb_queue_splice_init(&l_ptr->waiting_sks, &owner->waiting_sks);
437 owner->action_flags |= TIPC_WAKEUP_USERS;
438 }
439 l_ptr->retransm_queue_head = 0;
440 l_ptr->retransm_queue_size = 0;
441 l_ptr->last_out = NULL;
442 l_ptr->first_out = NULL;
443 l_ptr->next_out = NULL;
444 l_ptr->unacked_window = 0;
445 l_ptr->checkpoint = 1;
446 l_ptr->next_out_no = 1;
447 l_ptr->deferred_inqueue_sz = 0;
448 l_ptr->oldest_deferred_in = NULL;
449 l_ptr->newest_deferred_in = NULL;
450 l_ptr->fsm_msg_cnt = 0;
451 l_ptr->stale_count = 0;
452 link_reset_statistics(l_ptr);
453 }
454
455 void tipc_link_reset_list(unsigned int bearer_id)
456 {
457 struct tipc_link *l_ptr;
458 struct tipc_node *n_ptr;
459
460 rcu_read_lock();
461 list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
462 tipc_node_lock(n_ptr);
463 l_ptr = n_ptr->links[bearer_id];
464 if (l_ptr)
465 tipc_link_reset(l_ptr);
466 tipc_node_unlock(n_ptr);
467 }
468 rcu_read_unlock();
469 }
470
471 static void link_activate(struct tipc_link *l_ptr)
472 {
473 l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
474 tipc_node_link_up(l_ptr->owner, l_ptr);
475 tipc_bearer_add_dest(l_ptr->bearer_id, l_ptr->addr);
476 }
477
478 /**
479 * link_state_event - link finite state machine
480 * @l_ptr: pointer to link
481 * @event: state machine event to process
482 */
483 static void link_state_event(struct tipc_link *l_ptr, unsigned int event)
484 {
485 struct tipc_link *other;
486 u32 cont_intv = l_ptr->continuity_interval;
487
488 if (l_ptr->flags & LINK_STOPPED)
489 return;
490
491 if (!(l_ptr->flags & LINK_STARTED) && (event != STARTING_EVT))
492 return; /* Not yet. */
493
494 /* Check whether changeover is going on */
495 if (l_ptr->exp_msg_count) {
496 if (event == TIMEOUT_EVT)
497 link_set_timer(l_ptr, cont_intv);
498 return;
499 }
500
501 switch (l_ptr->state) {
502 case WORKING_WORKING:
503 switch (event) {
504 case TRAFFIC_MSG_EVT:
505 case ACTIVATE_MSG:
506 break;
507 case TIMEOUT_EVT:
508 if (l_ptr->next_in_no != l_ptr->checkpoint) {
509 l_ptr->checkpoint = l_ptr->next_in_no;
510 if (tipc_bclink_acks_missing(l_ptr->owner)) {
511 tipc_link_proto_xmit(l_ptr, STATE_MSG,
512 0, 0, 0, 0, 0);
513 l_ptr->fsm_msg_cnt++;
514 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
515 tipc_link_proto_xmit(l_ptr, STATE_MSG,
516 1, 0, 0, 0, 0);
517 l_ptr->fsm_msg_cnt++;
518 }
519 link_set_timer(l_ptr, cont_intv);
520 break;
521 }
522 l_ptr->state = WORKING_UNKNOWN;
523 l_ptr->fsm_msg_cnt = 0;
524 tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
525 l_ptr->fsm_msg_cnt++;
526 link_set_timer(l_ptr, cont_intv / 4);
527 break;
528 case RESET_MSG:
529 pr_info("%s<%s>, requested by peer\n", link_rst_msg,
530 l_ptr->name);
531 tipc_link_reset(l_ptr);
532 l_ptr->state = RESET_RESET;
533 l_ptr->fsm_msg_cnt = 0;
534 tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
535 0, 0, 0, 0, 0);
536 l_ptr->fsm_msg_cnt++;
537 link_set_timer(l_ptr, cont_intv);
538 break;
539 default:
540 pr_err("%s%u in WW state\n", link_unk_evt, event);
541 }
542 break;
543 case WORKING_UNKNOWN:
544 switch (event) {
545 case TRAFFIC_MSG_EVT:
546 case ACTIVATE_MSG:
547 l_ptr->state = WORKING_WORKING;
548 l_ptr->fsm_msg_cnt = 0;
549 link_set_timer(l_ptr, cont_intv);
550 break;
551 case RESET_MSG:
552 pr_info("%s<%s>, requested by peer while probing\n",
553 link_rst_msg, l_ptr->name);
554 tipc_link_reset(l_ptr);
555 l_ptr->state = RESET_RESET;
556 l_ptr->fsm_msg_cnt = 0;
557 tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
558 0, 0, 0, 0, 0);
559 l_ptr->fsm_msg_cnt++;
560 link_set_timer(l_ptr, cont_intv);
561 break;
562 case TIMEOUT_EVT:
563 if (l_ptr->next_in_no != l_ptr->checkpoint) {
564 l_ptr->state = WORKING_WORKING;
565 l_ptr->fsm_msg_cnt = 0;
566 l_ptr->checkpoint = l_ptr->next_in_no;
567 if (tipc_bclink_acks_missing(l_ptr->owner)) {
568 tipc_link_proto_xmit(l_ptr, STATE_MSG,
569 0, 0, 0, 0, 0);
570 l_ptr->fsm_msg_cnt++;
571 }
572 link_set_timer(l_ptr, cont_intv);
573 } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
574 tipc_link_proto_xmit(l_ptr, STATE_MSG,
575 1, 0, 0, 0, 0);
576 l_ptr->fsm_msg_cnt++;
577 link_set_timer(l_ptr, cont_intv / 4);
578 } else { /* Link has failed */
579 pr_warn("%s<%s>, peer not responding\n",
580 link_rst_msg, l_ptr->name);
581 tipc_link_reset(l_ptr);
582 l_ptr->state = RESET_UNKNOWN;
583 l_ptr->fsm_msg_cnt = 0;
584 tipc_link_proto_xmit(l_ptr, RESET_MSG,
585 0, 0, 0, 0, 0);
586 l_ptr->fsm_msg_cnt++;
587 link_set_timer(l_ptr, cont_intv);
588 }
589 break;
590 default:
591 pr_err("%s%u in WU state\n", link_unk_evt, event);
592 }
593 break;
594 case RESET_UNKNOWN:
595 switch (event) {
596 case TRAFFIC_MSG_EVT:
597 break;
598 case ACTIVATE_MSG:
599 other = l_ptr->owner->active_links[0];
600 if (other && link_working_unknown(other))
601 break;
602 l_ptr->state = WORKING_WORKING;
603 l_ptr->fsm_msg_cnt = 0;
604 link_activate(l_ptr);
605 tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
606 l_ptr->fsm_msg_cnt++;
607 if (l_ptr->owner->working_links == 1)
608 tipc_link_sync_xmit(l_ptr);
609 link_set_timer(l_ptr, cont_intv);
610 break;
611 case RESET_MSG:
612 l_ptr->state = RESET_RESET;
613 l_ptr->fsm_msg_cnt = 0;
614 tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
615 1, 0, 0, 0, 0);
616 l_ptr->fsm_msg_cnt++;
617 link_set_timer(l_ptr, cont_intv);
618 break;
619 case STARTING_EVT:
620 l_ptr->flags |= LINK_STARTED;
621 /* fall through */
622 case TIMEOUT_EVT:
623 tipc_link_proto_xmit(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
624 l_ptr->fsm_msg_cnt++;
625 link_set_timer(l_ptr, cont_intv);
626 break;
627 default:
628 pr_err("%s%u in RU state\n", link_unk_evt, event);
629 }
630 break;
631 case RESET_RESET:
632 switch (event) {
633 case TRAFFIC_MSG_EVT:
634 case ACTIVATE_MSG:
635 other = l_ptr->owner->active_links[0];
636 if (other && link_working_unknown(other))
637 break;
638 l_ptr->state = WORKING_WORKING;
639 l_ptr->fsm_msg_cnt = 0;
640 link_activate(l_ptr);
641 tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
642 l_ptr->fsm_msg_cnt++;
643 if (l_ptr->owner->working_links == 1)
644 tipc_link_sync_xmit(l_ptr);
645 link_set_timer(l_ptr, cont_intv);
646 break;
647 case RESET_MSG:
648 break;
649 case TIMEOUT_EVT:
650 tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
651 0, 0, 0, 0, 0);
652 l_ptr->fsm_msg_cnt++;
653 link_set_timer(l_ptr, cont_intv);
654 break;
655 default:
656 pr_err("%s%u in RR state\n", link_unk_evt, event);
657 }
658 break;
659 default:
660 pr_err("Unknown link state %u/%u\n", l_ptr->state, event);
661 }
662 }
663
664 /* tipc_link_cong: determine return value and how to treat the
665 * sent buffer during link congestion.
666 * - For plain, errorless user data messages we keep the buffer and
667 * return -ELINKONG.
668 * - For all other messages we discard the buffer and return -EHOSTUNREACH
669 * - For TIPC internal messages we also reset the link
670 */
671 static int tipc_link_cong(struct tipc_link *link, struct sk_buff *buf)
672 {
673 struct tipc_msg *msg = buf_msg(buf);
674 uint imp = tipc_msg_tot_importance(msg);
675 u32 oport = msg_tot_origport(msg);
676
677 if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
678 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
679 tipc_link_reset(link);
680 goto drop;
681 }
682 if (unlikely(msg_errcode(msg)))
683 goto drop;
684 if (unlikely(msg_reroute_cnt(msg)))
685 goto drop;
686 if (TIPC_SKB_CB(buf)->wakeup_pending)
687 return -ELINKCONG;
688 if (link_schedule_user(link, oport, TIPC_SKB_CB(buf)->chain_sz, imp))
689 return -ELINKCONG;
690 drop:
691 kfree_skb_list(buf);
692 return -EHOSTUNREACH;
693 }
694
695 /**
696 * __tipc_link_xmit(): same as tipc_link_xmit, but destlink is known & locked
697 * @link: link to use
698 * @buf: chain of buffers containing message
699 * Consumes the buffer chain, except when returning -ELINKCONG
700 * Returns 0 if success, otherwise errno: -ELINKCONG, -EMSGSIZE (plain socket
701 * user data messages) or -EHOSTUNREACH (all other messages/senders)
702 * Only the socket functions tipc_send_stream() and tipc_send_packet() need
703 * to act on the return value, since they may need to do more send attempts.
704 */
705 int __tipc_link_xmit(struct tipc_link *link, struct sk_buff *buf)
706 {
707 struct tipc_msg *msg = buf_msg(buf);
708 uint psz = msg_size(msg);
709 uint qsz = link->out_queue_size;
710 uint sndlim = link->queue_limit[0];
711 uint imp = tipc_msg_tot_importance(msg);
712 uint mtu = link->max_pkt;
713 uint ack = mod(link->next_in_no - 1);
714 uint seqno = link->next_out_no;
715 uint bc_last_in = link->owner->bclink.last_in;
716 struct tipc_media_addr *addr = &link->media_addr;
717 struct sk_buff *next = buf->next;
718
719 /* Match queue limits against msg importance: */
720 if (unlikely(qsz >= link->queue_limit[imp]))
721 return tipc_link_cong(link, buf);
722
723 /* Has valid packet limit been used ? */
724 if (unlikely(psz > mtu)) {
725 kfree_skb_list(buf);
726 return -EMSGSIZE;
727 }
728
729 /* Prepare each packet for sending, and add to outqueue: */
730 while (buf) {
731 next = buf->next;
732 msg = buf_msg(buf);
733 msg_set_word(msg, 2, ((ack << 16) | mod(seqno)));
734 msg_set_bcast_ack(msg, bc_last_in);
735
736 if (!link->first_out) {
737 link->first_out = buf;
738 } else if (qsz < sndlim) {
739 link->last_out->next = buf;
740 } else if (tipc_msg_bundle(link->last_out, buf, mtu)) {
741 link->stats.sent_bundled++;
742 buf = next;
743 next = buf->next;
744 continue;
745 } else if (tipc_msg_make_bundle(&buf, mtu, link->addr)) {
746 link->stats.sent_bundled++;
747 link->stats.sent_bundles++;
748 link->last_out->next = buf;
749 if (!link->next_out)
750 link->next_out = buf;
751 } else {
752 link->last_out->next = buf;
753 if (!link->next_out)
754 link->next_out = buf;
755 }
756
757 /* Send packet if possible: */
758 if (likely(++qsz <= sndlim)) {
759 tipc_bearer_send(link->bearer_id, buf, addr);
760 link->next_out = next;
761 link->unacked_window = 0;
762 }
763 seqno++;
764 link->last_out = buf;
765 buf = next;
766 }
767 link->next_out_no = seqno;
768 link->out_queue_size = qsz;
769 return 0;
770 }
771
772 /**
773 * tipc_link_xmit() is the general link level function for message sending
774 * @buf: chain of buffers containing message
775 * @dsz: amount of user data to be sent
776 * @dnode: address of destination node
777 * @selector: a number used for deterministic link selection
778 * Consumes the buffer chain, except when returning -ELINKCONG
779 * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
780 */
781 int tipc_link_xmit(struct sk_buff *buf, u32 dnode, u32 selector)
782 {
783 struct tipc_link *link = NULL;
784 struct tipc_node *node;
785 int rc = -EHOSTUNREACH;
786
787 node = tipc_node_find(dnode);
788 if (node) {
789 tipc_node_lock(node);
790 link = node->active_links[selector & 1];
791 if (link)
792 rc = __tipc_link_xmit(link, buf);
793 tipc_node_unlock(node);
794 }
795
796 if (link)
797 return rc;
798
799 if (likely(in_own_node(dnode)))
800 return tipc_sk_rcv(buf);
801
802 kfree_skb_list(buf);
803 return rc;
804 }
805
806 /*
807 * tipc_link_sync_xmit - synchronize broadcast link endpoints.
808 *
809 * Give a newly added peer node the sequence number where it should
810 * start receiving and acking broadcast packets.
811 *
812 * Called with node locked
813 */
814 static void tipc_link_sync_xmit(struct tipc_link *link)
815 {
816 struct sk_buff *buf;
817 struct tipc_msg *msg;
818
819 buf = tipc_buf_acquire(INT_H_SIZE);
820 if (!buf)
821 return;
822
823 msg = buf_msg(buf);
824 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE, link->addr);
825 msg_set_last_bcast(msg, link->owner->bclink.acked);
826 __tipc_link_xmit(link, buf);
827 }
828
829 /*
830 * tipc_link_sync_rcv - synchronize broadcast link endpoints.
831 * Receive the sequence number where we should start receiving and
832 * acking broadcast packets from a newly added peer node, and open
833 * up for reception of such packets.
834 *
835 * Called with node locked
836 */
837 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf)
838 {
839 struct tipc_msg *msg = buf_msg(buf);
840
841 n->bclink.last_sent = n->bclink.last_in = msg_last_bcast(msg);
842 n->bclink.recv_permitted = true;
843 kfree_skb(buf);
844 }
845
846 /*
847 * tipc_link_push_packet: Push one unsent packet to the media
848 */
849 static u32 tipc_link_push_packet(struct tipc_link *l_ptr)
850 {
851 struct sk_buff *buf = l_ptr->first_out;
852 u32 r_q_size = l_ptr->retransm_queue_size;
853 u32 r_q_head = l_ptr->retransm_queue_head;
854
855 /* Step to position where retransmission failed, if any, */
856 /* consider that buffers may have been released in meantime */
857 if (r_q_size && buf) {
858 u32 last = lesser(mod(r_q_head + r_q_size),
859 link_last_sent(l_ptr));
860 u32 first = buf_seqno(buf);
861
862 while (buf && less(first, r_q_head)) {
863 first = mod(first + 1);
864 buf = buf->next;
865 }
866 l_ptr->retransm_queue_head = r_q_head = first;
867 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
868 }
869
870 /* Continue retransmission now, if there is anything: */
871 if (r_q_size && buf) {
872 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
873 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
874 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
875 l_ptr->retransm_queue_head = mod(++r_q_head);
876 l_ptr->retransm_queue_size = --r_q_size;
877 l_ptr->stats.retransmitted++;
878 return 0;
879 }
880
881 /* Send deferred protocol message, if any: */
882 buf = l_ptr->proto_msg_queue;
883 if (buf) {
884 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
885 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
886 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
887 l_ptr->unacked_window = 0;
888 kfree_skb(buf);
889 l_ptr->proto_msg_queue = NULL;
890 return 0;
891 }
892
893 /* Send one deferred data message, if send window not full: */
894 buf = l_ptr->next_out;
895 if (buf) {
896 struct tipc_msg *msg = buf_msg(buf);
897 u32 next = msg_seqno(msg);
898 u32 first = buf_seqno(l_ptr->first_out);
899
900 if (mod(next - first) < l_ptr->queue_limit[0]) {
901 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
902 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
903 tipc_bearer_send(l_ptr->bearer_id, buf,
904 &l_ptr->media_addr);
905 if (msg_user(msg) == MSG_BUNDLER)
906 msg_set_type(msg, BUNDLE_CLOSED);
907 l_ptr->next_out = buf->next;
908 return 0;
909 }
910 }
911 return 1;
912 }
913
914 /*
915 * push_queue(): push out the unsent messages of a link where
916 * congestion has abated. Node is locked
917 */
918 void tipc_link_push_queue(struct tipc_link *l_ptr)
919 {
920 u32 res;
921
922 do {
923 res = tipc_link_push_packet(l_ptr);
924 } while (!res);
925 }
926
927 void tipc_link_reset_all(struct tipc_node *node)
928 {
929 char addr_string[16];
930 u32 i;
931
932 tipc_node_lock(node);
933
934 pr_warn("Resetting all links to %s\n",
935 tipc_addr_string_fill(addr_string, node->addr));
936
937 for (i = 0; i < MAX_BEARERS; i++) {
938 if (node->links[i]) {
939 link_print(node->links[i], "Resetting link\n");
940 tipc_link_reset(node->links[i]);
941 }
942 }
943
944 tipc_node_unlock(node);
945 }
946
947 static void link_retransmit_failure(struct tipc_link *l_ptr,
948 struct sk_buff *buf)
949 {
950 struct tipc_msg *msg = buf_msg(buf);
951
952 pr_warn("Retransmission failure on link <%s>\n", l_ptr->name);
953
954 if (l_ptr->addr) {
955 /* Handle failure on standard link */
956 link_print(l_ptr, "Resetting link\n");
957 tipc_link_reset(l_ptr);
958
959 } else {
960 /* Handle failure on broadcast link */
961 struct tipc_node *n_ptr;
962 char addr_string[16];
963
964 pr_info("Msg seq number: %u, ", msg_seqno(msg));
965 pr_cont("Outstanding acks: %lu\n",
966 (unsigned long) TIPC_SKB_CB(buf)->handle);
967
968 n_ptr = tipc_bclink_retransmit_to();
969 tipc_node_lock(n_ptr);
970
971 tipc_addr_string_fill(addr_string, n_ptr->addr);
972 pr_info("Broadcast link info for %s\n", addr_string);
973 pr_info("Reception permitted: %d, Acked: %u\n",
974 n_ptr->bclink.recv_permitted,
975 n_ptr->bclink.acked);
976 pr_info("Last in: %u, Oos state: %u, Last sent: %u\n",
977 n_ptr->bclink.last_in,
978 n_ptr->bclink.oos_state,
979 n_ptr->bclink.last_sent);
980
981 tipc_node_unlock(n_ptr);
982
983 tipc_bclink_set_flags(TIPC_BCLINK_RESET);
984 l_ptr->stale_count = 0;
985 }
986 }
987
988 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *buf,
989 u32 retransmits)
990 {
991 struct tipc_msg *msg;
992
993 if (!buf)
994 return;
995
996 msg = buf_msg(buf);
997
998 /* Detect repeated retransmit failures */
999 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1000 if (++l_ptr->stale_count > 100) {
1001 link_retransmit_failure(l_ptr, buf);
1002 return;
1003 }
1004 } else {
1005 l_ptr->last_retransmitted = msg_seqno(msg);
1006 l_ptr->stale_count = 1;
1007 }
1008
1009 while (retransmits && (buf != l_ptr->next_out) && buf) {
1010 msg = buf_msg(buf);
1011 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1012 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1013 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
1014 buf = buf->next;
1015 retransmits--;
1016 l_ptr->stats.retransmitted++;
1017 }
1018
1019 l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1020 }
1021
1022 /**
1023 * link_insert_deferred_queue - insert deferred messages back into receive chain
1024 */
1025 static struct sk_buff *link_insert_deferred_queue(struct tipc_link *l_ptr,
1026 struct sk_buff *buf)
1027 {
1028 u32 seq_no;
1029
1030 if (l_ptr->oldest_deferred_in == NULL)
1031 return buf;
1032
1033 seq_no = buf_seqno(l_ptr->oldest_deferred_in);
1034 if (seq_no == mod(l_ptr->next_in_no)) {
1035 l_ptr->newest_deferred_in->next = buf;
1036 buf = l_ptr->oldest_deferred_in;
1037 l_ptr->oldest_deferred_in = NULL;
1038 l_ptr->deferred_inqueue_sz = 0;
1039 }
1040 return buf;
1041 }
1042
1043 /**
1044 * link_recv_buf_validate - validate basic format of received message
1045 *
1046 * This routine ensures a TIPC message has an acceptable header, and at least
1047 * as much data as the header indicates it should. The routine also ensures
1048 * that the entire message header is stored in the main fragment of the message
1049 * buffer, to simplify future access to message header fields.
1050 *
1051 * Note: Having extra info present in the message header or data areas is OK.
1052 * TIPC will ignore the excess, under the assumption that it is optional info
1053 * introduced by a later release of the protocol.
1054 */
1055 static int link_recv_buf_validate(struct sk_buff *buf)
1056 {
1057 static u32 min_data_hdr_size[8] = {
1058 SHORT_H_SIZE, MCAST_H_SIZE, NAMED_H_SIZE, BASIC_H_SIZE,
1059 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1060 };
1061
1062 struct tipc_msg *msg;
1063 u32 tipc_hdr[2];
1064 u32 size;
1065 u32 hdr_size;
1066 u32 min_hdr_size;
1067
1068 /* If this packet comes from the defer queue, the skb has already
1069 * been validated
1070 */
1071 if (unlikely(TIPC_SKB_CB(buf)->deferred))
1072 return 1;
1073
1074 if (unlikely(buf->len < MIN_H_SIZE))
1075 return 0;
1076
1077 msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1078 if (msg == NULL)
1079 return 0;
1080
1081 if (unlikely(msg_version(msg) != TIPC_VERSION))
1082 return 0;
1083
1084 size = msg_size(msg);
1085 hdr_size = msg_hdr_sz(msg);
1086 min_hdr_size = msg_isdata(msg) ?
1087 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1088
1089 if (unlikely((hdr_size < min_hdr_size) ||
1090 (size < hdr_size) ||
1091 (buf->len < size) ||
1092 (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1093 return 0;
1094
1095 return pskb_may_pull(buf, hdr_size);
1096 }
1097
1098 /**
1099 * tipc_rcv - process TIPC packets/messages arriving from off-node
1100 * @head: pointer to message buffer chain
1101 * @b_ptr: pointer to bearer message arrived on
1102 *
1103 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1104 * structure (i.e. cannot be NULL), but bearer can be inactive.
1105 */
1106 void tipc_rcv(struct sk_buff *head, struct tipc_bearer *b_ptr)
1107 {
1108 while (head) {
1109 struct tipc_node *n_ptr;
1110 struct tipc_link *l_ptr;
1111 struct sk_buff *crs;
1112 struct sk_buff *buf = head;
1113 struct tipc_msg *msg;
1114 u32 seq_no;
1115 u32 ackd;
1116 u32 released = 0;
1117
1118 head = head->next;
1119 buf->next = NULL;
1120
1121 /* Ensure message is well-formed */
1122 if (unlikely(!link_recv_buf_validate(buf)))
1123 goto discard;
1124
1125 /* Ensure message data is a single contiguous unit */
1126 if (unlikely(skb_linearize(buf)))
1127 goto discard;
1128
1129 /* Handle arrival of a non-unicast link message */
1130 msg = buf_msg(buf);
1131
1132 if (unlikely(msg_non_seq(msg))) {
1133 if (msg_user(msg) == LINK_CONFIG)
1134 tipc_disc_rcv(buf, b_ptr);
1135 else
1136 tipc_bclink_rcv(buf);
1137 continue;
1138 }
1139
1140 /* Discard unicast link messages destined for another node */
1141 if (unlikely(!msg_short(msg) &&
1142 (msg_destnode(msg) != tipc_own_addr)))
1143 goto discard;
1144
1145 /* Locate neighboring node that sent message */
1146 n_ptr = tipc_node_find(msg_prevnode(msg));
1147 if (unlikely(!n_ptr))
1148 goto discard;
1149 tipc_node_lock(n_ptr);
1150
1151 /* Locate unicast link endpoint that should handle message */
1152 l_ptr = n_ptr->links[b_ptr->identity];
1153 if (unlikely(!l_ptr))
1154 goto unlock_discard;
1155
1156 /* Verify that communication with node is currently allowed */
1157 if ((n_ptr->action_flags & TIPC_WAIT_PEER_LINKS_DOWN) &&
1158 msg_user(msg) == LINK_PROTOCOL &&
1159 (msg_type(msg) == RESET_MSG ||
1160 msg_type(msg) == ACTIVATE_MSG) &&
1161 !msg_redundant_link(msg))
1162 n_ptr->action_flags &= ~TIPC_WAIT_PEER_LINKS_DOWN;
1163
1164 if (tipc_node_blocked(n_ptr))
1165 goto unlock_discard;
1166
1167 /* Validate message sequence number info */
1168 seq_no = msg_seqno(msg);
1169 ackd = msg_ack(msg);
1170
1171 /* Release acked messages */
1172 if (n_ptr->bclink.recv_permitted)
1173 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1174
1175 crs = l_ptr->first_out;
1176 while ((crs != l_ptr->next_out) &&
1177 less_eq(buf_seqno(crs), ackd)) {
1178 struct sk_buff *next = crs->next;
1179 kfree_skb(crs);
1180 crs = next;
1181 released++;
1182 }
1183 if (released) {
1184 l_ptr->first_out = crs;
1185 l_ptr->out_queue_size -= released;
1186 }
1187
1188 /* Try sending any messages link endpoint has pending */
1189 if (unlikely(l_ptr->next_out))
1190 tipc_link_push_queue(l_ptr);
1191
1192 if (released && !skb_queue_empty(&l_ptr->waiting_sks)) {
1193 link_prepare_wakeup(l_ptr);
1194 l_ptr->owner->action_flags |= TIPC_WAKEUP_USERS;
1195 }
1196
1197 /* Process the incoming packet */
1198 if (unlikely(!link_working_working(l_ptr))) {
1199 if (msg_user(msg) == LINK_PROTOCOL) {
1200 tipc_link_proto_rcv(l_ptr, buf);
1201 head = link_insert_deferred_queue(l_ptr, head);
1202 tipc_node_unlock(n_ptr);
1203 continue;
1204 }
1205
1206 /* Traffic message. Conditionally activate link */
1207 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1208
1209 if (link_working_working(l_ptr)) {
1210 /* Re-insert buffer in front of queue */
1211 buf->next = head;
1212 head = buf;
1213 tipc_node_unlock(n_ptr);
1214 continue;
1215 }
1216 goto unlock_discard;
1217 }
1218
1219 /* Link is now in state WORKING_WORKING */
1220 if (unlikely(seq_no != mod(l_ptr->next_in_no))) {
1221 link_handle_out_of_seq_msg(l_ptr, buf);
1222 head = link_insert_deferred_queue(l_ptr, head);
1223 tipc_node_unlock(n_ptr);
1224 continue;
1225 }
1226 l_ptr->next_in_no++;
1227 if (unlikely(l_ptr->oldest_deferred_in))
1228 head = link_insert_deferred_queue(l_ptr, head);
1229
1230 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1231 l_ptr->stats.sent_acks++;
1232 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1233 }
1234
1235 if (tipc_link_prepare_input(l_ptr, &buf)) {
1236 tipc_node_unlock(n_ptr);
1237 continue;
1238 }
1239 tipc_node_unlock(n_ptr);
1240 msg = buf_msg(buf);
1241 if (tipc_link_input(l_ptr, buf) != 0)
1242 goto discard;
1243 continue;
1244 unlock_discard:
1245 tipc_node_unlock(n_ptr);
1246 discard:
1247 kfree_skb(buf);
1248 }
1249 }
1250
1251 /**
1252 * tipc_link_prepare_input - process TIPC link messages
1253 *
1254 * returns nonzero if the message was consumed
1255 *
1256 * Node lock must be held
1257 */
1258 static int tipc_link_prepare_input(struct tipc_link *l, struct sk_buff **buf)
1259 {
1260 struct tipc_node *n;
1261 struct tipc_msg *msg;
1262 int res = -EINVAL;
1263
1264 n = l->owner;
1265 msg = buf_msg(*buf);
1266 switch (msg_user(msg)) {
1267 case CHANGEOVER_PROTOCOL:
1268 if (tipc_link_tunnel_rcv(n, buf))
1269 res = 0;
1270 break;
1271 case MSG_FRAGMENTER:
1272 l->stats.recv_fragments++;
1273 if (tipc_buf_append(&l->reasm_buf, buf)) {
1274 l->stats.recv_fragmented++;
1275 res = 0;
1276 } else if (!l->reasm_buf) {
1277 tipc_link_reset(l);
1278 }
1279 break;
1280 case MSG_BUNDLER:
1281 l->stats.recv_bundles++;
1282 l->stats.recv_bundled += msg_msgcnt(msg);
1283 res = 0;
1284 break;
1285 case NAME_DISTRIBUTOR:
1286 n->bclink.recv_permitted = true;
1287 res = 0;
1288 break;
1289 case BCAST_PROTOCOL:
1290 tipc_link_sync_rcv(n, *buf);
1291 break;
1292 default:
1293 res = 0;
1294 }
1295 return res;
1296 }
1297 /**
1298 * tipc_link_input - Deliver message too higher layers
1299 */
1300 static int tipc_link_input(struct tipc_link *l, struct sk_buff *buf)
1301 {
1302 struct tipc_msg *msg = buf_msg(buf);
1303 int res = 0;
1304
1305 switch (msg_user(msg)) {
1306 case TIPC_LOW_IMPORTANCE:
1307 case TIPC_MEDIUM_IMPORTANCE:
1308 case TIPC_HIGH_IMPORTANCE:
1309 case TIPC_CRITICAL_IMPORTANCE:
1310 case CONN_MANAGER:
1311 tipc_sk_rcv(buf);
1312 break;
1313 case NAME_DISTRIBUTOR:
1314 tipc_named_rcv(buf);
1315 break;
1316 case MSG_BUNDLER:
1317 tipc_link_bundle_rcv(buf);
1318 break;
1319 default:
1320 res = -EINVAL;
1321 }
1322 return res;
1323 }
1324
1325 /**
1326 * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1327 *
1328 * Returns increase in queue length (i.e. 0 or 1)
1329 */
1330 u32 tipc_link_defer_pkt(struct sk_buff **head, struct sk_buff **tail,
1331 struct sk_buff *buf)
1332 {
1333 struct sk_buff *queue_buf;
1334 struct sk_buff **prev;
1335 u32 seq_no = buf_seqno(buf);
1336
1337 buf->next = NULL;
1338
1339 /* Empty queue ? */
1340 if (*head == NULL) {
1341 *head = *tail = buf;
1342 return 1;
1343 }
1344
1345 /* Last ? */
1346 if (less(buf_seqno(*tail), seq_no)) {
1347 (*tail)->next = buf;
1348 *tail = buf;
1349 return 1;
1350 }
1351
1352 /* Locate insertion point in queue, then insert; discard if duplicate */
1353 prev = head;
1354 queue_buf = *head;
1355 for (;;) {
1356 u32 curr_seqno = buf_seqno(queue_buf);
1357
1358 if (seq_no == curr_seqno) {
1359 kfree_skb(buf);
1360 return 0;
1361 }
1362
1363 if (less(seq_no, curr_seqno))
1364 break;
1365
1366 prev = &queue_buf->next;
1367 queue_buf = queue_buf->next;
1368 }
1369
1370 buf->next = queue_buf;
1371 *prev = buf;
1372 return 1;
1373 }
1374
1375 /*
1376 * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
1377 */
1378 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
1379 struct sk_buff *buf)
1380 {
1381 u32 seq_no = buf_seqno(buf);
1382
1383 if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
1384 tipc_link_proto_rcv(l_ptr, buf);
1385 return;
1386 }
1387
1388 /* Record OOS packet arrival (force mismatch on next timeout) */
1389 l_ptr->checkpoint--;
1390
1391 /*
1392 * Discard packet if a duplicate; otherwise add it to deferred queue
1393 * and notify peer of gap as per protocol specification
1394 */
1395 if (less(seq_no, mod(l_ptr->next_in_no))) {
1396 l_ptr->stats.duplicates++;
1397 kfree_skb(buf);
1398 return;
1399 }
1400
1401 if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
1402 &l_ptr->newest_deferred_in, buf)) {
1403 l_ptr->deferred_inqueue_sz++;
1404 l_ptr->stats.deferred_recv++;
1405 TIPC_SKB_CB(buf)->deferred = true;
1406 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
1407 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1408 } else
1409 l_ptr->stats.duplicates++;
1410 }
1411
1412 /*
1413 * Send protocol message to the other endpoint.
1414 */
1415 void tipc_link_proto_xmit(struct tipc_link *l_ptr, u32 msg_typ, int probe_msg,
1416 u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
1417 {
1418 struct sk_buff *buf = NULL;
1419 struct tipc_msg *msg = l_ptr->pmsg;
1420 u32 msg_size = sizeof(l_ptr->proto_msg);
1421 int r_flag;
1422
1423 /* Discard any previous message that was deferred due to congestion */
1424 if (l_ptr->proto_msg_queue) {
1425 kfree_skb(l_ptr->proto_msg_queue);
1426 l_ptr->proto_msg_queue = NULL;
1427 }
1428
1429 /* Don't send protocol message during link changeover */
1430 if (l_ptr->exp_msg_count)
1431 return;
1432
1433 /* Abort non-RESET send if communication with node is prohibited */
1434 if ((tipc_node_blocked(l_ptr->owner)) && (msg_typ != RESET_MSG))
1435 return;
1436
1437 /* Create protocol message with "out-of-sequence" sequence number */
1438 msg_set_type(msg, msg_typ);
1439 msg_set_net_plane(msg, l_ptr->net_plane);
1440 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1441 msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
1442
1443 if (msg_typ == STATE_MSG) {
1444 u32 next_sent = mod(l_ptr->next_out_no);
1445
1446 if (!tipc_link_is_up(l_ptr))
1447 return;
1448 if (l_ptr->next_out)
1449 next_sent = buf_seqno(l_ptr->next_out);
1450 msg_set_next_sent(msg, next_sent);
1451 if (l_ptr->oldest_deferred_in) {
1452 u32 rec = buf_seqno(l_ptr->oldest_deferred_in);
1453 gap = mod(rec - mod(l_ptr->next_in_no));
1454 }
1455 msg_set_seq_gap(msg, gap);
1456 if (gap)
1457 l_ptr->stats.sent_nacks++;
1458 msg_set_link_tolerance(msg, tolerance);
1459 msg_set_linkprio(msg, priority);
1460 msg_set_max_pkt(msg, ack_mtu);
1461 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1462 msg_set_probe(msg, probe_msg != 0);
1463 if (probe_msg) {
1464 u32 mtu = l_ptr->max_pkt;
1465
1466 if ((mtu < l_ptr->max_pkt_target) &&
1467 link_working_working(l_ptr) &&
1468 l_ptr->fsm_msg_cnt) {
1469 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1470 if (l_ptr->max_pkt_probes == 10) {
1471 l_ptr->max_pkt_target = (msg_size - 4);
1472 l_ptr->max_pkt_probes = 0;
1473 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1474 }
1475 l_ptr->max_pkt_probes++;
1476 }
1477
1478 l_ptr->stats.sent_probes++;
1479 }
1480 l_ptr->stats.sent_states++;
1481 } else { /* RESET_MSG or ACTIVATE_MSG */
1482 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
1483 msg_set_seq_gap(msg, 0);
1484 msg_set_next_sent(msg, 1);
1485 msg_set_probe(msg, 0);
1486 msg_set_link_tolerance(msg, l_ptr->tolerance);
1487 msg_set_linkprio(msg, l_ptr->priority);
1488 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
1489 }
1490
1491 r_flag = (l_ptr->owner->working_links > tipc_link_is_up(l_ptr));
1492 msg_set_redundant_link(msg, r_flag);
1493 msg_set_linkprio(msg, l_ptr->priority);
1494 msg_set_size(msg, msg_size);
1495
1496 msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
1497
1498 buf = tipc_buf_acquire(msg_size);
1499 if (!buf)
1500 return;
1501
1502 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1503 buf->priority = TC_PRIO_CONTROL;
1504
1505 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
1506 l_ptr->unacked_window = 0;
1507 kfree_skb(buf);
1508 }
1509
1510 /*
1511 * Receive protocol message :
1512 * Note that network plane id propagates through the network, and may
1513 * change at any time. The node with lowest address rules
1514 */
1515 static void tipc_link_proto_rcv(struct tipc_link *l_ptr, struct sk_buff *buf)
1516 {
1517 u32 rec_gap = 0;
1518 u32 max_pkt_info;
1519 u32 max_pkt_ack;
1520 u32 msg_tol;
1521 struct tipc_msg *msg = buf_msg(buf);
1522
1523 /* Discard protocol message during link changeover */
1524 if (l_ptr->exp_msg_count)
1525 goto exit;
1526
1527 if (l_ptr->net_plane != msg_net_plane(msg))
1528 if (tipc_own_addr > msg_prevnode(msg))
1529 l_ptr->net_plane = msg_net_plane(msg);
1530
1531 switch (msg_type(msg)) {
1532
1533 case RESET_MSG:
1534 if (!link_working_unknown(l_ptr) &&
1535 (l_ptr->peer_session != INVALID_SESSION)) {
1536 if (less_eq(msg_session(msg), l_ptr->peer_session))
1537 break; /* duplicate or old reset: ignore */
1538 }
1539
1540 if (!msg_redundant_link(msg) && (link_working_working(l_ptr) ||
1541 link_working_unknown(l_ptr))) {
1542 /*
1543 * peer has lost contact -- don't allow peer's links
1544 * to reactivate before we recognize loss & clean up
1545 */
1546 l_ptr->owner->action_flags |= TIPC_WAIT_OWN_LINKS_DOWN;
1547 }
1548
1549 link_state_event(l_ptr, RESET_MSG);
1550
1551 /* fall thru' */
1552 case ACTIVATE_MSG:
1553 /* Update link settings according other endpoint's values */
1554 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
1555
1556 msg_tol = msg_link_tolerance(msg);
1557 if (msg_tol > l_ptr->tolerance)
1558 link_set_supervision_props(l_ptr, msg_tol);
1559
1560 if (msg_linkprio(msg) > l_ptr->priority)
1561 l_ptr->priority = msg_linkprio(msg);
1562
1563 max_pkt_info = msg_max_pkt(msg);
1564 if (max_pkt_info) {
1565 if (max_pkt_info < l_ptr->max_pkt_target)
1566 l_ptr->max_pkt_target = max_pkt_info;
1567 if (l_ptr->max_pkt > l_ptr->max_pkt_target)
1568 l_ptr->max_pkt = l_ptr->max_pkt_target;
1569 } else {
1570 l_ptr->max_pkt = l_ptr->max_pkt_target;
1571 }
1572
1573 /* Synchronize broadcast link info, if not done previously */
1574 if (!tipc_node_is_up(l_ptr->owner)) {
1575 l_ptr->owner->bclink.last_sent =
1576 l_ptr->owner->bclink.last_in =
1577 msg_last_bcast(msg);
1578 l_ptr->owner->bclink.oos_state = 0;
1579 }
1580
1581 l_ptr->peer_session = msg_session(msg);
1582 l_ptr->peer_bearer_id = msg_bearer_id(msg);
1583
1584 if (msg_type(msg) == ACTIVATE_MSG)
1585 link_state_event(l_ptr, ACTIVATE_MSG);
1586 break;
1587 case STATE_MSG:
1588
1589 msg_tol = msg_link_tolerance(msg);
1590 if (msg_tol)
1591 link_set_supervision_props(l_ptr, msg_tol);
1592
1593 if (msg_linkprio(msg) &&
1594 (msg_linkprio(msg) != l_ptr->priority)) {
1595 pr_warn("%s<%s>, priority change %u->%u\n",
1596 link_rst_msg, l_ptr->name, l_ptr->priority,
1597 msg_linkprio(msg));
1598 l_ptr->priority = msg_linkprio(msg);
1599 tipc_link_reset(l_ptr); /* Enforce change to take effect */
1600 break;
1601 }
1602
1603 /* Record reception; force mismatch at next timeout: */
1604 l_ptr->checkpoint--;
1605
1606 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1607 l_ptr->stats.recv_states++;
1608 if (link_reset_unknown(l_ptr))
1609 break;
1610
1611 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
1612 rec_gap = mod(msg_next_sent(msg) -
1613 mod(l_ptr->next_in_no));
1614 }
1615
1616 max_pkt_ack = msg_max_pkt(msg);
1617 if (max_pkt_ack > l_ptr->max_pkt) {
1618 l_ptr->max_pkt = max_pkt_ack;
1619 l_ptr->max_pkt_probes = 0;
1620 }
1621
1622 max_pkt_ack = 0;
1623 if (msg_probe(msg)) {
1624 l_ptr->stats.recv_probes++;
1625 if (msg_size(msg) > sizeof(l_ptr->proto_msg))
1626 max_pkt_ack = msg_size(msg);
1627 }
1628
1629 /* Protocol message before retransmits, reduce loss risk */
1630 if (l_ptr->owner->bclink.recv_permitted)
1631 tipc_bclink_update_link_state(l_ptr->owner,
1632 msg_last_bcast(msg));
1633
1634 if (rec_gap || (msg_probe(msg))) {
1635 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, rec_gap, 0,
1636 0, max_pkt_ack);
1637 }
1638 if (msg_seq_gap(msg)) {
1639 l_ptr->stats.recv_nacks++;
1640 tipc_link_retransmit(l_ptr, l_ptr->first_out,
1641 msg_seq_gap(msg));
1642 }
1643 break;
1644 }
1645 exit:
1646 kfree_skb(buf);
1647 }
1648
1649
1650 /* tipc_link_tunnel_xmit(): Tunnel one packet via a link belonging to
1651 * a different bearer. Owner node is locked.
1652 */
1653 static void tipc_link_tunnel_xmit(struct tipc_link *l_ptr,
1654 struct tipc_msg *tunnel_hdr,
1655 struct tipc_msg *msg,
1656 u32 selector)
1657 {
1658 struct tipc_link *tunnel;
1659 struct sk_buff *buf;
1660 u32 length = msg_size(msg);
1661
1662 tunnel = l_ptr->owner->active_links[selector & 1];
1663 if (!tipc_link_is_up(tunnel)) {
1664 pr_warn("%stunnel link no longer available\n", link_co_err);
1665 return;
1666 }
1667 msg_set_size(tunnel_hdr, length + INT_H_SIZE);
1668 buf = tipc_buf_acquire(length + INT_H_SIZE);
1669 if (!buf) {
1670 pr_warn("%sunable to send tunnel msg\n", link_co_err);
1671 return;
1672 }
1673 skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
1674 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
1675 __tipc_link_xmit(tunnel, buf);
1676 }
1677
1678
1679 /* tipc_link_failover_send_queue(): A link has gone down, but a second
1680 * link is still active. We can do failover. Tunnel the failing link's
1681 * whole send queue via the remaining link. This way, we don't lose
1682 * any packets, and sequence order is preserved for subsequent traffic
1683 * sent over the remaining link. Owner node is locked.
1684 */
1685 void tipc_link_failover_send_queue(struct tipc_link *l_ptr)
1686 {
1687 u32 msgcount = l_ptr->out_queue_size;
1688 struct sk_buff *crs = l_ptr->first_out;
1689 struct tipc_link *tunnel = l_ptr->owner->active_links[0];
1690 struct tipc_msg tunnel_hdr;
1691 int split_bundles;
1692
1693 if (!tunnel)
1694 return;
1695
1696 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
1697 ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
1698 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
1699 msg_set_msgcnt(&tunnel_hdr, msgcount);
1700
1701 if (!l_ptr->first_out) {
1702 struct sk_buff *buf;
1703
1704 buf = tipc_buf_acquire(INT_H_SIZE);
1705 if (buf) {
1706 skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
1707 msg_set_size(&tunnel_hdr, INT_H_SIZE);
1708 __tipc_link_xmit(tunnel, buf);
1709 } else {
1710 pr_warn("%sunable to send changeover msg\n",
1711 link_co_err);
1712 }
1713 return;
1714 }
1715
1716 split_bundles = (l_ptr->owner->active_links[0] !=
1717 l_ptr->owner->active_links[1]);
1718
1719 while (crs) {
1720 struct tipc_msg *msg = buf_msg(crs);
1721
1722 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
1723 struct tipc_msg *m = msg_get_wrapped(msg);
1724 unchar *pos = (unchar *)m;
1725
1726 msgcount = msg_msgcnt(msg);
1727 while (msgcount--) {
1728 msg_set_seqno(m, msg_seqno(msg));
1729 tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, m,
1730 msg_link_selector(m));
1731 pos += align(msg_size(m));
1732 m = (struct tipc_msg *)pos;
1733 }
1734 } else {
1735 tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, msg,
1736 msg_link_selector(msg));
1737 }
1738 crs = crs->next;
1739 }
1740 }
1741
1742 /* tipc_link_dup_queue_xmit(): A second link has become active. Tunnel a
1743 * duplicate of the first link's send queue via the new link. This way, we
1744 * are guaranteed that currently queued packets from a socket are delivered
1745 * before future traffic from the same socket, even if this is using the
1746 * new link. The last arriving copy of each duplicate packet is dropped at
1747 * the receiving end by the regular protocol check, so packet cardinality
1748 * and sequence order is preserved per sender/receiver socket pair.
1749 * Owner node is locked.
1750 */
1751 void tipc_link_dup_queue_xmit(struct tipc_link *l_ptr,
1752 struct tipc_link *tunnel)
1753 {
1754 struct sk_buff *iter;
1755 struct tipc_msg tunnel_hdr;
1756
1757 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
1758 DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
1759 msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
1760 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
1761 iter = l_ptr->first_out;
1762 while (iter) {
1763 struct sk_buff *outbuf;
1764 struct tipc_msg *msg = buf_msg(iter);
1765 u32 length = msg_size(msg);
1766
1767 if (msg_user(msg) == MSG_BUNDLER)
1768 msg_set_type(msg, CLOSED_MSG);
1769 msg_set_ack(msg, mod(l_ptr->next_in_no - 1)); /* Update */
1770 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1771 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
1772 outbuf = tipc_buf_acquire(length + INT_H_SIZE);
1773 if (outbuf == NULL) {
1774 pr_warn("%sunable to send duplicate msg\n",
1775 link_co_err);
1776 return;
1777 }
1778 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
1779 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
1780 length);
1781 __tipc_link_xmit(tunnel, outbuf);
1782 if (!tipc_link_is_up(l_ptr))
1783 return;
1784 iter = iter->next;
1785 }
1786 }
1787
1788 /**
1789 * buf_extract - extracts embedded TIPC message from another message
1790 * @skb: encapsulating message buffer
1791 * @from_pos: offset to extract from
1792 *
1793 * Returns a new message buffer containing an embedded message. The
1794 * encapsulating message itself is left unchanged.
1795 */
1796 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
1797 {
1798 struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
1799 u32 size = msg_size(msg);
1800 struct sk_buff *eb;
1801
1802 eb = tipc_buf_acquire(size);
1803 if (eb)
1804 skb_copy_to_linear_data(eb, msg, size);
1805 return eb;
1806 }
1807
1808
1809
1810 /* tipc_link_dup_rcv(): Receive a tunnelled DUPLICATE_MSG packet.
1811 * Owner node is locked.
1812 */
1813 static void tipc_link_dup_rcv(struct tipc_link *l_ptr,
1814 struct sk_buff *t_buf)
1815 {
1816 struct sk_buff *buf;
1817
1818 if (!tipc_link_is_up(l_ptr))
1819 return;
1820
1821 buf = buf_extract(t_buf, INT_H_SIZE);
1822 if (buf == NULL) {
1823 pr_warn("%sfailed to extract inner dup pkt\n", link_co_err);
1824 return;
1825 }
1826
1827 /* Add buffer to deferred queue, if applicable: */
1828 link_handle_out_of_seq_msg(l_ptr, buf);
1829 }
1830
1831 /* tipc_link_failover_rcv(): Receive a tunnelled ORIGINAL_MSG packet
1832 * Owner node is locked.
1833 */
1834 static struct sk_buff *tipc_link_failover_rcv(struct tipc_link *l_ptr,
1835 struct sk_buff *t_buf)
1836 {
1837 struct tipc_msg *t_msg = buf_msg(t_buf);
1838 struct sk_buff *buf = NULL;
1839 struct tipc_msg *msg;
1840
1841 if (tipc_link_is_up(l_ptr))
1842 tipc_link_reset(l_ptr);
1843
1844 /* First failover packet? */
1845 if (l_ptr->exp_msg_count == START_CHANGEOVER)
1846 l_ptr->exp_msg_count = msg_msgcnt(t_msg);
1847
1848 /* Should there be an inner packet? */
1849 if (l_ptr->exp_msg_count) {
1850 l_ptr->exp_msg_count--;
1851 buf = buf_extract(t_buf, INT_H_SIZE);
1852 if (buf == NULL) {
1853 pr_warn("%sno inner failover pkt\n", link_co_err);
1854 goto exit;
1855 }
1856 msg = buf_msg(buf);
1857
1858 if (less(msg_seqno(msg), l_ptr->reset_checkpoint)) {
1859 kfree_skb(buf);
1860 buf = NULL;
1861 goto exit;
1862 }
1863 if (msg_user(msg) == MSG_FRAGMENTER) {
1864 l_ptr->stats.recv_fragments++;
1865 tipc_buf_append(&l_ptr->reasm_buf, &buf);
1866 }
1867 }
1868 exit:
1869 if ((l_ptr->exp_msg_count == 0) && (l_ptr->flags & LINK_STOPPED)) {
1870 tipc_node_detach_link(l_ptr->owner, l_ptr);
1871 kfree(l_ptr);
1872 }
1873 return buf;
1874 }
1875
1876 /* tipc_link_tunnel_rcv(): Receive a tunnelled packet, sent
1877 * via other link as result of a failover (ORIGINAL_MSG) or
1878 * a new active link (DUPLICATE_MSG). Failover packets are
1879 * returned to the active link for delivery upwards.
1880 * Owner node is locked.
1881 */
1882 static int tipc_link_tunnel_rcv(struct tipc_node *n_ptr,
1883 struct sk_buff **buf)
1884 {
1885 struct sk_buff *t_buf = *buf;
1886 struct tipc_link *l_ptr;
1887 struct tipc_msg *t_msg = buf_msg(t_buf);
1888 u32 bearer_id = msg_bearer_id(t_msg);
1889
1890 *buf = NULL;
1891
1892 if (bearer_id >= MAX_BEARERS)
1893 goto exit;
1894
1895 l_ptr = n_ptr->links[bearer_id];
1896 if (!l_ptr)
1897 goto exit;
1898
1899 if (msg_type(t_msg) == DUPLICATE_MSG)
1900 tipc_link_dup_rcv(l_ptr, t_buf);
1901 else if (msg_type(t_msg) == ORIGINAL_MSG)
1902 *buf = tipc_link_failover_rcv(l_ptr, t_buf);
1903 else
1904 pr_warn("%sunknown tunnel pkt received\n", link_co_err);
1905 exit:
1906 kfree_skb(t_buf);
1907 return *buf != NULL;
1908 }
1909
1910 /*
1911 * Bundler functionality:
1912 */
1913 void tipc_link_bundle_rcv(struct sk_buff *buf)
1914 {
1915 u32 msgcount = msg_msgcnt(buf_msg(buf));
1916 u32 pos = INT_H_SIZE;
1917 struct sk_buff *obuf;
1918 struct tipc_msg *omsg;
1919
1920 while (msgcount--) {
1921 obuf = buf_extract(buf, pos);
1922 if (obuf == NULL) {
1923 pr_warn("Link unable to unbundle message(s)\n");
1924 break;
1925 }
1926 omsg = buf_msg(obuf);
1927 pos += align(msg_size(omsg));
1928 if (msg_isdata(omsg) || (msg_user(omsg) == CONN_MANAGER)) {
1929 tipc_sk_rcv(obuf);
1930 } else if (msg_user(omsg) == NAME_DISTRIBUTOR) {
1931 tipc_named_rcv(obuf);
1932 } else {
1933 pr_warn("Illegal bundled msg: %u\n", msg_user(omsg));
1934 kfree_skb(obuf);
1935 }
1936 }
1937 kfree_skb(buf);
1938 }
1939
1940 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance)
1941 {
1942 if ((tolerance < TIPC_MIN_LINK_TOL) || (tolerance > TIPC_MAX_LINK_TOL))
1943 return;
1944
1945 l_ptr->tolerance = tolerance;
1946 l_ptr->continuity_interval =
1947 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
1948 l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
1949 }
1950
1951 void tipc_link_set_queue_limits(struct tipc_link *l_ptr, u32 window)
1952 {
1953 /* Data messages from this node, inclusive FIRST_FRAGM */
1954 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
1955 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
1956 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
1957 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
1958 /* Transiting data messages,inclusive FIRST_FRAGM */
1959 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
1960 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
1961 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
1962 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
1963 l_ptr->queue_limit[CONN_MANAGER] = 1200;
1964 l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
1965 l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
1966 /* FRAGMENT and LAST_FRAGMENT packets */
1967 l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
1968 }
1969
1970 /* tipc_link_find_owner - locate owner node of link by link's name
1971 * @name: pointer to link name string
1972 * @bearer_id: pointer to index in 'node->links' array where the link was found.
1973 *
1974 * Returns pointer to node owning the link, or 0 if no matching link is found.
1975 */
1976 static struct tipc_node *tipc_link_find_owner(const char *link_name,
1977 unsigned int *bearer_id)
1978 {
1979 struct tipc_link *l_ptr;
1980 struct tipc_node *n_ptr;
1981 struct tipc_node *found_node = 0;
1982 int i;
1983
1984 *bearer_id = 0;
1985 rcu_read_lock();
1986 list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
1987 tipc_node_lock(n_ptr);
1988 for (i = 0; i < MAX_BEARERS; i++) {
1989 l_ptr = n_ptr->links[i];
1990 if (l_ptr && !strcmp(l_ptr->name, link_name)) {
1991 *bearer_id = i;
1992 found_node = n_ptr;
1993 break;
1994 }
1995 }
1996 tipc_node_unlock(n_ptr);
1997 if (found_node)
1998 break;
1999 }
2000 rcu_read_unlock();
2001
2002 return found_node;
2003 }
2004
2005 /**
2006 * link_value_is_valid -- validate proposed link tolerance/priority/window
2007 *
2008 * @cmd: value type (TIPC_CMD_SET_LINK_*)
2009 * @new_value: the new value
2010 *
2011 * Returns 1 if value is within range, 0 if not.
2012 */
2013 static int link_value_is_valid(u16 cmd, u32 new_value)
2014 {
2015 switch (cmd) {
2016 case TIPC_CMD_SET_LINK_TOL:
2017 return (new_value >= TIPC_MIN_LINK_TOL) &&
2018 (new_value <= TIPC_MAX_LINK_TOL);
2019 case TIPC_CMD_SET_LINK_PRI:
2020 return (new_value <= TIPC_MAX_LINK_PRI);
2021 case TIPC_CMD_SET_LINK_WINDOW:
2022 return (new_value >= TIPC_MIN_LINK_WIN) &&
2023 (new_value <= TIPC_MAX_LINK_WIN);
2024 }
2025 return 0;
2026 }
2027
2028 /**
2029 * link_cmd_set_value - change priority/tolerance/window for link/bearer/media
2030 * @name: ptr to link, bearer, or media name
2031 * @new_value: new value of link, bearer, or media setting
2032 * @cmd: which link, bearer, or media attribute to set (TIPC_CMD_SET_LINK_*)
2033 *
2034 * Caller must hold RTNL lock to ensure link/bearer/media is not deleted.
2035 *
2036 * Returns 0 if value updated and negative value on error.
2037 */
2038 static int link_cmd_set_value(const char *name, u32 new_value, u16 cmd)
2039 {
2040 struct tipc_node *node;
2041 struct tipc_link *l_ptr;
2042 struct tipc_bearer *b_ptr;
2043 struct tipc_media *m_ptr;
2044 int bearer_id;
2045 int res = 0;
2046
2047 node = tipc_link_find_owner(name, &bearer_id);
2048 if (node) {
2049 tipc_node_lock(node);
2050 l_ptr = node->links[bearer_id];
2051
2052 if (l_ptr) {
2053 switch (cmd) {
2054 case TIPC_CMD_SET_LINK_TOL:
2055 link_set_supervision_props(l_ptr, new_value);
2056 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0,
2057 new_value, 0, 0);
2058 break;
2059 case TIPC_CMD_SET_LINK_PRI:
2060 l_ptr->priority = new_value;
2061 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0,
2062 0, new_value, 0);
2063 break;
2064 case TIPC_CMD_SET_LINK_WINDOW:
2065 tipc_link_set_queue_limits(l_ptr, new_value);
2066 break;
2067 default:
2068 res = -EINVAL;
2069 break;
2070 }
2071 }
2072 tipc_node_unlock(node);
2073 return res;
2074 }
2075
2076 b_ptr = tipc_bearer_find(name);
2077 if (b_ptr) {
2078 switch (cmd) {
2079 case TIPC_CMD_SET_LINK_TOL:
2080 b_ptr->tolerance = new_value;
2081 break;
2082 case TIPC_CMD_SET_LINK_PRI:
2083 b_ptr->priority = new_value;
2084 break;
2085 case TIPC_CMD_SET_LINK_WINDOW:
2086 b_ptr->window = new_value;
2087 break;
2088 default:
2089 res = -EINVAL;
2090 break;
2091 }
2092 return res;
2093 }
2094
2095 m_ptr = tipc_media_find(name);
2096 if (!m_ptr)
2097 return -ENODEV;
2098 switch (cmd) {
2099 case TIPC_CMD_SET_LINK_TOL:
2100 m_ptr->tolerance = new_value;
2101 break;
2102 case TIPC_CMD_SET_LINK_PRI:
2103 m_ptr->priority = new_value;
2104 break;
2105 case TIPC_CMD_SET_LINK_WINDOW:
2106 m_ptr->window = new_value;
2107 break;
2108 default:
2109 res = -EINVAL;
2110 break;
2111 }
2112 return res;
2113 }
2114
2115 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2116 u16 cmd)
2117 {
2118 struct tipc_link_config *args;
2119 u32 new_value;
2120 int res;
2121
2122 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2123 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2124
2125 args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2126 new_value = ntohl(args->value);
2127
2128 if (!link_value_is_valid(cmd, new_value))
2129 return tipc_cfg_reply_error_string(
2130 "cannot change, value invalid");
2131
2132 if (!strcmp(args->name, tipc_bclink_name)) {
2133 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2134 (tipc_bclink_set_queue_limits(new_value) == 0))
2135 return tipc_cfg_reply_none();
2136 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2137 " (cannot change setting on broadcast link)");
2138 }
2139
2140 res = link_cmd_set_value(args->name, new_value, cmd);
2141 if (res)
2142 return tipc_cfg_reply_error_string("cannot change link setting");
2143
2144 return tipc_cfg_reply_none();
2145 }
2146
2147 /**
2148 * link_reset_statistics - reset link statistics
2149 * @l_ptr: pointer to link
2150 */
2151 static void link_reset_statistics(struct tipc_link *l_ptr)
2152 {
2153 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2154 l_ptr->stats.sent_info = l_ptr->next_out_no;
2155 l_ptr->stats.recv_info = l_ptr->next_in_no;
2156 }
2157
2158 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2159 {
2160 char *link_name;
2161 struct tipc_link *l_ptr;
2162 struct tipc_node *node;
2163 unsigned int bearer_id;
2164
2165 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2166 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2167
2168 link_name = (char *)TLV_DATA(req_tlv_area);
2169 if (!strcmp(link_name, tipc_bclink_name)) {
2170 if (tipc_bclink_reset_stats())
2171 return tipc_cfg_reply_error_string("link not found");
2172 return tipc_cfg_reply_none();
2173 }
2174 node = tipc_link_find_owner(link_name, &bearer_id);
2175 if (!node)
2176 return tipc_cfg_reply_error_string("link not found");
2177
2178 tipc_node_lock(node);
2179 l_ptr = node->links[bearer_id];
2180 if (!l_ptr) {
2181 tipc_node_unlock(node);
2182 return tipc_cfg_reply_error_string("link not found");
2183 }
2184 link_reset_statistics(l_ptr);
2185 tipc_node_unlock(node);
2186 return tipc_cfg_reply_none();
2187 }
2188
2189 /**
2190 * percent - convert count to a percentage of total (rounding up or down)
2191 */
2192 static u32 percent(u32 count, u32 total)
2193 {
2194 return (count * 100 + (total / 2)) / total;
2195 }
2196
2197 /**
2198 * tipc_link_stats - print link statistics
2199 * @name: link name
2200 * @buf: print buffer area
2201 * @buf_size: size of print buffer area
2202 *
2203 * Returns length of print buffer data string (or 0 if error)
2204 */
2205 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
2206 {
2207 struct tipc_link *l;
2208 struct tipc_stats *s;
2209 struct tipc_node *node;
2210 char *status;
2211 u32 profile_total = 0;
2212 unsigned int bearer_id;
2213 int ret;
2214
2215 if (!strcmp(name, tipc_bclink_name))
2216 return tipc_bclink_stats(buf, buf_size);
2217
2218 node = tipc_link_find_owner(name, &bearer_id);
2219 if (!node)
2220 return 0;
2221
2222 tipc_node_lock(node);
2223
2224 l = node->links[bearer_id];
2225 if (!l) {
2226 tipc_node_unlock(node);
2227 return 0;
2228 }
2229
2230 s = &l->stats;
2231
2232 if (tipc_link_is_active(l))
2233 status = "ACTIVE";
2234 else if (tipc_link_is_up(l))
2235 status = "STANDBY";
2236 else
2237 status = "DEFUNCT";
2238
2239 ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
2240 " %s MTU:%u Priority:%u Tolerance:%u ms"
2241 " Window:%u packets\n",
2242 l->name, status, l->max_pkt, l->priority,
2243 l->tolerance, l->queue_limit[0]);
2244
2245 ret += tipc_snprintf(buf + ret, buf_size - ret,
2246 " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
2247 l->next_in_no - s->recv_info, s->recv_fragments,
2248 s->recv_fragmented, s->recv_bundles,
2249 s->recv_bundled);
2250
2251 ret += tipc_snprintf(buf + ret, buf_size - ret,
2252 " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
2253 l->next_out_no - s->sent_info, s->sent_fragments,
2254 s->sent_fragmented, s->sent_bundles,
2255 s->sent_bundled);
2256
2257 profile_total = s->msg_length_counts;
2258 if (!profile_total)
2259 profile_total = 1;
2260
2261 ret += tipc_snprintf(buf + ret, buf_size - ret,
2262 " TX profile sample:%u packets average:%u octets\n"
2263 " 0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
2264 "-16384:%u%% -32768:%u%% -66000:%u%%\n",
2265 s->msg_length_counts,
2266 s->msg_lengths_total / profile_total,
2267 percent(s->msg_length_profile[0], profile_total),
2268 percent(s->msg_length_profile[1], profile_total),
2269 percent(s->msg_length_profile[2], profile_total),
2270 percent(s->msg_length_profile[3], profile_total),
2271 percent(s->msg_length_profile[4], profile_total),
2272 percent(s->msg_length_profile[5], profile_total),
2273 percent(s->msg_length_profile[6], profile_total));
2274
2275 ret += tipc_snprintf(buf + ret, buf_size - ret,
2276 " RX states:%u probes:%u naks:%u defs:%u"
2277 " dups:%u\n", s->recv_states, s->recv_probes,
2278 s->recv_nacks, s->deferred_recv, s->duplicates);
2279
2280 ret += tipc_snprintf(buf + ret, buf_size - ret,
2281 " TX states:%u probes:%u naks:%u acks:%u"
2282 " dups:%u\n", s->sent_states, s->sent_probes,
2283 s->sent_nacks, s->sent_acks, s->retransmitted);
2284
2285 ret += tipc_snprintf(buf + ret, buf_size - ret,
2286 " Congestion link:%u Send queue"
2287 " max:%u avg:%u\n", s->link_congs,
2288 s->max_queue_sz, s->queue_sz_counts ?
2289 (s->accu_queue_sz / s->queue_sz_counts) : 0);
2290
2291 tipc_node_unlock(node);
2292 return ret;
2293 }
2294
2295 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
2296 {
2297 struct sk_buff *buf;
2298 struct tlv_desc *rep_tlv;
2299 int str_len;
2300 int pb_len;
2301 char *pb;
2302
2303 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2304 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2305
2306 buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2307 if (!buf)
2308 return NULL;
2309
2310 rep_tlv = (struct tlv_desc *)buf->data;
2311 pb = TLV_DATA(rep_tlv);
2312 pb_len = ULTRA_STRING_MAX_LEN;
2313 str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
2314 pb, pb_len);
2315 if (!str_len) {
2316 kfree_skb(buf);
2317 return tipc_cfg_reply_error_string("link not found");
2318 }
2319 str_len += 1; /* for "\0" */
2320 skb_put(buf, TLV_SPACE(str_len));
2321 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2322
2323 return buf;
2324 }
2325
2326 /**
2327 * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
2328 * @dest: network address of destination node
2329 * @selector: used to select from set of active links
2330 *
2331 * If no active link can be found, uses default maximum packet size.
2332 */
2333 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
2334 {
2335 struct tipc_node *n_ptr;
2336 struct tipc_link *l_ptr;
2337 u32 res = MAX_PKT_DEFAULT;
2338
2339 if (dest == tipc_own_addr)
2340 return MAX_MSG_SIZE;
2341
2342 n_ptr = tipc_node_find(dest);
2343 if (n_ptr) {
2344 tipc_node_lock(n_ptr);
2345 l_ptr = n_ptr->active_links[selector & 1];
2346 if (l_ptr)
2347 res = l_ptr->max_pkt;
2348 tipc_node_unlock(n_ptr);
2349 }
2350 return res;
2351 }
2352
2353 static void link_print(struct tipc_link *l_ptr, const char *str)
2354 {
2355 struct tipc_bearer *b_ptr;
2356
2357 rcu_read_lock();
2358 b_ptr = rcu_dereference_rtnl(bearer_list[l_ptr->bearer_id]);
2359 if (b_ptr)
2360 pr_info("%s Link %x<%s>:", str, l_ptr->addr, b_ptr->name);
2361 rcu_read_unlock();
2362
2363 if (link_working_unknown(l_ptr))
2364 pr_cont(":WU\n");
2365 else if (link_reset_reset(l_ptr))
2366 pr_cont(":RR\n");
2367 else if (link_reset_unknown(l_ptr))
2368 pr_cont(":RU\n");
2369 else if (link_working_working(l_ptr))
2370 pr_cont(":WW\n");
2371 else
2372 pr_cont("\n");
2373 }
Linux kernel - Deliverables
This page took 0.10811 seconds and 6 git commands to generate.