Loading...
1/*
2 * net/tipc/link.c: TIPC link code
3 *
4 * Copyright (c) 1996-2007, Ericsson AB
5 * Copyright (c) 2004-2007, 2010-2011, 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 "name_distr.h"
41#include "discover.h"
42#include "config.h"
43
44
45/*
46 * Out-of-range value for link session numbers
47 */
48
49#define INVALID_SESSION 0x10000
50
51/*
52 * Link state events:
53 */
54
55#define STARTING_EVT 856384768 /* link processing trigger */
56#define TRAFFIC_MSG_EVT 560815u /* rx'd ??? */
57#define TIMEOUT_EVT 560817u /* link timer expired */
58
59/*
60 * The following two 'message types' is really just implementation
61 * data conveniently stored in the message header.
62 * They must not be considered part of the protocol
63 */
64#define OPEN_MSG 0
65#define CLOSED_MSG 1
66
67/*
68 * State value stored in 'exp_msg_count'
69 */
70
71#define START_CHANGEOVER 100000u
72
73/**
74 * struct link_name - deconstructed link name
75 * @addr_local: network address of node at this end
76 * @if_local: name of interface at this end
77 * @addr_peer: network address of node at far end
78 * @if_peer: name of interface at far end
79 */
80
81struct link_name {
82 u32 addr_local;
83 char if_local[TIPC_MAX_IF_NAME];
84 u32 addr_peer;
85 char if_peer[TIPC_MAX_IF_NAME];
86};
87
88static void link_handle_out_of_seq_msg(struct link *l_ptr,
89 struct sk_buff *buf);
90static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf);
91static int link_recv_changeover_msg(struct link **l_ptr, struct sk_buff **buf);
92static void link_set_supervision_props(struct link *l_ptr, u32 tolerance);
93static int link_send_sections_long(struct tipc_port *sender,
94 struct iovec const *msg_sect,
95 u32 num_sect, unsigned int total_len,
96 u32 destnode);
97static void link_check_defragm_bufs(struct link *l_ptr);
98static void link_state_event(struct link *l_ptr, u32 event);
99static void link_reset_statistics(struct link *l_ptr);
100static void link_print(struct link *l_ptr, const char *str);
101static void link_start(struct link *l_ptr);
102static int link_send_long_buf(struct link *l_ptr, struct sk_buff *buf);
103
104/*
105 * Simple link routines
106 */
107
108static unsigned int align(unsigned int i)
109{
110 return (i + 3) & ~3u;
111}
112
113static void link_init_max_pkt(struct link *l_ptr)
114{
115 u32 max_pkt;
116
117 max_pkt = (l_ptr->b_ptr->mtu & ~3);
118 if (max_pkt > MAX_MSG_SIZE)
119 max_pkt = MAX_MSG_SIZE;
120
121 l_ptr->max_pkt_target = max_pkt;
122 if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
123 l_ptr->max_pkt = l_ptr->max_pkt_target;
124 else
125 l_ptr->max_pkt = MAX_PKT_DEFAULT;
126
127 l_ptr->max_pkt_probes = 0;
128}
129
130static u32 link_next_sent(struct link *l_ptr)
131{
132 if (l_ptr->next_out)
133 return msg_seqno(buf_msg(l_ptr->next_out));
134 return mod(l_ptr->next_out_no);
135}
136
137static u32 link_last_sent(struct link *l_ptr)
138{
139 return mod(link_next_sent(l_ptr) - 1);
140}
141
142/*
143 * Simple non-static link routines (i.e. referenced outside this file)
144 */
145
146int tipc_link_is_up(struct link *l_ptr)
147{
148 if (!l_ptr)
149 return 0;
150 return link_working_working(l_ptr) || link_working_unknown(l_ptr);
151}
152
153int tipc_link_is_active(struct link *l_ptr)
154{
155 return (l_ptr->owner->active_links[0] == l_ptr) ||
156 (l_ptr->owner->active_links[1] == l_ptr);
157}
158
159/**
160 * link_name_validate - validate & (optionally) deconstruct link name
161 * @name - ptr to link name string
162 * @name_parts - ptr to area for link name components (or NULL if not needed)
163 *
164 * Returns 1 if link name is valid, otherwise 0.
165 */
166
167static int link_name_validate(const char *name, struct link_name *name_parts)
168{
169 char name_copy[TIPC_MAX_LINK_NAME];
170 char *addr_local;
171 char *if_local;
172 char *addr_peer;
173 char *if_peer;
174 char dummy;
175 u32 z_local, c_local, n_local;
176 u32 z_peer, c_peer, n_peer;
177 u32 if_local_len;
178 u32 if_peer_len;
179
180 /* copy link name & ensure length is OK */
181
182 name_copy[TIPC_MAX_LINK_NAME - 1] = 0;
183 /* need above in case non-Posix strncpy() doesn't pad with nulls */
184 strncpy(name_copy, name, TIPC_MAX_LINK_NAME);
185 if (name_copy[TIPC_MAX_LINK_NAME - 1] != 0)
186 return 0;
187
188 /* ensure all component parts of link name are present */
189
190 addr_local = name_copy;
191 if_local = strchr(addr_local, ':');
192 if (if_local == NULL)
193 return 0;
194 *(if_local++) = 0;
195 addr_peer = strchr(if_local, '-');
196 if (addr_peer == NULL)
197 return 0;
198 *(addr_peer++) = 0;
199 if_local_len = addr_peer - if_local;
200 if_peer = strchr(addr_peer, ':');
201 if (if_peer == NULL)
202 return 0;
203 *(if_peer++) = 0;
204 if_peer_len = strlen(if_peer) + 1;
205
206 /* validate component parts of link name */
207
208 if ((sscanf(addr_local, "%u.%u.%u%c",
209 &z_local, &c_local, &n_local, &dummy) != 3) ||
210 (sscanf(addr_peer, "%u.%u.%u%c",
211 &z_peer, &c_peer, &n_peer, &dummy) != 3) ||
212 (z_local > 255) || (c_local > 4095) || (n_local > 4095) ||
213 (z_peer > 255) || (c_peer > 4095) || (n_peer > 4095) ||
214 (if_local_len <= 1) || (if_local_len > TIPC_MAX_IF_NAME) ||
215 (if_peer_len <= 1) || (if_peer_len > TIPC_MAX_IF_NAME) ||
216 (strspn(if_local, tipc_alphabet) != (if_local_len - 1)) ||
217 (strspn(if_peer, tipc_alphabet) != (if_peer_len - 1)))
218 return 0;
219
220 /* return link name components, if necessary */
221
222 if (name_parts) {
223 name_parts->addr_local = tipc_addr(z_local, c_local, n_local);
224 strcpy(name_parts->if_local, if_local);
225 name_parts->addr_peer = tipc_addr(z_peer, c_peer, n_peer);
226 strcpy(name_parts->if_peer, if_peer);
227 }
228 return 1;
229}
230
231/**
232 * link_timeout - handle expiration of link timer
233 * @l_ptr: pointer to link
234 *
235 * This routine must not grab "tipc_net_lock" to avoid a potential deadlock conflict
236 * with tipc_link_delete(). (There is no risk that the node will be deleted by
237 * another thread because tipc_link_delete() always cancels the link timer before
238 * tipc_node_delete() is called.)
239 */
240
241static void link_timeout(struct link *l_ptr)
242{
243 tipc_node_lock(l_ptr->owner);
244
245 /* update counters used in statistical profiling of send traffic */
246
247 l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
248 l_ptr->stats.queue_sz_counts++;
249
250 if (l_ptr->first_out) {
251 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
252 u32 length = msg_size(msg);
253
254 if ((msg_user(msg) == MSG_FRAGMENTER) &&
255 (msg_type(msg) == FIRST_FRAGMENT)) {
256 length = msg_size(msg_get_wrapped(msg));
257 }
258 if (length) {
259 l_ptr->stats.msg_lengths_total += length;
260 l_ptr->stats.msg_length_counts++;
261 if (length <= 64)
262 l_ptr->stats.msg_length_profile[0]++;
263 else if (length <= 256)
264 l_ptr->stats.msg_length_profile[1]++;
265 else if (length <= 1024)
266 l_ptr->stats.msg_length_profile[2]++;
267 else if (length <= 4096)
268 l_ptr->stats.msg_length_profile[3]++;
269 else if (length <= 16384)
270 l_ptr->stats.msg_length_profile[4]++;
271 else if (length <= 32768)
272 l_ptr->stats.msg_length_profile[5]++;
273 else
274 l_ptr->stats.msg_length_profile[6]++;
275 }
276 }
277
278 /* do all other link processing performed on a periodic basis */
279
280 link_check_defragm_bufs(l_ptr);
281
282 link_state_event(l_ptr, TIMEOUT_EVT);
283
284 if (l_ptr->next_out)
285 tipc_link_push_queue(l_ptr);
286
287 tipc_node_unlock(l_ptr->owner);
288}
289
290static void link_set_timer(struct link *l_ptr, u32 time)
291{
292 k_start_timer(&l_ptr->timer, time);
293}
294
295/**
296 * tipc_link_create - create a new link
297 * @n_ptr: pointer to associated node
298 * @b_ptr: pointer to associated bearer
299 * @media_addr: media address to use when sending messages over link
300 *
301 * Returns pointer to link.
302 */
303
304struct link *tipc_link_create(struct tipc_node *n_ptr,
305 struct tipc_bearer *b_ptr,
306 const struct tipc_media_addr *media_addr)
307{
308 struct link *l_ptr;
309 struct tipc_msg *msg;
310 char *if_name;
311 char addr_string[16];
312 u32 peer = n_ptr->addr;
313
314 if (n_ptr->link_cnt >= 2) {
315 tipc_addr_string_fill(addr_string, n_ptr->addr);
316 err("Attempt to establish third link to %s\n", addr_string);
317 return NULL;
318 }
319
320 if (n_ptr->links[b_ptr->identity]) {
321 tipc_addr_string_fill(addr_string, n_ptr->addr);
322 err("Attempt to establish second link on <%s> to %s\n",
323 b_ptr->name, addr_string);
324 return NULL;
325 }
326
327 l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
328 if (!l_ptr) {
329 warn("Link creation failed, no memory\n");
330 return NULL;
331 }
332
333 l_ptr->addr = peer;
334 if_name = strchr(b_ptr->name, ':') + 1;
335 sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:",
336 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
337 tipc_node(tipc_own_addr),
338 if_name,
339 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
340 /* note: peer i/f is appended to link name by reset/activate */
341 memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
342 l_ptr->owner = n_ptr;
343 l_ptr->checkpoint = 1;
344 l_ptr->b_ptr = b_ptr;
345 link_set_supervision_props(l_ptr, b_ptr->media->tolerance);
346 l_ptr->state = RESET_UNKNOWN;
347
348 l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
349 msg = l_ptr->pmsg;
350 tipc_msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
351 msg_set_size(msg, sizeof(l_ptr->proto_msg));
352 msg_set_session(msg, (tipc_random & 0xffff));
353 msg_set_bearer_id(msg, b_ptr->identity);
354 strcpy((char *)msg_data(msg), if_name);
355
356 l_ptr->priority = b_ptr->priority;
357 tipc_link_set_queue_limits(l_ptr, b_ptr->media->window);
358
359 link_init_max_pkt(l_ptr);
360
361 l_ptr->next_out_no = 1;
362 INIT_LIST_HEAD(&l_ptr->waiting_ports);
363
364 link_reset_statistics(l_ptr);
365
366 tipc_node_attach_link(n_ptr, l_ptr);
367
368 k_init_timer(&l_ptr->timer, (Handler)link_timeout, (unsigned long)l_ptr);
369 list_add_tail(&l_ptr->link_list, &b_ptr->links);
370 tipc_k_signal((Handler)link_start, (unsigned long)l_ptr);
371
372 return l_ptr;
373}
374
375/**
376 * tipc_link_delete - delete a link
377 * @l_ptr: pointer to link
378 *
379 * Note: 'tipc_net_lock' is write_locked, bearer is locked.
380 * This routine must not grab the node lock until after link timer cancellation
381 * to avoid a potential deadlock situation.
382 */
383
384void tipc_link_delete(struct link *l_ptr)
385{
386 if (!l_ptr) {
387 err("Attempt to delete non-existent link\n");
388 return;
389 }
390
391 k_cancel_timer(&l_ptr->timer);
392
393 tipc_node_lock(l_ptr->owner);
394 tipc_link_reset(l_ptr);
395 tipc_node_detach_link(l_ptr->owner, l_ptr);
396 tipc_link_stop(l_ptr);
397 list_del_init(&l_ptr->link_list);
398 tipc_node_unlock(l_ptr->owner);
399 k_term_timer(&l_ptr->timer);
400 kfree(l_ptr);
401}
402
403static void link_start(struct link *l_ptr)
404{
405 tipc_node_lock(l_ptr->owner);
406 link_state_event(l_ptr, STARTING_EVT);
407 tipc_node_unlock(l_ptr->owner);
408}
409
410/**
411 * link_schedule_port - schedule port for deferred sending
412 * @l_ptr: pointer to link
413 * @origport: reference to sending port
414 * @sz: amount of data to be sent
415 *
416 * Schedules port for renewed sending of messages after link congestion
417 * has abated.
418 */
419
420static int link_schedule_port(struct link *l_ptr, u32 origport, u32 sz)
421{
422 struct tipc_port *p_ptr;
423
424 spin_lock_bh(&tipc_port_list_lock);
425 p_ptr = tipc_port_lock(origport);
426 if (p_ptr) {
427 if (!p_ptr->wakeup)
428 goto exit;
429 if (!list_empty(&p_ptr->wait_list))
430 goto exit;
431 p_ptr->congested = 1;
432 p_ptr->waiting_pkts = 1 + ((sz - 1) / l_ptr->max_pkt);
433 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
434 l_ptr->stats.link_congs++;
435exit:
436 tipc_port_unlock(p_ptr);
437 }
438 spin_unlock_bh(&tipc_port_list_lock);
439 return -ELINKCONG;
440}
441
442void tipc_link_wakeup_ports(struct link *l_ptr, int all)
443{
444 struct tipc_port *p_ptr;
445 struct tipc_port *temp_p_ptr;
446 int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
447
448 if (all)
449 win = 100000;
450 if (win <= 0)
451 return;
452 if (!spin_trylock_bh(&tipc_port_list_lock))
453 return;
454 if (link_congested(l_ptr))
455 goto exit;
456 list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
457 wait_list) {
458 if (win <= 0)
459 break;
460 list_del_init(&p_ptr->wait_list);
461 spin_lock_bh(p_ptr->lock);
462 p_ptr->congested = 0;
463 p_ptr->wakeup(p_ptr);
464 win -= p_ptr->waiting_pkts;
465 spin_unlock_bh(p_ptr->lock);
466 }
467
468exit:
469 spin_unlock_bh(&tipc_port_list_lock);
470}
471
472/**
473 * link_release_outqueue - purge link's outbound message queue
474 * @l_ptr: pointer to link
475 */
476
477static void link_release_outqueue(struct link *l_ptr)
478{
479 struct sk_buff *buf = l_ptr->first_out;
480 struct sk_buff *next;
481
482 while (buf) {
483 next = buf->next;
484 buf_discard(buf);
485 buf = next;
486 }
487 l_ptr->first_out = NULL;
488 l_ptr->out_queue_size = 0;
489}
490
491/**
492 * tipc_link_reset_fragments - purge link's inbound message fragments queue
493 * @l_ptr: pointer to link
494 */
495
496void tipc_link_reset_fragments(struct link *l_ptr)
497{
498 struct sk_buff *buf = l_ptr->defragm_buf;
499 struct sk_buff *next;
500
501 while (buf) {
502 next = buf->next;
503 buf_discard(buf);
504 buf = next;
505 }
506 l_ptr->defragm_buf = NULL;
507}
508
509/**
510 * tipc_link_stop - purge all inbound and outbound messages associated with link
511 * @l_ptr: pointer to link
512 */
513
514void tipc_link_stop(struct link *l_ptr)
515{
516 struct sk_buff *buf;
517 struct sk_buff *next;
518
519 buf = l_ptr->oldest_deferred_in;
520 while (buf) {
521 next = buf->next;
522 buf_discard(buf);
523 buf = next;
524 }
525
526 buf = l_ptr->first_out;
527 while (buf) {
528 next = buf->next;
529 buf_discard(buf);
530 buf = next;
531 }
532
533 tipc_link_reset_fragments(l_ptr);
534
535 buf_discard(l_ptr->proto_msg_queue);
536 l_ptr->proto_msg_queue = NULL;
537}
538
539/* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
540#define link_send_event(fcn, l_ptr, up) do { } while (0)
541
542void tipc_link_reset(struct link *l_ptr)
543{
544 struct sk_buff *buf;
545 u32 prev_state = l_ptr->state;
546 u32 checkpoint = l_ptr->next_in_no;
547 int was_active_link = tipc_link_is_active(l_ptr);
548
549 msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
550
551 /* Link is down, accept any session */
552 l_ptr->peer_session = INVALID_SESSION;
553
554 /* Prepare for max packet size negotiation */
555 link_init_max_pkt(l_ptr);
556
557 l_ptr->state = RESET_UNKNOWN;
558
559 if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
560 return;
561
562 tipc_node_link_down(l_ptr->owner, l_ptr);
563 tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
564
565 if (was_active_link && tipc_node_active_links(l_ptr->owner) &&
566 l_ptr->owner->permit_changeover) {
567 l_ptr->reset_checkpoint = checkpoint;
568 l_ptr->exp_msg_count = START_CHANGEOVER;
569 }
570
571 /* Clean up all queues: */
572
573 link_release_outqueue(l_ptr);
574 buf_discard(l_ptr->proto_msg_queue);
575 l_ptr->proto_msg_queue = NULL;
576 buf = l_ptr->oldest_deferred_in;
577 while (buf) {
578 struct sk_buff *next = buf->next;
579 buf_discard(buf);
580 buf = next;
581 }
582 if (!list_empty(&l_ptr->waiting_ports))
583 tipc_link_wakeup_ports(l_ptr, 1);
584
585 l_ptr->retransm_queue_head = 0;
586 l_ptr->retransm_queue_size = 0;
587 l_ptr->last_out = NULL;
588 l_ptr->first_out = NULL;
589 l_ptr->next_out = NULL;
590 l_ptr->unacked_window = 0;
591 l_ptr->checkpoint = 1;
592 l_ptr->next_out_no = 1;
593 l_ptr->deferred_inqueue_sz = 0;
594 l_ptr->oldest_deferred_in = NULL;
595 l_ptr->newest_deferred_in = NULL;
596 l_ptr->fsm_msg_cnt = 0;
597 l_ptr->stale_count = 0;
598 link_reset_statistics(l_ptr);
599
600 link_send_event(tipc_cfg_link_event, l_ptr, 0);
601 if (!in_own_cluster(l_ptr->addr))
602 link_send_event(tipc_disc_link_event, l_ptr, 0);
603}
604
605
606static void link_activate(struct link *l_ptr)
607{
608 l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
609 tipc_node_link_up(l_ptr->owner, l_ptr);
610 tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
611 link_send_event(tipc_cfg_link_event, l_ptr, 1);
612 if (!in_own_cluster(l_ptr->addr))
613 link_send_event(tipc_disc_link_event, l_ptr, 1);
614}
615
616/**
617 * link_state_event - link finite state machine
618 * @l_ptr: pointer to link
619 * @event: state machine event to process
620 */
621
622static void link_state_event(struct link *l_ptr, unsigned event)
623{
624 struct link *other;
625 u32 cont_intv = l_ptr->continuity_interval;
626
627 if (!l_ptr->started && (event != STARTING_EVT))
628 return; /* Not yet. */
629
630 if (link_blocked(l_ptr)) {
631 if (event == TIMEOUT_EVT)
632 link_set_timer(l_ptr, cont_intv);
633 return; /* Changeover going on */
634 }
635
636 switch (l_ptr->state) {
637 case WORKING_WORKING:
638 switch (event) {
639 case TRAFFIC_MSG_EVT:
640 case ACTIVATE_MSG:
641 break;
642 case TIMEOUT_EVT:
643 if (l_ptr->next_in_no != l_ptr->checkpoint) {
644 l_ptr->checkpoint = l_ptr->next_in_no;
645 if (tipc_bclink_acks_missing(l_ptr->owner)) {
646 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
647 0, 0, 0, 0, 0);
648 l_ptr->fsm_msg_cnt++;
649 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
650 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
651 1, 0, 0, 0, 0);
652 l_ptr->fsm_msg_cnt++;
653 }
654 link_set_timer(l_ptr, cont_intv);
655 break;
656 }
657 l_ptr->state = WORKING_UNKNOWN;
658 l_ptr->fsm_msg_cnt = 0;
659 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
660 l_ptr->fsm_msg_cnt++;
661 link_set_timer(l_ptr, cont_intv / 4);
662 break;
663 case RESET_MSG:
664 info("Resetting link <%s>, requested by peer\n",
665 l_ptr->name);
666 tipc_link_reset(l_ptr);
667 l_ptr->state = RESET_RESET;
668 l_ptr->fsm_msg_cnt = 0;
669 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
670 l_ptr->fsm_msg_cnt++;
671 link_set_timer(l_ptr, cont_intv);
672 break;
673 default:
674 err("Unknown link event %u in WW state\n", event);
675 }
676 break;
677 case WORKING_UNKNOWN:
678 switch (event) {
679 case TRAFFIC_MSG_EVT:
680 case ACTIVATE_MSG:
681 l_ptr->state = WORKING_WORKING;
682 l_ptr->fsm_msg_cnt = 0;
683 link_set_timer(l_ptr, cont_intv);
684 break;
685 case RESET_MSG:
686 info("Resetting link <%s>, requested by peer "
687 "while probing\n", l_ptr->name);
688 tipc_link_reset(l_ptr);
689 l_ptr->state = RESET_RESET;
690 l_ptr->fsm_msg_cnt = 0;
691 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
692 l_ptr->fsm_msg_cnt++;
693 link_set_timer(l_ptr, cont_intv);
694 break;
695 case TIMEOUT_EVT:
696 if (l_ptr->next_in_no != l_ptr->checkpoint) {
697 l_ptr->state = WORKING_WORKING;
698 l_ptr->fsm_msg_cnt = 0;
699 l_ptr->checkpoint = l_ptr->next_in_no;
700 if (tipc_bclink_acks_missing(l_ptr->owner)) {
701 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
702 0, 0, 0, 0, 0);
703 l_ptr->fsm_msg_cnt++;
704 }
705 link_set_timer(l_ptr, cont_intv);
706 } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
707 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
708 1, 0, 0, 0, 0);
709 l_ptr->fsm_msg_cnt++;
710 link_set_timer(l_ptr, cont_intv / 4);
711 } else { /* Link has failed */
712 warn("Resetting link <%s>, peer not responding\n",
713 l_ptr->name);
714 tipc_link_reset(l_ptr);
715 l_ptr->state = RESET_UNKNOWN;
716 l_ptr->fsm_msg_cnt = 0;
717 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
718 0, 0, 0, 0, 0);
719 l_ptr->fsm_msg_cnt++;
720 link_set_timer(l_ptr, cont_intv);
721 }
722 break;
723 default:
724 err("Unknown link event %u in WU state\n", event);
725 }
726 break;
727 case RESET_UNKNOWN:
728 switch (event) {
729 case TRAFFIC_MSG_EVT:
730 break;
731 case ACTIVATE_MSG:
732 other = l_ptr->owner->active_links[0];
733 if (other && link_working_unknown(other))
734 break;
735 l_ptr->state = WORKING_WORKING;
736 l_ptr->fsm_msg_cnt = 0;
737 link_activate(l_ptr);
738 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
739 l_ptr->fsm_msg_cnt++;
740 link_set_timer(l_ptr, cont_intv);
741 break;
742 case RESET_MSG:
743 l_ptr->state = RESET_RESET;
744 l_ptr->fsm_msg_cnt = 0;
745 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
746 l_ptr->fsm_msg_cnt++;
747 link_set_timer(l_ptr, cont_intv);
748 break;
749 case STARTING_EVT:
750 l_ptr->started = 1;
751 /* fall through */
752 case TIMEOUT_EVT:
753 tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
754 l_ptr->fsm_msg_cnt++;
755 link_set_timer(l_ptr, cont_intv);
756 break;
757 default:
758 err("Unknown link event %u in RU state\n", event);
759 }
760 break;
761 case RESET_RESET:
762 switch (event) {
763 case TRAFFIC_MSG_EVT:
764 case ACTIVATE_MSG:
765 other = l_ptr->owner->active_links[0];
766 if (other && link_working_unknown(other))
767 break;
768 l_ptr->state = WORKING_WORKING;
769 l_ptr->fsm_msg_cnt = 0;
770 link_activate(l_ptr);
771 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
772 l_ptr->fsm_msg_cnt++;
773 link_set_timer(l_ptr, cont_intv);
774 break;
775 case RESET_MSG:
776 break;
777 case TIMEOUT_EVT:
778 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
779 l_ptr->fsm_msg_cnt++;
780 link_set_timer(l_ptr, cont_intv);
781 break;
782 default:
783 err("Unknown link event %u in RR state\n", event);
784 }
785 break;
786 default:
787 err("Unknown link state %u/%u\n", l_ptr->state, event);
788 }
789}
790
791/*
792 * link_bundle_buf(): Append contents of a buffer to
793 * the tail of an existing one.
794 */
795
796static int link_bundle_buf(struct link *l_ptr,
797 struct sk_buff *bundler,
798 struct sk_buff *buf)
799{
800 struct tipc_msg *bundler_msg = buf_msg(bundler);
801 struct tipc_msg *msg = buf_msg(buf);
802 u32 size = msg_size(msg);
803 u32 bundle_size = msg_size(bundler_msg);
804 u32 to_pos = align(bundle_size);
805 u32 pad = to_pos - bundle_size;
806
807 if (msg_user(bundler_msg) != MSG_BUNDLER)
808 return 0;
809 if (msg_type(bundler_msg) != OPEN_MSG)
810 return 0;
811 if (skb_tailroom(bundler) < (pad + size))
812 return 0;
813 if (l_ptr->max_pkt < (to_pos + size))
814 return 0;
815
816 skb_put(bundler, pad + size);
817 skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
818 msg_set_size(bundler_msg, to_pos + size);
819 msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
820 buf_discard(buf);
821 l_ptr->stats.sent_bundled++;
822 return 1;
823}
824
825static void link_add_to_outqueue(struct link *l_ptr,
826 struct sk_buff *buf,
827 struct tipc_msg *msg)
828{
829 u32 ack = mod(l_ptr->next_in_no - 1);
830 u32 seqno = mod(l_ptr->next_out_no++);
831
832 msg_set_word(msg, 2, ((ack << 16) | seqno));
833 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
834 buf->next = NULL;
835 if (l_ptr->first_out) {
836 l_ptr->last_out->next = buf;
837 l_ptr->last_out = buf;
838 } else
839 l_ptr->first_out = l_ptr->last_out = buf;
840
841 l_ptr->out_queue_size++;
842 if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
843 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
844}
845
846static void link_add_chain_to_outqueue(struct link *l_ptr,
847 struct sk_buff *buf_chain,
848 u32 long_msgno)
849{
850 struct sk_buff *buf;
851 struct tipc_msg *msg;
852
853 if (!l_ptr->next_out)
854 l_ptr->next_out = buf_chain;
855 while (buf_chain) {
856 buf = buf_chain;
857 buf_chain = buf_chain->next;
858
859 msg = buf_msg(buf);
860 msg_set_long_msgno(msg, long_msgno);
861 link_add_to_outqueue(l_ptr, buf, msg);
862 }
863}
864
865/*
866 * tipc_link_send_buf() is the 'full path' for messages, called from
867 * inside TIPC when the 'fast path' in tipc_send_buf
868 * has failed, and from link_send()
869 */
870
871int tipc_link_send_buf(struct link *l_ptr, struct sk_buff *buf)
872{
873 struct tipc_msg *msg = buf_msg(buf);
874 u32 size = msg_size(msg);
875 u32 dsz = msg_data_sz(msg);
876 u32 queue_size = l_ptr->out_queue_size;
877 u32 imp = tipc_msg_tot_importance(msg);
878 u32 queue_limit = l_ptr->queue_limit[imp];
879 u32 max_packet = l_ptr->max_pkt;
880
881 msg_set_prevnode(msg, tipc_own_addr); /* If routed message */
882
883 /* Match msg importance against queue limits: */
884
885 if (unlikely(queue_size >= queue_limit)) {
886 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
887 link_schedule_port(l_ptr, msg_origport(msg), size);
888 buf_discard(buf);
889 return -ELINKCONG;
890 }
891 buf_discard(buf);
892 if (imp > CONN_MANAGER) {
893 warn("Resetting link <%s>, send queue full", l_ptr->name);
894 tipc_link_reset(l_ptr);
895 }
896 return dsz;
897 }
898
899 /* Fragmentation needed ? */
900
901 if (size > max_packet)
902 return link_send_long_buf(l_ptr, buf);
903
904 /* Packet can be queued or sent: */
905
906 if (likely(!tipc_bearer_congested(l_ptr->b_ptr, l_ptr) &&
907 !link_congested(l_ptr))) {
908 link_add_to_outqueue(l_ptr, buf, msg);
909
910 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr))) {
911 l_ptr->unacked_window = 0;
912 } else {
913 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
914 l_ptr->stats.bearer_congs++;
915 l_ptr->next_out = buf;
916 }
917 return dsz;
918 }
919 /* Congestion: can message be bundled ?: */
920
921 if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
922 (msg_user(msg) != MSG_FRAGMENTER)) {
923
924 /* Try adding message to an existing bundle */
925
926 if (l_ptr->next_out &&
927 link_bundle_buf(l_ptr, l_ptr->last_out, buf)) {
928 tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
929 return dsz;
930 }
931
932 /* Try creating a new bundle */
933
934 if (size <= max_packet * 2 / 3) {
935 struct sk_buff *bundler = tipc_buf_acquire(max_packet);
936 struct tipc_msg bundler_hdr;
937
938 if (bundler) {
939 tipc_msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
940 INT_H_SIZE, l_ptr->addr);
941 skb_copy_to_linear_data(bundler, &bundler_hdr,
942 INT_H_SIZE);
943 skb_trim(bundler, INT_H_SIZE);
944 link_bundle_buf(l_ptr, bundler, buf);
945 buf = bundler;
946 msg = buf_msg(buf);
947 l_ptr->stats.sent_bundles++;
948 }
949 }
950 }
951 if (!l_ptr->next_out)
952 l_ptr->next_out = buf;
953 link_add_to_outqueue(l_ptr, buf, msg);
954 tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
955 return dsz;
956}
957
958/*
959 * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
960 * not been selected yet, and the the owner node is not locked
961 * Called by TIPC internal users, e.g. the name distributor
962 */
963
964int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
965{
966 struct link *l_ptr;
967 struct tipc_node *n_ptr;
968 int res = -ELINKCONG;
969
970 read_lock_bh(&tipc_net_lock);
971 n_ptr = tipc_node_find(dest);
972 if (n_ptr) {
973 tipc_node_lock(n_ptr);
974 l_ptr = n_ptr->active_links[selector & 1];
975 if (l_ptr)
976 res = tipc_link_send_buf(l_ptr, buf);
977 else
978 buf_discard(buf);
979 tipc_node_unlock(n_ptr);
980 } else {
981 buf_discard(buf);
982 }
983 read_unlock_bh(&tipc_net_lock);
984 return res;
985}
986
987/*
988 * link_send_buf_fast: Entry for data messages where the
989 * destination link is known and the header is complete,
990 * inclusive total message length. Very time critical.
991 * Link is locked. Returns user data length.
992 */
993
994static int link_send_buf_fast(struct link *l_ptr, struct sk_buff *buf,
995 u32 *used_max_pkt)
996{
997 struct tipc_msg *msg = buf_msg(buf);
998 int res = msg_data_sz(msg);
999
1000 if (likely(!link_congested(l_ptr))) {
1001 if (likely(msg_size(msg) <= l_ptr->max_pkt)) {
1002 if (likely(list_empty(&l_ptr->b_ptr->cong_links))) {
1003 link_add_to_outqueue(l_ptr, buf, msg);
1004 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf,
1005 &l_ptr->media_addr))) {
1006 l_ptr->unacked_window = 0;
1007 return res;
1008 }
1009 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1010 l_ptr->stats.bearer_congs++;
1011 l_ptr->next_out = buf;
1012 return res;
1013 }
1014 } else
1015 *used_max_pkt = l_ptr->max_pkt;
1016 }
1017 return tipc_link_send_buf(l_ptr, buf); /* All other cases */
1018}
1019
1020/*
1021 * tipc_send_buf_fast: Entry for data messages where the
1022 * destination node is known and the header is complete,
1023 * inclusive total message length.
1024 * Returns user data length.
1025 */
1026int tipc_send_buf_fast(struct sk_buff *buf, u32 destnode)
1027{
1028 struct link *l_ptr;
1029 struct tipc_node *n_ptr;
1030 int res;
1031 u32 selector = msg_origport(buf_msg(buf)) & 1;
1032 u32 dummy;
1033
1034 if (destnode == tipc_own_addr)
1035 return tipc_port_recv_msg(buf);
1036
1037 read_lock_bh(&tipc_net_lock);
1038 n_ptr = tipc_node_find(destnode);
1039 if (likely(n_ptr)) {
1040 tipc_node_lock(n_ptr);
1041 l_ptr = n_ptr->active_links[selector];
1042 if (likely(l_ptr)) {
1043 res = link_send_buf_fast(l_ptr, buf, &dummy);
1044 tipc_node_unlock(n_ptr);
1045 read_unlock_bh(&tipc_net_lock);
1046 return res;
1047 }
1048 tipc_node_unlock(n_ptr);
1049 }
1050 read_unlock_bh(&tipc_net_lock);
1051 res = msg_data_sz(buf_msg(buf));
1052 tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1053 return res;
1054}
1055
1056
1057/*
1058 * tipc_link_send_sections_fast: Entry for messages where the
1059 * destination processor is known and the header is complete,
1060 * except for total message length.
1061 * Returns user data length or errno.
1062 */
1063int tipc_link_send_sections_fast(struct tipc_port *sender,
1064 struct iovec const *msg_sect,
1065 const u32 num_sect,
1066 unsigned int total_len,
1067 u32 destaddr)
1068{
1069 struct tipc_msg *hdr = &sender->phdr;
1070 struct link *l_ptr;
1071 struct sk_buff *buf;
1072 struct tipc_node *node;
1073 int res;
1074 u32 selector = msg_origport(hdr) & 1;
1075
1076again:
1077 /*
1078 * Try building message using port's max_pkt hint.
1079 * (Must not hold any locks while building message.)
1080 */
1081
1082 res = tipc_msg_build(hdr, msg_sect, num_sect, total_len,
1083 sender->max_pkt, !sender->user_port, &buf);
1084
1085 read_lock_bh(&tipc_net_lock);
1086 node = tipc_node_find(destaddr);
1087 if (likely(node)) {
1088 tipc_node_lock(node);
1089 l_ptr = node->active_links[selector];
1090 if (likely(l_ptr)) {
1091 if (likely(buf)) {
1092 res = link_send_buf_fast(l_ptr, buf,
1093 &sender->max_pkt);
1094exit:
1095 tipc_node_unlock(node);
1096 read_unlock_bh(&tipc_net_lock);
1097 return res;
1098 }
1099
1100 /* Exit if build request was invalid */
1101
1102 if (unlikely(res < 0))
1103 goto exit;
1104
1105 /* Exit if link (or bearer) is congested */
1106
1107 if (link_congested(l_ptr) ||
1108 !list_empty(&l_ptr->b_ptr->cong_links)) {
1109 res = link_schedule_port(l_ptr,
1110 sender->ref, res);
1111 goto exit;
1112 }
1113
1114 /*
1115 * Message size exceeds max_pkt hint; update hint,
1116 * then re-try fast path or fragment the message
1117 */
1118
1119 sender->max_pkt = l_ptr->max_pkt;
1120 tipc_node_unlock(node);
1121 read_unlock_bh(&tipc_net_lock);
1122
1123
1124 if ((msg_hdr_sz(hdr) + res) <= sender->max_pkt)
1125 goto again;
1126
1127 return link_send_sections_long(sender, msg_sect,
1128 num_sect, total_len,
1129 destaddr);
1130 }
1131 tipc_node_unlock(node);
1132 }
1133 read_unlock_bh(&tipc_net_lock);
1134
1135 /* Couldn't find a link to the destination node */
1136
1137 if (buf)
1138 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1139 if (res >= 0)
1140 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1141 total_len, TIPC_ERR_NO_NODE);
1142 return res;
1143}
1144
1145/*
1146 * link_send_sections_long(): Entry for long messages where the
1147 * destination node is known and the header is complete,
1148 * inclusive total message length.
1149 * Link and bearer congestion status have been checked to be ok,
1150 * and are ignored if they change.
1151 *
1152 * Note that fragments do not use the full link MTU so that they won't have
1153 * to undergo refragmentation if link changeover causes them to be sent
1154 * over another link with an additional tunnel header added as prefix.
1155 * (Refragmentation will still occur if the other link has a smaller MTU.)
1156 *
1157 * Returns user data length or errno.
1158 */
1159static int link_send_sections_long(struct tipc_port *sender,
1160 struct iovec const *msg_sect,
1161 u32 num_sect,
1162 unsigned int total_len,
1163 u32 destaddr)
1164{
1165 struct link *l_ptr;
1166 struct tipc_node *node;
1167 struct tipc_msg *hdr = &sender->phdr;
1168 u32 dsz = total_len;
1169 u32 max_pkt, fragm_sz, rest;
1170 struct tipc_msg fragm_hdr;
1171 struct sk_buff *buf, *buf_chain, *prev;
1172 u32 fragm_crs, fragm_rest, hsz, sect_rest;
1173 const unchar *sect_crs;
1174 int curr_sect;
1175 u32 fragm_no;
1176
1177again:
1178 fragm_no = 1;
1179 max_pkt = sender->max_pkt - INT_H_SIZE;
1180 /* leave room for tunnel header in case of link changeover */
1181 fragm_sz = max_pkt - INT_H_SIZE;
1182 /* leave room for fragmentation header in each fragment */
1183 rest = dsz;
1184 fragm_crs = 0;
1185 fragm_rest = 0;
1186 sect_rest = 0;
1187 sect_crs = NULL;
1188 curr_sect = -1;
1189
1190 /* Prepare reusable fragment header: */
1191
1192 tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1193 INT_H_SIZE, msg_destnode(hdr));
1194 msg_set_size(&fragm_hdr, max_pkt);
1195 msg_set_fragm_no(&fragm_hdr, 1);
1196
1197 /* Prepare header of first fragment: */
1198
1199 buf_chain = buf = tipc_buf_acquire(max_pkt);
1200 if (!buf)
1201 return -ENOMEM;
1202 buf->next = NULL;
1203 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1204 hsz = msg_hdr_sz(hdr);
1205 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1206
1207 /* Chop up message: */
1208
1209 fragm_crs = INT_H_SIZE + hsz;
1210 fragm_rest = fragm_sz - hsz;
1211
1212 do { /* For all sections */
1213 u32 sz;
1214
1215 if (!sect_rest) {
1216 sect_rest = msg_sect[++curr_sect].iov_len;
1217 sect_crs = (const unchar *)msg_sect[curr_sect].iov_base;
1218 }
1219
1220 if (sect_rest < fragm_rest)
1221 sz = sect_rest;
1222 else
1223 sz = fragm_rest;
1224
1225 if (likely(!sender->user_port)) {
1226 if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1227error:
1228 for (; buf_chain; buf_chain = buf) {
1229 buf = buf_chain->next;
1230 buf_discard(buf_chain);
1231 }
1232 return -EFAULT;
1233 }
1234 } else
1235 skb_copy_to_linear_data_offset(buf, fragm_crs,
1236 sect_crs, sz);
1237 sect_crs += sz;
1238 sect_rest -= sz;
1239 fragm_crs += sz;
1240 fragm_rest -= sz;
1241 rest -= sz;
1242
1243 if (!fragm_rest && rest) {
1244
1245 /* Initiate new fragment: */
1246 if (rest <= fragm_sz) {
1247 fragm_sz = rest;
1248 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
1249 } else {
1250 msg_set_type(&fragm_hdr, FRAGMENT);
1251 }
1252 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1253 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1254 prev = buf;
1255 buf = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
1256 if (!buf)
1257 goto error;
1258
1259 buf->next = NULL;
1260 prev->next = buf;
1261 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1262 fragm_crs = INT_H_SIZE;
1263 fragm_rest = fragm_sz;
1264 }
1265 } while (rest > 0);
1266
1267 /*
1268 * Now we have a buffer chain. Select a link and check
1269 * that packet size is still OK
1270 */
1271 node = tipc_node_find(destaddr);
1272 if (likely(node)) {
1273 tipc_node_lock(node);
1274 l_ptr = node->active_links[sender->ref & 1];
1275 if (!l_ptr) {
1276 tipc_node_unlock(node);
1277 goto reject;
1278 }
1279 if (l_ptr->max_pkt < max_pkt) {
1280 sender->max_pkt = l_ptr->max_pkt;
1281 tipc_node_unlock(node);
1282 for (; buf_chain; buf_chain = buf) {
1283 buf = buf_chain->next;
1284 buf_discard(buf_chain);
1285 }
1286 goto again;
1287 }
1288 } else {
1289reject:
1290 for (; buf_chain; buf_chain = buf) {
1291 buf = buf_chain->next;
1292 buf_discard(buf_chain);
1293 }
1294 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1295 total_len, TIPC_ERR_NO_NODE);
1296 }
1297
1298 /* Append chain of fragments to send queue & send them */
1299
1300 l_ptr->long_msg_seq_no++;
1301 link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
1302 l_ptr->stats.sent_fragments += fragm_no;
1303 l_ptr->stats.sent_fragmented++;
1304 tipc_link_push_queue(l_ptr);
1305 tipc_node_unlock(node);
1306 return dsz;
1307}
1308
1309/*
1310 * tipc_link_push_packet: Push one unsent packet to the media
1311 */
1312u32 tipc_link_push_packet(struct link *l_ptr)
1313{
1314 struct sk_buff *buf = l_ptr->first_out;
1315 u32 r_q_size = l_ptr->retransm_queue_size;
1316 u32 r_q_head = l_ptr->retransm_queue_head;
1317
1318 /* Step to position where retransmission failed, if any, */
1319 /* consider that buffers may have been released in meantime */
1320
1321 if (r_q_size && buf) {
1322 u32 last = lesser(mod(r_q_head + r_q_size),
1323 link_last_sent(l_ptr));
1324 u32 first = msg_seqno(buf_msg(buf));
1325
1326 while (buf && less(first, r_q_head)) {
1327 first = mod(first + 1);
1328 buf = buf->next;
1329 }
1330 l_ptr->retransm_queue_head = r_q_head = first;
1331 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1332 }
1333
1334 /* Continue retransmission now, if there is anything: */
1335
1336 if (r_q_size && buf) {
1337 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1338 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1339 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1340 l_ptr->retransm_queue_head = mod(++r_q_head);
1341 l_ptr->retransm_queue_size = --r_q_size;
1342 l_ptr->stats.retransmitted++;
1343 return 0;
1344 } else {
1345 l_ptr->stats.bearer_congs++;
1346 return PUSH_FAILED;
1347 }
1348 }
1349
1350 /* Send deferred protocol message, if any: */
1351
1352 buf = l_ptr->proto_msg_queue;
1353 if (buf) {
1354 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1355 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1356 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1357 l_ptr->unacked_window = 0;
1358 buf_discard(buf);
1359 l_ptr->proto_msg_queue = NULL;
1360 return 0;
1361 } else {
1362 l_ptr->stats.bearer_congs++;
1363 return PUSH_FAILED;
1364 }
1365 }
1366
1367 /* Send one deferred data message, if send window not full: */
1368
1369 buf = l_ptr->next_out;
1370 if (buf) {
1371 struct tipc_msg *msg = buf_msg(buf);
1372 u32 next = msg_seqno(msg);
1373 u32 first = msg_seqno(buf_msg(l_ptr->first_out));
1374
1375 if (mod(next - first) < l_ptr->queue_limit[0]) {
1376 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1377 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1378 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1379 if (msg_user(msg) == MSG_BUNDLER)
1380 msg_set_type(msg, CLOSED_MSG);
1381 l_ptr->next_out = buf->next;
1382 return 0;
1383 } else {
1384 l_ptr->stats.bearer_congs++;
1385 return PUSH_FAILED;
1386 }
1387 }
1388 }
1389 return PUSH_FINISHED;
1390}
1391
1392/*
1393 * push_queue(): push out the unsent messages of a link where
1394 * congestion has abated. Node is locked
1395 */
1396void tipc_link_push_queue(struct link *l_ptr)
1397{
1398 u32 res;
1399
1400 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr))
1401 return;
1402
1403 do {
1404 res = tipc_link_push_packet(l_ptr);
1405 } while (!res);
1406
1407 if (res == PUSH_FAILED)
1408 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1409}
1410
1411static void link_reset_all(unsigned long addr)
1412{
1413 struct tipc_node *n_ptr;
1414 char addr_string[16];
1415 u32 i;
1416
1417 read_lock_bh(&tipc_net_lock);
1418 n_ptr = tipc_node_find((u32)addr);
1419 if (!n_ptr) {
1420 read_unlock_bh(&tipc_net_lock);
1421 return; /* node no longer exists */
1422 }
1423
1424 tipc_node_lock(n_ptr);
1425
1426 warn("Resetting all links to %s\n",
1427 tipc_addr_string_fill(addr_string, n_ptr->addr));
1428
1429 for (i = 0; i < MAX_BEARERS; i++) {
1430 if (n_ptr->links[i]) {
1431 link_print(n_ptr->links[i], "Resetting link\n");
1432 tipc_link_reset(n_ptr->links[i]);
1433 }
1434 }
1435
1436 tipc_node_unlock(n_ptr);
1437 read_unlock_bh(&tipc_net_lock);
1438}
1439
1440static void link_retransmit_failure(struct link *l_ptr, struct sk_buff *buf)
1441{
1442 struct tipc_msg *msg = buf_msg(buf);
1443
1444 warn("Retransmission failure on link <%s>\n", l_ptr->name);
1445
1446 if (l_ptr->addr) {
1447
1448 /* Handle failure on standard link */
1449
1450 link_print(l_ptr, "Resetting link\n");
1451 tipc_link_reset(l_ptr);
1452
1453 } else {
1454
1455 /* Handle failure on broadcast link */
1456
1457 struct tipc_node *n_ptr;
1458 char addr_string[16];
1459
1460 info("Msg seq number: %u, ", msg_seqno(msg));
1461 info("Outstanding acks: %lu\n",
1462 (unsigned long) TIPC_SKB_CB(buf)->handle);
1463
1464 n_ptr = tipc_bclink_retransmit_to();
1465 tipc_node_lock(n_ptr);
1466
1467 tipc_addr_string_fill(addr_string, n_ptr->addr);
1468 info("Multicast link info for %s\n", addr_string);
1469 info("Supported: %d, ", n_ptr->bclink.supported);
1470 info("Acked: %u\n", n_ptr->bclink.acked);
1471 info("Last in: %u, ", n_ptr->bclink.last_in);
1472 info("Gap after: %u, ", n_ptr->bclink.gap_after);
1473 info("Gap to: %u\n", n_ptr->bclink.gap_to);
1474 info("Nack sync: %u\n\n", n_ptr->bclink.nack_sync);
1475
1476 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1477
1478 tipc_node_unlock(n_ptr);
1479
1480 l_ptr->stale_count = 0;
1481 }
1482}
1483
1484void tipc_link_retransmit(struct link *l_ptr, struct sk_buff *buf,
1485 u32 retransmits)
1486{
1487 struct tipc_msg *msg;
1488
1489 if (!buf)
1490 return;
1491
1492 msg = buf_msg(buf);
1493
1494 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1495 if (l_ptr->retransm_queue_size == 0) {
1496 l_ptr->retransm_queue_head = msg_seqno(msg);
1497 l_ptr->retransm_queue_size = retransmits;
1498 } else {
1499 err("Unexpected retransmit on link %s (qsize=%d)\n",
1500 l_ptr->name, l_ptr->retransm_queue_size);
1501 }
1502 return;
1503 } else {
1504 /* Detect repeated retransmit failures on uncongested bearer */
1505
1506 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1507 if (++l_ptr->stale_count > 100) {
1508 link_retransmit_failure(l_ptr, buf);
1509 return;
1510 }
1511 } else {
1512 l_ptr->last_retransmitted = msg_seqno(msg);
1513 l_ptr->stale_count = 1;
1514 }
1515 }
1516
1517 while (retransmits && (buf != l_ptr->next_out) && buf) {
1518 msg = buf_msg(buf);
1519 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1520 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1521 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1522 buf = buf->next;
1523 retransmits--;
1524 l_ptr->stats.retransmitted++;
1525 } else {
1526 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1527 l_ptr->stats.bearer_congs++;
1528 l_ptr->retransm_queue_head = msg_seqno(buf_msg(buf));
1529 l_ptr->retransm_queue_size = retransmits;
1530 return;
1531 }
1532 }
1533
1534 l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1535}
1536
1537/**
1538 * link_insert_deferred_queue - insert deferred messages back into receive chain
1539 */
1540
1541static struct sk_buff *link_insert_deferred_queue(struct link *l_ptr,
1542 struct sk_buff *buf)
1543{
1544 u32 seq_no;
1545
1546 if (l_ptr->oldest_deferred_in == NULL)
1547 return buf;
1548
1549 seq_no = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1550 if (seq_no == mod(l_ptr->next_in_no)) {
1551 l_ptr->newest_deferred_in->next = buf;
1552 buf = l_ptr->oldest_deferred_in;
1553 l_ptr->oldest_deferred_in = NULL;
1554 l_ptr->deferred_inqueue_sz = 0;
1555 }
1556 return buf;
1557}
1558
1559/**
1560 * link_recv_buf_validate - validate basic format of received message
1561 *
1562 * This routine ensures a TIPC message has an acceptable header, and at least
1563 * as much data as the header indicates it should. The routine also ensures
1564 * that the entire message header is stored in the main fragment of the message
1565 * buffer, to simplify future access to message header fields.
1566 *
1567 * Note: Having extra info present in the message header or data areas is OK.
1568 * TIPC will ignore the excess, under the assumption that it is optional info
1569 * introduced by a later release of the protocol.
1570 */
1571
1572static int link_recv_buf_validate(struct sk_buff *buf)
1573{
1574 static u32 min_data_hdr_size[8] = {
1575 SHORT_H_SIZE, MCAST_H_SIZE, NAMED_H_SIZE, BASIC_H_SIZE,
1576 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1577 };
1578
1579 struct tipc_msg *msg;
1580 u32 tipc_hdr[2];
1581 u32 size;
1582 u32 hdr_size;
1583 u32 min_hdr_size;
1584
1585 if (unlikely(buf->len < MIN_H_SIZE))
1586 return 0;
1587
1588 msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1589 if (msg == NULL)
1590 return 0;
1591
1592 if (unlikely(msg_version(msg) != TIPC_VERSION))
1593 return 0;
1594
1595 size = msg_size(msg);
1596 hdr_size = msg_hdr_sz(msg);
1597 min_hdr_size = msg_isdata(msg) ?
1598 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1599
1600 if (unlikely((hdr_size < min_hdr_size) ||
1601 (size < hdr_size) ||
1602 (buf->len < size) ||
1603 (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1604 return 0;
1605
1606 return pskb_may_pull(buf, hdr_size);
1607}
1608
1609/**
1610 * tipc_recv_msg - process TIPC messages arriving from off-node
1611 * @head: pointer to message buffer chain
1612 * @tb_ptr: pointer to bearer message arrived on
1613 *
1614 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1615 * structure (i.e. cannot be NULL), but bearer can be inactive.
1616 */
1617
1618void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *b_ptr)
1619{
1620 read_lock_bh(&tipc_net_lock);
1621 while (head) {
1622 struct tipc_node *n_ptr;
1623 struct link *l_ptr;
1624 struct sk_buff *crs;
1625 struct sk_buff *buf = head;
1626 struct tipc_msg *msg;
1627 u32 seq_no;
1628 u32 ackd;
1629 u32 released = 0;
1630 int type;
1631
1632 head = head->next;
1633
1634 /* Ensure bearer is still enabled */
1635
1636 if (unlikely(!b_ptr->active))
1637 goto cont;
1638
1639 /* Ensure message is well-formed */
1640
1641 if (unlikely(!link_recv_buf_validate(buf)))
1642 goto cont;
1643
1644 /* Ensure message data is a single contiguous unit */
1645
1646 if (unlikely(buf_linearize(buf)))
1647 goto cont;
1648
1649 /* Handle arrival of a non-unicast link message */
1650
1651 msg = buf_msg(buf);
1652
1653 if (unlikely(msg_non_seq(msg))) {
1654 if (msg_user(msg) == LINK_CONFIG)
1655 tipc_disc_recv_msg(buf, b_ptr);
1656 else
1657 tipc_bclink_recv_pkt(buf);
1658 continue;
1659 }
1660
1661 if (unlikely(!msg_short(msg) &&
1662 (msg_destnode(msg) != tipc_own_addr)))
1663 goto cont;
1664
1665 /* Discard non-routeable messages destined for another node */
1666
1667 if (unlikely(!msg_isdata(msg) &&
1668 (msg_destnode(msg) != tipc_own_addr))) {
1669 if ((msg_user(msg) != CONN_MANAGER) &&
1670 (msg_user(msg) != MSG_FRAGMENTER))
1671 goto cont;
1672 }
1673
1674 /* Locate neighboring node that sent message */
1675
1676 n_ptr = tipc_node_find(msg_prevnode(msg));
1677 if (unlikely(!n_ptr))
1678 goto cont;
1679 tipc_node_lock(n_ptr);
1680
1681 /* Don't talk to neighbor during cleanup after last session */
1682
1683 if (n_ptr->cleanup_required) {
1684 tipc_node_unlock(n_ptr);
1685 goto cont;
1686 }
1687
1688 /* Locate unicast link endpoint that should handle message */
1689
1690 l_ptr = n_ptr->links[b_ptr->identity];
1691 if (unlikely(!l_ptr)) {
1692 tipc_node_unlock(n_ptr);
1693 goto cont;
1694 }
1695
1696 /* Validate message sequence number info */
1697
1698 seq_no = msg_seqno(msg);
1699 ackd = msg_ack(msg);
1700
1701 /* Release acked messages */
1702
1703 if (less(n_ptr->bclink.acked, msg_bcast_ack(msg))) {
1704 if (tipc_node_is_up(n_ptr) && n_ptr->bclink.supported)
1705 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1706 }
1707
1708 crs = l_ptr->first_out;
1709 while ((crs != l_ptr->next_out) &&
1710 less_eq(msg_seqno(buf_msg(crs)), ackd)) {
1711 struct sk_buff *next = crs->next;
1712
1713 buf_discard(crs);
1714 crs = next;
1715 released++;
1716 }
1717 if (released) {
1718 l_ptr->first_out = crs;
1719 l_ptr->out_queue_size -= released;
1720 }
1721
1722 /* Try sending any messages link endpoint has pending */
1723
1724 if (unlikely(l_ptr->next_out))
1725 tipc_link_push_queue(l_ptr);
1726 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1727 tipc_link_wakeup_ports(l_ptr, 0);
1728 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1729 l_ptr->stats.sent_acks++;
1730 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1731 }
1732
1733 /* Now (finally!) process the incoming message */
1734
1735protocol_check:
1736 if (likely(link_working_working(l_ptr))) {
1737 if (likely(seq_no == mod(l_ptr->next_in_no))) {
1738 l_ptr->next_in_no++;
1739 if (unlikely(l_ptr->oldest_deferred_in))
1740 head = link_insert_deferred_queue(l_ptr,
1741 head);
1742 if (likely(msg_is_dest(msg, tipc_own_addr))) {
1743deliver:
1744 if (likely(msg_isdata(msg))) {
1745 tipc_node_unlock(n_ptr);
1746 tipc_port_recv_msg(buf);
1747 continue;
1748 }
1749 switch (msg_user(msg)) {
1750 case MSG_BUNDLER:
1751 l_ptr->stats.recv_bundles++;
1752 l_ptr->stats.recv_bundled +=
1753 msg_msgcnt(msg);
1754 tipc_node_unlock(n_ptr);
1755 tipc_link_recv_bundle(buf);
1756 continue;
1757 case NAME_DISTRIBUTOR:
1758 tipc_node_unlock(n_ptr);
1759 tipc_named_recv(buf);
1760 continue;
1761 case CONN_MANAGER:
1762 tipc_node_unlock(n_ptr);
1763 tipc_port_recv_proto_msg(buf);
1764 continue;
1765 case MSG_FRAGMENTER:
1766 l_ptr->stats.recv_fragments++;
1767 if (tipc_link_recv_fragment(&l_ptr->defragm_buf,
1768 &buf, &msg)) {
1769 l_ptr->stats.recv_fragmented++;
1770 goto deliver;
1771 }
1772 break;
1773 case CHANGEOVER_PROTOCOL:
1774 type = msg_type(msg);
1775 if (link_recv_changeover_msg(&l_ptr, &buf)) {
1776 msg = buf_msg(buf);
1777 seq_no = msg_seqno(msg);
1778 if (type == ORIGINAL_MSG)
1779 goto deliver;
1780 goto protocol_check;
1781 }
1782 break;
1783 default:
1784 buf_discard(buf);
1785 buf = NULL;
1786 break;
1787 }
1788 }
1789 tipc_node_unlock(n_ptr);
1790 tipc_net_route_msg(buf);
1791 continue;
1792 }
1793 link_handle_out_of_seq_msg(l_ptr, buf);
1794 head = link_insert_deferred_queue(l_ptr, head);
1795 tipc_node_unlock(n_ptr);
1796 continue;
1797 }
1798
1799 if (msg_user(msg) == LINK_PROTOCOL) {
1800 link_recv_proto_msg(l_ptr, buf);
1801 head = link_insert_deferred_queue(l_ptr, head);
1802 tipc_node_unlock(n_ptr);
1803 continue;
1804 }
1805 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1806
1807 if (link_working_working(l_ptr)) {
1808 /* Re-insert in front of queue */
1809 buf->next = head;
1810 head = buf;
1811 tipc_node_unlock(n_ptr);
1812 continue;
1813 }
1814 tipc_node_unlock(n_ptr);
1815cont:
1816 buf_discard(buf);
1817 }
1818 read_unlock_bh(&tipc_net_lock);
1819}
1820
1821/*
1822 * link_defer_buf(): Sort a received out-of-sequence packet
1823 * into the deferred reception queue.
1824 * Returns the increase of the queue length,i.e. 0 or 1
1825 */
1826
1827u32 tipc_link_defer_pkt(struct sk_buff **head,
1828 struct sk_buff **tail,
1829 struct sk_buff *buf)
1830{
1831 struct sk_buff *prev = NULL;
1832 struct sk_buff *crs = *head;
1833 u32 seq_no = msg_seqno(buf_msg(buf));
1834
1835 buf->next = NULL;
1836
1837 /* Empty queue ? */
1838 if (*head == NULL) {
1839 *head = *tail = buf;
1840 return 1;
1841 }
1842
1843 /* Last ? */
1844 if (less(msg_seqno(buf_msg(*tail)), seq_no)) {
1845 (*tail)->next = buf;
1846 *tail = buf;
1847 return 1;
1848 }
1849
1850 /* Scan through queue and sort it in */
1851 do {
1852 struct tipc_msg *msg = buf_msg(crs);
1853
1854 if (less(seq_no, msg_seqno(msg))) {
1855 buf->next = crs;
1856 if (prev)
1857 prev->next = buf;
1858 else
1859 *head = buf;
1860 return 1;
1861 }
1862 if (seq_no == msg_seqno(msg))
1863 break;
1864 prev = crs;
1865 crs = crs->next;
1866 } while (crs);
1867
1868 /* Message is a duplicate of an existing message */
1869
1870 buf_discard(buf);
1871 return 0;
1872}
1873
1874/**
1875 * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
1876 */
1877
1878static void link_handle_out_of_seq_msg(struct link *l_ptr,
1879 struct sk_buff *buf)
1880{
1881 u32 seq_no = msg_seqno(buf_msg(buf));
1882
1883 if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
1884 link_recv_proto_msg(l_ptr, buf);
1885 return;
1886 }
1887
1888 /* Record OOS packet arrival (force mismatch on next timeout) */
1889
1890 l_ptr->checkpoint--;
1891
1892 /*
1893 * Discard packet if a duplicate; otherwise add it to deferred queue
1894 * and notify peer of gap as per protocol specification
1895 */
1896
1897 if (less(seq_no, mod(l_ptr->next_in_no))) {
1898 l_ptr->stats.duplicates++;
1899 buf_discard(buf);
1900 return;
1901 }
1902
1903 if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
1904 &l_ptr->newest_deferred_in, buf)) {
1905 l_ptr->deferred_inqueue_sz++;
1906 l_ptr->stats.deferred_recv++;
1907 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
1908 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1909 } else
1910 l_ptr->stats.duplicates++;
1911}
1912
1913/*
1914 * Send protocol message to the other endpoint.
1915 */
1916void tipc_link_send_proto_msg(struct link *l_ptr, u32 msg_typ, int probe_msg,
1917 u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
1918{
1919 struct sk_buff *buf = NULL;
1920 struct tipc_msg *msg = l_ptr->pmsg;
1921 u32 msg_size = sizeof(l_ptr->proto_msg);
1922 int r_flag;
1923
1924 if (link_blocked(l_ptr))
1925 return;
1926 msg_set_type(msg, msg_typ);
1927 msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
1928 msg_set_bcast_ack(msg, mod(l_ptr->owner->bclink.last_in));
1929 msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
1930
1931 if (msg_typ == STATE_MSG) {
1932 u32 next_sent = mod(l_ptr->next_out_no);
1933
1934 if (!tipc_link_is_up(l_ptr))
1935 return;
1936 if (l_ptr->next_out)
1937 next_sent = msg_seqno(buf_msg(l_ptr->next_out));
1938 msg_set_next_sent(msg, next_sent);
1939 if (l_ptr->oldest_deferred_in) {
1940 u32 rec = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1941 gap = mod(rec - mod(l_ptr->next_in_no));
1942 }
1943 msg_set_seq_gap(msg, gap);
1944 if (gap)
1945 l_ptr->stats.sent_nacks++;
1946 msg_set_link_tolerance(msg, tolerance);
1947 msg_set_linkprio(msg, priority);
1948 msg_set_max_pkt(msg, ack_mtu);
1949 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1950 msg_set_probe(msg, probe_msg != 0);
1951 if (probe_msg) {
1952 u32 mtu = l_ptr->max_pkt;
1953
1954 if ((mtu < l_ptr->max_pkt_target) &&
1955 link_working_working(l_ptr) &&
1956 l_ptr->fsm_msg_cnt) {
1957 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1958 if (l_ptr->max_pkt_probes == 10) {
1959 l_ptr->max_pkt_target = (msg_size - 4);
1960 l_ptr->max_pkt_probes = 0;
1961 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1962 }
1963 l_ptr->max_pkt_probes++;
1964 }
1965
1966 l_ptr->stats.sent_probes++;
1967 }
1968 l_ptr->stats.sent_states++;
1969 } else { /* RESET_MSG or ACTIVATE_MSG */
1970 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
1971 msg_set_seq_gap(msg, 0);
1972 msg_set_next_sent(msg, 1);
1973 msg_set_probe(msg, 0);
1974 msg_set_link_tolerance(msg, l_ptr->tolerance);
1975 msg_set_linkprio(msg, l_ptr->priority);
1976 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
1977 }
1978
1979 r_flag = (l_ptr->owner->working_links > tipc_link_is_up(l_ptr));
1980 msg_set_redundant_link(msg, r_flag);
1981 msg_set_linkprio(msg, l_ptr->priority);
1982
1983 /* Ensure sequence number will not fit : */
1984
1985 msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
1986
1987 /* Congestion? */
1988
1989 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1990 if (!l_ptr->proto_msg_queue) {
1991 l_ptr->proto_msg_queue =
1992 tipc_buf_acquire(sizeof(l_ptr->proto_msg));
1993 }
1994 buf = l_ptr->proto_msg_queue;
1995 if (!buf)
1996 return;
1997 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1998 return;
1999 }
2000
2001 /* Message can be sent */
2002
2003 buf = tipc_buf_acquire(msg_size);
2004 if (!buf)
2005 return;
2006
2007 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2008 msg_set_size(buf_msg(buf), msg_size);
2009
2010 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
2011 l_ptr->unacked_window = 0;
2012 buf_discard(buf);
2013 return;
2014 }
2015
2016 /* New congestion */
2017 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
2018 l_ptr->proto_msg_queue = buf;
2019 l_ptr->stats.bearer_congs++;
2020}
2021
2022/*
2023 * Receive protocol message :
2024 * Note that network plane id propagates through the network, and may
2025 * change at any time. The node with lowest address rules
2026 */
2027
2028static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf)
2029{
2030 u32 rec_gap = 0;
2031 u32 max_pkt_info;
2032 u32 max_pkt_ack;
2033 u32 msg_tol;
2034 struct tipc_msg *msg = buf_msg(buf);
2035
2036 if (link_blocked(l_ptr))
2037 goto exit;
2038
2039 /* record unnumbered packet arrival (force mismatch on next timeout) */
2040
2041 l_ptr->checkpoint--;
2042
2043 if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
2044 if (tipc_own_addr > msg_prevnode(msg))
2045 l_ptr->b_ptr->net_plane = msg_net_plane(msg);
2046
2047 l_ptr->owner->permit_changeover = msg_redundant_link(msg);
2048
2049 switch (msg_type(msg)) {
2050
2051 case RESET_MSG:
2052 if (!link_working_unknown(l_ptr) &&
2053 (l_ptr->peer_session != INVALID_SESSION)) {
2054 if (msg_session(msg) == l_ptr->peer_session)
2055 break; /* duplicate: ignore */
2056 }
2057 /* fall thru' */
2058 case ACTIVATE_MSG:
2059 /* Update link settings according other endpoint's values */
2060
2061 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2062
2063 msg_tol = msg_link_tolerance(msg);
2064 if (msg_tol > l_ptr->tolerance)
2065 link_set_supervision_props(l_ptr, msg_tol);
2066
2067 if (msg_linkprio(msg) > l_ptr->priority)
2068 l_ptr->priority = msg_linkprio(msg);
2069
2070 max_pkt_info = msg_max_pkt(msg);
2071 if (max_pkt_info) {
2072 if (max_pkt_info < l_ptr->max_pkt_target)
2073 l_ptr->max_pkt_target = max_pkt_info;
2074 if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2075 l_ptr->max_pkt = l_ptr->max_pkt_target;
2076 } else {
2077 l_ptr->max_pkt = l_ptr->max_pkt_target;
2078 }
2079 l_ptr->owner->bclink.supported = (max_pkt_info != 0);
2080
2081 link_state_event(l_ptr, msg_type(msg));
2082
2083 l_ptr->peer_session = msg_session(msg);
2084 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2085
2086 /* Synchronize broadcast sequence numbers */
2087 if (!tipc_node_redundant_links(l_ptr->owner))
2088 l_ptr->owner->bclink.last_in = mod(msg_last_bcast(msg));
2089 break;
2090 case STATE_MSG:
2091
2092 msg_tol = msg_link_tolerance(msg);
2093 if (msg_tol)
2094 link_set_supervision_props(l_ptr, msg_tol);
2095
2096 if (msg_linkprio(msg) &&
2097 (msg_linkprio(msg) != l_ptr->priority)) {
2098 warn("Resetting link <%s>, priority change %u->%u\n",
2099 l_ptr->name, l_ptr->priority, msg_linkprio(msg));
2100 l_ptr->priority = msg_linkprio(msg);
2101 tipc_link_reset(l_ptr); /* Enforce change to take effect */
2102 break;
2103 }
2104 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2105 l_ptr->stats.recv_states++;
2106 if (link_reset_unknown(l_ptr))
2107 break;
2108
2109 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2110 rec_gap = mod(msg_next_sent(msg) -
2111 mod(l_ptr->next_in_no));
2112 }
2113
2114 max_pkt_ack = msg_max_pkt(msg);
2115 if (max_pkt_ack > l_ptr->max_pkt) {
2116 l_ptr->max_pkt = max_pkt_ack;
2117 l_ptr->max_pkt_probes = 0;
2118 }
2119
2120 max_pkt_ack = 0;
2121 if (msg_probe(msg)) {
2122 l_ptr->stats.recv_probes++;
2123 if (msg_size(msg) > sizeof(l_ptr->proto_msg))
2124 max_pkt_ack = msg_size(msg);
2125 }
2126
2127 /* Protocol message before retransmits, reduce loss risk */
2128
2129 tipc_bclink_check_gap(l_ptr->owner, msg_last_bcast(msg));
2130
2131 if (rec_gap || (msg_probe(msg))) {
2132 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2133 0, rec_gap, 0, 0, max_pkt_ack);
2134 }
2135 if (msg_seq_gap(msg)) {
2136 l_ptr->stats.recv_nacks++;
2137 tipc_link_retransmit(l_ptr, l_ptr->first_out,
2138 msg_seq_gap(msg));
2139 }
2140 break;
2141 }
2142exit:
2143 buf_discard(buf);
2144}
2145
2146
2147/*
2148 * tipc_link_tunnel(): Send one message via a link belonging to
2149 * another bearer. Owner node is locked.
2150 */
2151static void tipc_link_tunnel(struct link *l_ptr,
2152 struct tipc_msg *tunnel_hdr,
2153 struct tipc_msg *msg,
2154 u32 selector)
2155{
2156 struct link *tunnel;
2157 struct sk_buff *buf;
2158 u32 length = msg_size(msg);
2159
2160 tunnel = l_ptr->owner->active_links[selector & 1];
2161 if (!tipc_link_is_up(tunnel)) {
2162 warn("Link changeover error, "
2163 "tunnel link no longer available\n");
2164 return;
2165 }
2166 msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2167 buf = tipc_buf_acquire(length + INT_H_SIZE);
2168 if (!buf) {
2169 warn("Link changeover error, "
2170 "unable to send tunnel msg\n");
2171 return;
2172 }
2173 skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
2174 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
2175 tipc_link_send_buf(tunnel, buf);
2176}
2177
2178
2179
2180/*
2181 * changeover(): Send whole message queue via the remaining link
2182 * Owner node is locked.
2183 */
2184
2185void tipc_link_changeover(struct link *l_ptr)
2186{
2187 u32 msgcount = l_ptr->out_queue_size;
2188 struct sk_buff *crs = l_ptr->first_out;
2189 struct link *tunnel = l_ptr->owner->active_links[0];
2190 struct tipc_msg tunnel_hdr;
2191 int split_bundles;
2192
2193 if (!tunnel)
2194 return;
2195
2196 if (!l_ptr->owner->permit_changeover) {
2197 warn("Link changeover error, "
2198 "peer did not permit changeover\n");
2199 return;
2200 }
2201
2202 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2203 ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
2204 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2205 msg_set_msgcnt(&tunnel_hdr, msgcount);
2206
2207 if (!l_ptr->first_out) {
2208 struct sk_buff *buf;
2209
2210 buf = tipc_buf_acquire(INT_H_SIZE);
2211 if (buf) {
2212 skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2213 msg_set_size(&tunnel_hdr, INT_H_SIZE);
2214 tipc_link_send_buf(tunnel, buf);
2215 } else {
2216 warn("Link changeover error, "
2217 "unable to send changeover msg\n");
2218 }
2219 return;
2220 }
2221
2222 split_bundles = (l_ptr->owner->active_links[0] !=
2223 l_ptr->owner->active_links[1]);
2224
2225 while (crs) {
2226 struct tipc_msg *msg = buf_msg(crs);
2227
2228 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2229 struct tipc_msg *m = msg_get_wrapped(msg);
2230 unchar *pos = (unchar *)m;
2231
2232 msgcount = msg_msgcnt(msg);
2233 while (msgcount--) {
2234 msg_set_seqno(m, msg_seqno(msg));
2235 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2236 msg_link_selector(m));
2237 pos += align(msg_size(m));
2238 m = (struct tipc_msg *)pos;
2239 }
2240 } else {
2241 tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2242 msg_link_selector(msg));
2243 }
2244 crs = crs->next;
2245 }
2246}
2247
2248void tipc_link_send_duplicate(struct link *l_ptr, struct link *tunnel)
2249{
2250 struct sk_buff *iter;
2251 struct tipc_msg tunnel_hdr;
2252
2253 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2254 DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
2255 msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2256 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2257 iter = l_ptr->first_out;
2258 while (iter) {
2259 struct sk_buff *outbuf;
2260 struct tipc_msg *msg = buf_msg(iter);
2261 u32 length = msg_size(msg);
2262
2263 if (msg_user(msg) == MSG_BUNDLER)
2264 msg_set_type(msg, CLOSED_MSG);
2265 msg_set_ack(msg, mod(l_ptr->next_in_no - 1)); /* Update */
2266 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2267 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2268 outbuf = tipc_buf_acquire(length + INT_H_SIZE);
2269 if (outbuf == NULL) {
2270 warn("Link changeover error, "
2271 "unable to send duplicate msg\n");
2272 return;
2273 }
2274 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2275 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2276 length);
2277 tipc_link_send_buf(tunnel, outbuf);
2278 if (!tipc_link_is_up(l_ptr))
2279 return;
2280 iter = iter->next;
2281 }
2282}
2283
2284
2285
2286/**
2287 * buf_extract - extracts embedded TIPC message from another message
2288 * @skb: encapsulating message buffer
2289 * @from_pos: offset to extract from
2290 *
2291 * Returns a new message buffer containing an embedded message. The
2292 * encapsulating message itself is left unchanged.
2293 */
2294
2295static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2296{
2297 struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2298 u32 size = msg_size(msg);
2299 struct sk_buff *eb;
2300
2301 eb = tipc_buf_acquire(size);
2302 if (eb)
2303 skb_copy_to_linear_data(eb, msg, size);
2304 return eb;
2305}
2306
2307/*
2308 * link_recv_changeover_msg(): Receive tunneled packet sent
2309 * via other link. Node is locked. Return extracted buffer.
2310 */
2311
2312static int link_recv_changeover_msg(struct link **l_ptr,
2313 struct sk_buff **buf)
2314{
2315 struct sk_buff *tunnel_buf = *buf;
2316 struct link *dest_link;
2317 struct tipc_msg *msg;
2318 struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2319 u32 msg_typ = msg_type(tunnel_msg);
2320 u32 msg_count = msg_msgcnt(tunnel_msg);
2321
2322 dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2323 if (!dest_link)
2324 goto exit;
2325 if (dest_link == *l_ptr) {
2326 err("Unexpected changeover message on link <%s>\n",
2327 (*l_ptr)->name);
2328 goto exit;
2329 }
2330 *l_ptr = dest_link;
2331 msg = msg_get_wrapped(tunnel_msg);
2332
2333 if (msg_typ == DUPLICATE_MSG) {
2334 if (less(msg_seqno(msg), mod(dest_link->next_in_no)))
2335 goto exit;
2336 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2337 if (*buf == NULL) {
2338 warn("Link changeover error, duplicate msg dropped\n");
2339 goto exit;
2340 }
2341 buf_discard(tunnel_buf);
2342 return 1;
2343 }
2344
2345 /* First original message ?: */
2346
2347 if (tipc_link_is_up(dest_link)) {
2348 info("Resetting link <%s>, changeover initiated by peer\n",
2349 dest_link->name);
2350 tipc_link_reset(dest_link);
2351 dest_link->exp_msg_count = msg_count;
2352 if (!msg_count)
2353 goto exit;
2354 } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2355 dest_link->exp_msg_count = msg_count;
2356 if (!msg_count)
2357 goto exit;
2358 }
2359
2360 /* Receive original message */
2361
2362 if (dest_link->exp_msg_count == 0) {
2363 warn("Link switchover error, "
2364 "got too many tunnelled messages\n");
2365 goto exit;
2366 }
2367 dest_link->exp_msg_count--;
2368 if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2369 goto exit;
2370 } else {
2371 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2372 if (*buf != NULL) {
2373 buf_discard(tunnel_buf);
2374 return 1;
2375 } else {
2376 warn("Link changeover error, original msg dropped\n");
2377 }
2378 }
2379exit:
2380 *buf = NULL;
2381 buf_discard(tunnel_buf);
2382 return 0;
2383}
2384
2385/*
2386 * Bundler functionality:
2387 */
2388void tipc_link_recv_bundle(struct sk_buff *buf)
2389{
2390 u32 msgcount = msg_msgcnt(buf_msg(buf));
2391 u32 pos = INT_H_SIZE;
2392 struct sk_buff *obuf;
2393
2394 while (msgcount--) {
2395 obuf = buf_extract(buf, pos);
2396 if (obuf == NULL) {
2397 warn("Link unable to unbundle message(s)\n");
2398 break;
2399 }
2400 pos += align(msg_size(buf_msg(obuf)));
2401 tipc_net_route_msg(obuf);
2402 }
2403 buf_discard(buf);
2404}
2405
2406/*
2407 * Fragmentation/defragmentation:
2408 */
2409
2410
2411/*
2412 * link_send_long_buf: Entry for buffers needing fragmentation.
2413 * The buffer is complete, inclusive total message length.
2414 * Returns user data length.
2415 */
2416static int link_send_long_buf(struct link *l_ptr, struct sk_buff *buf)
2417{
2418 struct sk_buff *buf_chain = NULL;
2419 struct sk_buff *buf_chain_tail = (struct sk_buff *)&buf_chain;
2420 struct tipc_msg *inmsg = buf_msg(buf);
2421 struct tipc_msg fragm_hdr;
2422 u32 insize = msg_size(inmsg);
2423 u32 dsz = msg_data_sz(inmsg);
2424 unchar *crs = buf->data;
2425 u32 rest = insize;
2426 u32 pack_sz = l_ptr->max_pkt;
2427 u32 fragm_sz = pack_sz - INT_H_SIZE;
2428 u32 fragm_no = 0;
2429 u32 destaddr;
2430
2431 if (msg_short(inmsg))
2432 destaddr = l_ptr->addr;
2433 else
2434 destaddr = msg_destnode(inmsg);
2435
2436 /* Prepare reusable fragment header: */
2437
2438 tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2439 INT_H_SIZE, destaddr);
2440
2441 /* Chop up message: */
2442
2443 while (rest > 0) {
2444 struct sk_buff *fragm;
2445
2446 if (rest <= fragm_sz) {
2447 fragm_sz = rest;
2448 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2449 }
2450 fragm = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
2451 if (fragm == NULL) {
2452 buf_discard(buf);
2453 while (buf_chain) {
2454 buf = buf_chain;
2455 buf_chain = buf_chain->next;
2456 buf_discard(buf);
2457 }
2458 return -ENOMEM;
2459 }
2460 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2461 fragm_no++;
2462 msg_set_fragm_no(&fragm_hdr, fragm_no);
2463 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2464 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2465 fragm_sz);
2466 buf_chain_tail->next = fragm;
2467 buf_chain_tail = fragm;
2468
2469 rest -= fragm_sz;
2470 crs += fragm_sz;
2471 msg_set_type(&fragm_hdr, FRAGMENT);
2472 }
2473 buf_discard(buf);
2474
2475 /* Append chain of fragments to send queue & send them */
2476
2477 l_ptr->long_msg_seq_no++;
2478 link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
2479 l_ptr->stats.sent_fragments += fragm_no;
2480 l_ptr->stats.sent_fragmented++;
2481 tipc_link_push_queue(l_ptr);
2482
2483 return dsz;
2484}
2485
2486/*
2487 * A pending message being re-assembled must store certain values
2488 * to handle subsequent fragments correctly. The following functions
2489 * help storing these values in unused, available fields in the
2490 * pending message. This makes dynamic memory allocation unnecessary.
2491 */
2492
2493static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2494{
2495 msg_set_seqno(buf_msg(buf), seqno);
2496}
2497
2498static u32 get_fragm_size(struct sk_buff *buf)
2499{
2500 return msg_ack(buf_msg(buf));
2501}
2502
2503static void set_fragm_size(struct sk_buff *buf, u32 sz)
2504{
2505 msg_set_ack(buf_msg(buf), sz);
2506}
2507
2508static u32 get_expected_frags(struct sk_buff *buf)
2509{
2510 return msg_bcast_ack(buf_msg(buf));
2511}
2512
2513static void set_expected_frags(struct sk_buff *buf, u32 exp)
2514{
2515 msg_set_bcast_ack(buf_msg(buf), exp);
2516}
2517
2518static u32 get_timer_cnt(struct sk_buff *buf)
2519{
2520 return msg_reroute_cnt(buf_msg(buf));
2521}
2522
2523static void incr_timer_cnt(struct sk_buff *buf)
2524{
2525 msg_incr_reroute_cnt(buf_msg(buf));
2526}
2527
2528/*
2529 * tipc_link_recv_fragment(): Called with node lock on. Returns
2530 * the reassembled buffer if message is complete.
2531 */
2532int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2533 struct tipc_msg **m)
2534{
2535 struct sk_buff *prev = NULL;
2536 struct sk_buff *fbuf = *fb;
2537 struct tipc_msg *fragm = buf_msg(fbuf);
2538 struct sk_buff *pbuf = *pending;
2539 u32 long_msg_seq_no = msg_long_msgno(fragm);
2540
2541 *fb = NULL;
2542
2543 /* Is there an incomplete message waiting for this fragment? */
2544
2545 while (pbuf && ((msg_seqno(buf_msg(pbuf)) != long_msg_seq_no) ||
2546 (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2547 prev = pbuf;
2548 pbuf = pbuf->next;
2549 }
2550
2551 if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2552 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2553 u32 msg_sz = msg_size(imsg);
2554 u32 fragm_sz = msg_data_sz(fragm);
2555 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2556 u32 max = TIPC_MAX_USER_MSG_SIZE + NAMED_H_SIZE;
2557 if (msg_type(imsg) == TIPC_MCAST_MSG)
2558 max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2559 if (msg_size(imsg) > max) {
2560 buf_discard(fbuf);
2561 return 0;
2562 }
2563 pbuf = tipc_buf_acquire(msg_size(imsg));
2564 if (pbuf != NULL) {
2565 pbuf->next = *pending;
2566 *pending = pbuf;
2567 skb_copy_to_linear_data(pbuf, imsg,
2568 msg_data_sz(fragm));
2569 /* Prepare buffer for subsequent fragments. */
2570
2571 set_long_msg_seqno(pbuf, long_msg_seq_no);
2572 set_fragm_size(pbuf, fragm_sz);
2573 set_expected_frags(pbuf, exp_fragm_cnt - 1);
2574 } else {
2575 warn("Link unable to reassemble fragmented message\n");
2576 }
2577 buf_discard(fbuf);
2578 return 0;
2579 } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2580 u32 dsz = msg_data_sz(fragm);
2581 u32 fsz = get_fragm_size(pbuf);
2582 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2583 u32 exp_frags = get_expected_frags(pbuf) - 1;
2584 skb_copy_to_linear_data_offset(pbuf, crs,
2585 msg_data(fragm), dsz);
2586 buf_discard(fbuf);
2587
2588 /* Is message complete? */
2589
2590 if (exp_frags == 0) {
2591 if (prev)
2592 prev->next = pbuf->next;
2593 else
2594 *pending = pbuf->next;
2595 msg_reset_reroute_cnt(buf_msg(pbuf));
2596 *fb = pbuf;
2597 *m = buf_msg(pbuf);
2598 return 1;
2599 }
2600 set_expected_frags(pbuf, exp_frags);
2601 return 0;
2602 }
2603 buf_discard(fbuf);
2604 return 0;
2605}
2606
2607/**
2608 * link_check_defragm_bufs - flush stale incoming message fragments
2609 * @l_ptr: pointer to link
2610 */
2611
2612static void link_check_defragm_bufs(struct link *l_ptr)
2613{
2614 struct sk_buff *prev = NULL;
2615 struct sk_buff *next = NULL;
2616 struct sk_buff *buf = l_ptr->defragm_buf;
2617
2618 if (!buf)
2619 return;
2620 if (!link_working_working(l_ptr))
2621 return;
2622 while (buf) {
2623 u32 cnt = get_timer_cnt(buf);
2624
2625 next = buf->next;
2626 if (cnt < 4) {
2627 incr_timer_cnt(buf);
2628 prev = buf;
2629 } else {
2630 if (prev)
2631 prev->next = buf->next;
2632 else
2633 l_ptr->defragm_buf = buf->next;
2634 buf_discard(buf);
2635 }
2636 buf = next;
2637 }
2638}
2639
2640
2641
2642static void link_set_supervision_props(struct link *l_ptr, u32 tolerance)
2643{
2644 if ((tolerance < TIPC_MIN_LINK_TOL) || (tolerance > TIPC_MAX_LINK_TOL))
2645 return;
2646
2647 l_ptr->tolerance = tolerance;
2648 l_ptr->continuity_interval =
2649 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2650 l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2651}
2652
2653
2654void tipc_link_set_queue_limits(struct link *l_ptr, u32 window)
2655{
2656 /* Data messages from this node, inclusive FIRST_FRAGM */
2657 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2658 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2659 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2660 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2661 /* Transiting data messages,inclusive FIRST_FRAGM */
2662 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2663 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2664 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2665 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2666 l_ptr->queue_limit[CONN_MANAGER] = 1200;
2667 l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2668 l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2669 /* FRAGMENT and LAST_FRAGMENT packets */
2670 l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2671}
2672
2673/**
2674 * link_find_link - locate link by name
2675 * @name - ptr to link name string
2676 * @node - ptr to area to be filled with ptr to associated node
2677 *
2678 * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2679 * this also prevents link deletion.
2680 *
2681 * Returns pointer to link (or 0 if invalid link name).
2682 */
2683
2684static struct link *link_find_link(const char *name, struct tipc_node **node)
2685{
2686 struct link_name link_name_parts;
2687 struct tipc_bearer *b_ptr;
2688 struct link *l_ptr;
2689
2690 if (!link_name_validate(name, &link_name_parts))
2691 return NULL;
2692
2693 b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2694 if (!b_ptr)
2695 return NULL;
2696
2697 *node = tipc_node_find(link_name_parts.addr_peer);
2698 if (!*node)
2699 return NULL;
2700
2701 l_ptr = (*node)->links[b_ptr->identity];
2702 if (!l_ptr || strcmp(l_ptr->name, name))
2703 return NULL;
2704
2705 return l_ptr;
2706}
2707
2708struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2709 u16 cmd)
2710{
2711 struct tipc_link_config *args;
2712 u32 new_value;
2713 struct link *l_ptr;
2714 struct tipc_node *node;
2715 int res;
2716
2717 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2718 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2719
2720 args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2721 new_value = ntohl(args->value);
2722
2723 if (!strcmp(args->name, tipc_bclink_name)) {
2724 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2725 (tipc_bclink_set_queue_limits(new_value) == 0))
2726 return tipc_cfg_reply_none();
2727 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2728 " (cannot change setting on broadcast link)");
2729 }
2730
2731 read_lock_bh(&tipc_net_lock);
2732 l_ptr = link_find_link(args->name, &node);
2733 if (!l_ptr) {
2734 read_unlock_bh(&tipc_net_lock);
2735 return tipc_cfg_reply_error_string("link not found");
2736 }
2737
2738 tipc_node_lock(node);
2739 res = -EINVAL;
2740 switch (cmd) {
2741 case TIPC_CMD_SET_LINK_TOL:
2742 if ((new_value >= TIPC_MIN_LINK_TOL) &&
2743 (new_value <= TIPC_MAX_LINK_TOL)) {
2744 link_set_supervision_props(l_ptr, new_value);
2745 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2746 0, 0, new_value, 0, 0);
2747 res = 0;
2748 }
2749 break;
2750 case TIPC_CMD_SET_LINK_PRI:
2751 if ((new_value >= TIPC_MIN_LINK_PRI) &&
2752 (new_value <= TIPC_MAX_LINK_PRI)) {
2753 l_ptr->priority = new_value;
2754 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2755 0, 0, 0, new_value, 0);
2756 res = 0;
2757 }
2758 break;
2759 case TIPC_CMD_SET_LINK_WINDOW:
2760 if ((new_value >= TIPC_MIN_LINK_WIN) &&
2761 (new_value <= TIPC_MAX_LINK_WIN)) {
2762 tipc_link_set_queue_limits(l_ptr, new_value);
2763 res = 0;
2764 }
2765 break;
2766 }
2767 tipc_node_unlock(node);
2768
2769 read_unlock_bh(&tipc_net_lock);
2770 if (res)
2771 return tipc_cfg_reply_error_string("cannot change link setting");
2772
2773 return tipc_cfg_reply_none();
2774}
2775
2776/**
2777 * link_reset_statistics - reset link statistics
2778 * @l_ptr: pointer to link
2779 */
2780
2781static void link_reset_statistics(struct link *l_ptr)
2782{
2783 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2784 l_ptr->stats.sent_info = l_ptr->next_out_no;
2785 l_ptr->stats.recv_info = l_ptr->next_in_no;
2786}
2787
2788struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2789{
2790 char *link_name;
2791 struct link *l_ptr;
2792 struct tipc_node *node;
2793
2794 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2795 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2796
2797 link_name = (char *)TLV_DATA(req_tlv_area);
2798 if (!strcmp(link_name, tipc_bclink_name)) {
2799 if (tipc_bclink_reset_stats())
2800 return tipc_cfg_reply_error_string("link not found");
2801 return tipc_cfg_reply_none();
2802 }
2803
2804 read_lock_bh(&tipc_net_lock);
2805 l_ptr = link_find_link(link_name, &node);
2806 if (!l_ptr) {
2807 read_unlock_bh(&tipc_net_lock);
2808 return tipc_cfg_reply_error_string("link not found");
2809 }
2810
2811 tipc_node_lock(node);
2812 link_reset_statistics(l_ptr);
2813 tipc_node_unlock(node);
2814 read_unlock_bh(&tipc_net_lock);
2815 return tipc_cfg_reply_none();
2816}
2817
2818/**
2819 * percent - convert count to a percentage of total (rounding up or down)
2820 */
2821
2822static u32 percent(u32 count, u32 total)
2823{
2824 return (count * 100 + (total / 2)) / total;
2825}
2826
2827/**
2828 * tipc_link_stats - print link statistics
2829 * @name: link name
2830 * @buf: print buffer area
2831 * @buf_size: size of print buffer area
2832 *
2833 * Returns length of print buffer data string (or 0 if error)
2834 */
2835
2836static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
2837{
2838 struct print_buf pb;
2839 struct link *l_ptr;
2840 struct tipc_node *node;
2841 char *status;
2842 u32 profile_total = 0;
2843
2844 if (!strcmp(name, tipc_bclink_name))
2845 return tipc_bclink_stats(buf, buf_size);
2846
2847 tipc_printbuf_init(&pb, buf, buf_size);
2848
2849 read_lock_bh(&tipc_net_lock);
2850 l_ptr = link_find_link(name, &node);
2851 if (!l_ptr) {
2852 read_unlock_bh(&tipc_net_lock);
2853 return 0;
2854 }
2855 tipc_node_lock(node);
2856
2857 if (tipc_link_is_active(l_ptr))
2858 status = "ACTIVE";
2859 else if (tipc_link_is_up(l_ptr))
2860 status = "STANDBY";
2861 else
2862 status = "DEFUNCT";
2863 tipc_printf(&pb, "Link <%s>\n"
2864 " %s MTU:%u Priority:%u Tolerance:%u ms"
2865 " Window:%u packets\n",
2866 l_ptr->name, status, l_ptr->max_pkt,
2867 l_ptr->priority, l_ptr->tolerance, l_ptr->queue_limit[0]);
2868 tipc_printf(&pb, " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
2869 l_ptr->next_in_no - l_ptr->stats.recv_info,
2870 l_ptr->stats.recv_fragments,
2871 l_ptr->stats.recv_fragmented,
2872 l_ptr->stats.recv_bundles,
2873 l_ptr->stats.recv_bundled);
2874 tipc_printf(&pb, " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
2875 l_ptr->next_out_no - l_ptr->stats.sent_info,
2876 l_ptr->stats.sent_fragments,
2877 l_ptr->stats.sent_fragmented,
2878 l_ptr->stats.sent_bundles,
2879 l_ptr->stats.sent_bundled);
2880 profile_total = l_ptr->stats.msg_length_counts;
2881 if (!profile_total)
2882 profile_total = 1;
2883 tipc_printf(&pb, " TX profile sample:%u packets average:%u octets\n"
2884 " 0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
2885 "-16384:%u%% -32768:%u%% -66000:%u%%\n",
2886 l_ptr->stats.msg_length_counts,
2887 l_ptr->stats.msg_lengths_total / profile_total,
2888 percent(l_ptr->stats.msg_length_profile[0], profile_total),
2889 percent(l_ptr->stats.msg_length_profile[1], profile_total),
2890 percent(l_ptr->stats.msg_length_profile[2], profile_total),
2891 percent(l_ptr->stats.msg_length_profile[3], profile_total),
2892 percent(l_ptr->stats.msg_length_profile[4], profile_total),
2893 percent(l_ptr->stats.msg_length_profile[5], profile_total),
2894 percent(l_ptr->stats.msg_length_profile[6], profile_total));
2895 tipc_printf(&pb, " RX states:%u probes:%u naks:%u defs:%u dups:%u\n",
2896 l_ptr->stats.recv_states,
2897 l_ptr->stats.recv_probes,
2898 l_ptr->stats.recv_nacks,
2899 l_ptr->stats.deferred_recv,
2900 l_ptr->stats.duplicates);
2901 tipc_printf(&pb, " TX states:%u probes:%u naks:%u acks:%u dups:%u\n",
2902 l_ptr->stats.sent_states,
2903 l_ptr->stats.sent_probes,
2904 l_ptr->stats.sent_nacks,
2905 l_ptr->stats.sent_acks,
2906 l_ptr->stats.retransmitted);
2907 tipc_printf(&pb, " Congestion bearer:%u link:%u Send queue max:%u avg:%u\n",
2908 l_ptr->stats.bearer_congs,
2909 l_ptr->stats.link_congs,
2910 l_ptr->stats.max_queue_sz,
2911 l_ptr->stats.queue_sz_counts
2912 ? (l_ptr->stats.accu_queue_sz / l_ptr->stats.queue_sz_counts)
2913 : 0);
2914
2915 tipc_node_unlock(node);
2916 read_unlock_bh(&tipc_net_lock);
2917 return tipc_printbuf_validate(&pb);
2918}
2919
2920#define MAX_LINK_STATS_INFO 2000
2921
2922struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
2923{
2924 struct sk_buff *buf;
2925 struct tlv_desc *rep_tlv;
2926 int str_len;
2927
2928 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2929 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2930
2931 buf = tipc_cfg_reply_alloc(TLV_SPACE(MAX_LINK_STATS_INFO));
2932 if (!buf)
2933 return NULL;
2934
2935 rep_tlv = (struct tlv_desc *)buf->data;
2936
2937 str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
2938 (char *)TLV_DATA(rep_tlv), MAX_LINK_STATS_INFO);
2939 if (!str_len) {
2940 buf_discard(buf);
2941 return tipc_cfg_reply_error_string("link not found");
2942 }
2943
2944 skb_put(buf, TLV_SPACE(str_len));
2945 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2946
2947 return buf;
2948}
2949
2950/**
2951 * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
2952 * @dest: network address of destination node
2953 * @selector: used to select from set of active links
2954 *
2955 * If no active link can be found, uses default maximum packet size.
2956 */
2957
2958u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
2959{
2960 struct tipc_node *n_ptr;
2961 struct link *l_ptr;
2962 u32 res = MAX_PKT_DEFAULT;
2963
2964 if (dest == tipc_own_addr)
2965 return MAX_MSG_SIZE;
2966
2967 read_lock_bh(&tipc_net_lock);
2968 n_ptr = tipc_node_find(dest);
2969 if (n_ptr) {
2970 tipc_node_lock(n_ptr);
2971 l_ptr = n_ptr->active_links[selector & 1];
2972 if (l_ptr)
2973 res = l_ptr->max_pkt;
2974 tipc_node_unlock(n_ptr);
2975 }
2976 read_unlock_bh(&tipc_net_lock);
2977 return res;
2978}
2979
2980static void link_print(struct link *l_ptr, const char *str)
2981{
2982 char print_area[256];
2983 struct print_buf pb;
2984 struct print_buf *buf = &pb;
2985
2986 tipc_printbuf_init(buf, print_area, sizeof(print_area));
2987
2988 tipc_printf(buf, str);
2989 tipc_printf(buf, "Link %x<%s>:",
2990 l_ptr->addr, l_ptr->b_ptr->name);
2991
2992#ifdef CONFIG_TIPC_DEBUG
2993 if (link_reset_reset(l_ptr) || link_reset_unknown(l_ptr))
2994 goto print_state;
2995
2996 tipc_printf(buf, ": NXO(%u):", mod(l_ptr->next_out_no));
2997 tipc_printf(buf, "NXI(%u):", mod(l_ptr->next_in_no));
2998 tipc_printf(buf, "SQUE");
2999 if (l_ptr->first_out) {
3000 tipc_printf(buf, "[%u..", msg_seqno(buf_msg(l_ptr->first_out)));
3001 if (l_ptr->next_out)
3002 tipc_printf(buf, "%u..",
3003 msg_seqno(buf_msg(l_ptr->next_out)));
3004 tipc_printf(buf, "%u]", msg_seqno(buf_msg(l_ptr->last_out)));
3005 if ((mod(msg_seqno(buf_msg(l_ptr->last_out)) -
3006 msg_seqno(buf_msg(l_ptr->first_out)))
3007 != (l_ptr->out_queue_size - 1)) ||
3008 (l_ptr->last_out->next != NULL)) {
3009 tipc_printf(buf, "\nSend queue inconsistency\n");
3010 tipc_printf(buf, "first_out= %p ", l_ptr->first_out);
3011 tipc_printf(buf, "next_out= %p ", l_ptr->next_out);
3012 tipc_printf(buf, "last_out= %p ", l_ptr->last_out);
3013 }
3014 } else
3015 tipc_printf(buf, "[]");
3016 tipc_printf(buf, "SQSIZ(%u)", l_ptr->out_queue_size);
3017 if (l_ptr->oldest_deferred_in) {
3018 u32 o = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
3019 u32 n = msg_seqno(buf_msg(l_ptr->newest_deferred_in));
3020 tipc_printf(buf, ":RQUE[%u..%u]", o, n);
3021 if (l_ptr->deferred_inqueue_sz != mod((n + 1) - o)) {
3022 tipc_printf(buf, ":RQSIZ(%u)",
3023 l_ptr->deferred_inqueue_sz);
3024 }
3025 }
3026print_state:
3027#endif
3028
3029 if (link_working_unknown(l_ptr))
3030 tipc_printf(buf, ":WU");
3031 else if (link_reset_reset(l_ptr))
3032 tipc_printf(buf, ":RR");
3033 else if (link_reset_unknown(l_ptr))
3034 tipc_printf(buf, ":RU");
3035 else if (link_working_working(l_ptr))
3036 tipc_printf(buf, ":WW");
3037 tipc_printf(buf, "\n");
3038
3039 tipc_printbuf_validate(buf);
3040 info("%s", print_area);
3041}
3042
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#include "crypto.h"
48
49#include <linux/pkt_sched.h>
50
51struct tipc_stats {
52 u32 sent_pkts;
53 u32 recv_pkts;
54 u32 sent_states;
55 u32 recv_states;
56 u32 sent_probes;
57 u32 recv_probes;
58 u32 sent_nacks;
59 u32 recv_nacks;
60 u32 sent_acks;
61 u32 sent_bundled;
62 u32 sent_bundles;
63 u32 recv_bundled;
64 u32 recv_bundles;
65 u32 retransmitted;
66 u32 sent_fragmented;
67 u32 sent_fragments;
68 u32 recv_fragmented;
69 u32 recv_fragments;
70 u32 link_congs; /* # port sends blocked by congestion */
71 u32 deferred_recv;
72 u32 duplicates;
73 u32 max_queue_sz; /* send queue size high water mark */
74 u32 accu_queue_sz; /* used for send queue size profiling */
75 u32 queue_sz_counts; /* used for send queue size profiling */
76 u32 msg_length_counts; /* used for message length profiling */
77 u32 msg_lengths_total; /* used for message length profiling */
78 u32 msg_length_profile[7]; /* used for msg. length profiling */
79};
80
81/**
82 * struct tipc_link - TIPC link data structure
83 * @addr: network address of link's peer node
84 * @name: link name character string
85 * @net: pointer to namespace struct
86 * @peer_session: link session # being used by peer end of link
87 * @peer_bearer_id: bearer id used by link's peer endpoint
88 * @bearer_id: local bearer id used by link
89 * @tolerance: minimum link continuity loss needed to reset link [in ms]
90 * @abort_limit: # of unacknowledged continuity probes needed to reset link
91 * @state: current state of link FSM
92 * @peer_caps: bitmap describing capabilities of peer node
93 * @silent_intv_cnt: # of timer intervals without any reception from peer
94 * @priority: current link priority
95 * @net_plane: current link network plane ('A' through 'H')
96 * @mon_state: cookie with information needed by link monitor
97 * @mtu: current maximum packet size for this link
98 * @advertised_mtu: advertised own mtu when link is being established
99 * @backlogq: queue for messages waiting to be sent
100 * @ackers: # of peers that needs to ack each packet before it can be released
101 * @acked: # last packet acked by a certain peer. Used for broadcast.
102 * @rcv_nxt: next sequence number to expect for inbound messages
103 * @inputq: buffer queue for messages to be delivered upwards
104 * @namedq: buffer queue for name table messages to be delivered upwards
105 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
106 * @reasm_buf: head of partially reassembled inbound message fragments
107 * @stats: collects statistics regarding link activity
108 * @session: session to be used by link
109 * @snd_nxt_state: next send seq number
110 * @rcv_nxt_state: next rcv seq number
111 * @in_session: have received ACTIVATE_MSG from peer
112 * @active: link is active
113 * @if_name: associated interface name
114 * @rst_cnt: link reset counter
115 * @drop_point: seq number for failover handling (FIXME)
116 * @failover_reasm_skb: saved failover msg ptr (FIXME)
117 * @failover_deferdq: deferred message queue for failover processing (FIXME)
118 * @transmq: the link's transmit queue
119 * @backlog: link's backlog by priority (importance)
120 * @snd_nxt: next sequence number to be used
121 * @rcv_unacked: # messages read by user, but not yet acked back to peer
122 * @deferdq: deferred receive queue
123 * @window: sliding window size for congestion handling
124 * @min_win: minimal send window to be used by link
125 * @ssthresh: slow start threshold for congestion handling
126 * @max_win: maximal send window to be used by link
127 * @cong_acks: congestion acks for congestion avoidance (FIXME)
128 * @checkpoint: seq number for congestion window size handling
129 * @reasm_tnlmsg: fragmentation/reassembly area for tunnel protocol message
130 * @last_gap: last gap ack blocks for bcast (FIXME)
131 * @last_ga: ptr to gap ack blocks
132 * @bc_rcvlink: the peer specific link used for broadcast reception
133 * @bc_sndlink: the namespace global link used for broadcast sending
134 * @nack_state: bcast nack state
135 * @bc_peer_is_up: peer has acked the bcast init msg
136 */
137struct tipc_link {
138 u32 addr;
139 char name[TIPC_MAX_LINK_NAME];
140 struct net *net;
141
142 /* Management and link supervision data */
143 u16 peer_session;
144 u16 session;
145 u16 snd_nxt_state;
146 u16 rcv_nxt_state;
147 u32 peer_bearer_id;
148 u32 bearer_id;
149 u32 tolerance;
150 u32 abort_limit;
151 u32 state;
152 u16 peer_caps;
153 bool in_session;
154 bool active;
155 u32 silent_intv_cnt;
156 char if_name[TIPC_MAX_IF_NAME];
157 u32 priority;
158 char net_plane;
159 struct tipc_mon_state mon_state;
160 u16 rst_cnt;
161
162 /* Failover/synch */
163 u16 drop_point;
164 struct sk_buff *failover_reasm_skb;
165 struct sk_buff_head failover_deferdq;
166
167 /* Max packet negotiation */
168 u16 mtu;
169 u16 advertised_mtu;
170
171 /* Sending */
172 struct sk_buff_head transmq;
173 struct sk_buff_head backlogq;
174 struct {
175 u16 len;
176 u16 limit;
177 struct sk_buff *target_bskb;
178 } backlog[5];
179 u16 snd_nxt;
180
181 /* Reception */
182 u16 rcv_nxt;
183 u32 rcv_unacked;
184 struct sk_buff_head deferdq;
185 struct sk_buff_head *inputq;
186 struct sk_buff_head *namedq;
187
188 /* Congestion handling */
189 struct sk_buff_head wakeupq;
190 u16 window;
191 u16 min_win;
192 u16 ssthresh;
193 u16 max_win;
194 u16 cong_acks;
195 u16 checkpoint;
196
197 /* Fragmentation/reassembly */
198 struct sk_buff *reasm_buf;
199 struct sk_buff *reasm_tnlmsg;
200
201 /* Broadcast */
202 u16 ackers;
203 u16 acked;
204 u16 last_gap;
205 struct tipc_gap_ack_blks *last_ga;
206 struct tipc_link *bc_rcvlink;
207 struct tipc_link *bc_sndlink;
208 u8 nack_state;
209 bool bc_peer_is_up;
210
211 /* Statistics */
212 struct tipc_stats stats;
213};
214
215/*
216 * Error message prefixes
217 */
218static const char *link_co_err = "Link tunneling error, ";
219static const char *link_rst_msg = "Resetting link ";
220
221/* Send states for broadcast NACKs
222 */
223enum {
224 BC_NACK_SND_CONDITIONAL,
225 BC_NACK_SND_UNCONDITIONAL,
226 BC_NACK_SND_SUPPRESS,
227};
228
229#define TIPC_BC_RETR_LIM (jiffies + msecs_to_jiffies(10))
230#define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
231
232/* Link FSM states:
233 */
234enum {
235 LINK_ESTABLISHED = 0xe,
236 LINK_ESTABLISHING = 0xe << 4,
237 LINK_RESET = 0x1 << 8,
238 LINK_RESETTING = 0x2 << 12,
239 LINK_PEER_RESET = 0xd << 16,
240 LINK_FAILINGOVER = 0xf << 20,
241 LINK_SYNCHING = 0xc << 24
242};
243
244static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
245 struct sk_buff_head *xmitq);
246static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
247 bool probe_reply, u16 rcvgap,
248 int tolerance, int priority,
249 struct sk_buff_head *xmitq);
250static void link_print(struct tipc_link *l, const char *str);
251static int tipc_link_build_nack_msg(struct tipc_link *l,
252 struct sk_buff_head *xmitq);
253static void tipc_link_build_bc_init_msg(struct tipc_link *l,
254 struct sk_buff_head *xmitq);
255static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
256 struct tipc_link *l, u8 start_index);
257static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr);
258static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r,
259 u16 acked, u16 gap,
260 struct tipc_gap_ack_blks *ga,
261 struct sk_buff_head *xmitq,
262 bool *retransmitted, int *rc);
263static void tipc_link_update_cwin(struct tipc_link *l, int released,
264 bool retransmitted);
265/*
266 * Simple non-static link routines (i.e. referenced outside this file)
267 */
268bool tipc_link_is_up(struct tipc_link *l)
269{
270 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
271}
272
273bool tipc_link_peer_is_down(struct tipc_link *l)
274{
275 return l->state == LINK_PEER_RESET;
276}
277
278bool tipc_link_is_reset(struct tipc_link *l)
279{
280 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
281}
282
283bool tipc_link_is_establishing(struct tipc_link *l)
284{
285 return l->state == LINK_ESTABLISHING;
286}
287
288bool tipc_link_is_synching(struct tipc_link *l)
289{
290 return l->state == LINK_SYNCHING;
291}
292
293bool tipc_link_is_failingover(struct tipc_link *l)
294{
295 return l->state == LINK_FAILINGOVER;
296}
297
298bool tipc_link_is_blocked(struct tipc_link *l)
299{
300 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
301}
302
303static bool link_is_bc_sndlink(struct tipc_link *l)
304{
305 return !l->bc_sndlink;
306}
307
308static bool link_is_bc_rcvlink(struct tipc_link *l)
309{
310 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
311}
312
313void tipc_link_set_active(struct tipc_link *l, bool active)
314{
315 l->active = active;
316}
317
318u32 tipc_link_id(struct tipc_link *l)
319{
320 return l->peer_bearer_id << 16 | l->bearer_id;
321}
322
323int tipc_link_min_win(struct tipc_link *l)
324{
325 return l->min_win;
326}
327
328int tipc_link_max_win(struct tipc_link *l)
329{
330 return l->max_win;
331}
332
333int tipc_link_prio(struct tipc_link *l)
334{
335 return l->priority;
336}
337
338unsigned long tipc_link_tolerance(struct tipc_link *l)
339{
340 return l->tolerance;
341}
342
343struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
344{
345 return l->inputq;
346}
347
348char tipc_link_plane(struct tipc_link *l)
349{
350 return l->net_plane;
351}
352
353struct net *tipc_link_net(struct tipc_link *l)
354{
355 return l->net;
356}
357
358void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
359{
360 l->peer_caps = capabilities;
361}
362
363void tipc_link_add_bc_peer(struct tipc_link *snd_l,
364 struct tipc_link *uc_l,
365 struct sk_buff_head *xmitq)
366{
367 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
368
369 snd_l->ackers++;
370 rcv_l->acked = snd_l->snd_nxt - 1;
371 snd_l->state = LINK_ESTABLISHED;
372 tipc_link_build_bc_init_msg(uc_l, xmitq);
373}
374
375void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
376 struct tipc_link *rcv_l,
377 struct sk_buff_head *xmitq)
378{
379 u16 ack = snd_l->snd_nxt - 1;
380
381 snd_l->ackers--;
382 rcv_l->bc_peer_is_up = true;
383 rcv_l->state = LINK_ESTABLISHED;
384 tipc_link_bc_ack_rcv(rcv_l, ack, 0, NULL, xmitq, NULL);
385 trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
386 tipc_link_reset(rcv_l);
387 rcv_l->state = LINK_RESET;
388 if (!snd_l->ackers) {
389 trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
390 tipc_link_reset(snd_l);
391 snd_l->state = LINK_RESET;
392 __skb_queue_purge(xmitq);
393 }
394}
395
396int tipc_link_bc_peers(struct tipc_link *l)
397{
398 return l->ackers;
399}
400
401static u16 link_bc_rcv_gap(struct tipc_link *l)
402{
403 struct sk_buff *skb = skb_peek(&l->deferdq);
404 u16 gap = 0;
405
406 if (more(l->snd_nxt, l->rcv_nxt))
407 gap = l->snd_nxt - l->rcv_nxt;
408 if (skb)
409 gap = buf_seqno(skb) - l->rcv_nxt;
410 return gap;
411}
412
413void tipc_link_set_mtu(struct tipc_link *l, int mtu)
414{
415 l->mtu = mtu;
416}
417
418int tipc_link_mtu(struct tipc_link *l)
419{
420 return l->mtu;
421}
422
423int tipc_link_mss(struct tipc_link *l)
424{
425#ifdef CONFIG_TIPC_CRYPTO
426 return l->mtu - INT_H_SIZE - EMSG_OVERHEAD;
427#else
428 return l->mtu - INT_H_SIZE;
429#endif
430}
431
432u16 tipc_link_rcv_nxt(struct tipc_link *l)
433{
434 return l->rcv_nxt;
435}
436
437u16 tipc_link_acked(struct tipc_link *l)
438{
439 return l->acked;
440}
441
442char *tipc_link_name(struct tipc_link *l)
443{
444 return l->name;
445}
446
447u32 tipc_link_state(struct tipc_link *l)
448{
449 return l->state;
450}
451
452/**
453 * tipc_link_create - create a new link
454 * @net: pointer to associated network namespace
455 * @if_name: associated interface name
456 * @bearer_id: id (index) of associated bearer
457 * @tolerance: link tolerance to be used by link
458 * @net_plane: network plane (A,B,c..) this link belongs to
459 * @mtu: mtu to be advertised by link
460 * @priority: priority to be used by link
461 * @min_win: minimal send window to be used by link
462 * @max_win: maximal send window to be used by link
463 * @session: session to be used by link
464 * @peer: node id of peer node
465 * @peer_caps: bitmap describing peer node capabilities
466 * @bc_sndlink: the namespace global link used for broadcast sending
467 * @bc_rcvlink: the peer specific link used for broadcast reception
468 * @inputq: queue to put messages ready for delivery
469 * @namedq: queue to put binding table update messages ready for delivery
470 * @link: return value, pointer to put the created link
471 * @self: local unicast link id
472 * @peer_id: 128-bit ID of peer
473 *
474 * Return: true if link was created, otherwise false
475 */
476bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
477 int tolerance, char net_plane, u32 mtu, int priority,
478 u32 min_win, u32 max_win, u32 session, u32 self,
479 u32 peer, u8 *peer_id, u16 peer_caps,
480 struct tipc_link *bc_sndlink,
481 struct tipc_link *bc_rcvlink,
482 struct sk_buff_head *inputq,
483 struct sk_buff_head *namedq,
484 struct tipc_link **link)
485{
486 char peer_str[NODE_ID_STR_LEN] = {0,};
487 char self_str[NODE_ID_STR_LEN] = {0,};
488 struct tipc_link *l;
489
490 l = kzalloc(sizeof(*l), GFP_ATOMIC);
491 if (!l)
492 return false;
493 *link = l;
494 l->session = session;
495
496 /* Set link name for unicast links only */
497 if (peer_id) {
498 tipc_nodeid2string(self_str, tipc_own_id(net));
499 if (strlen(self_str) > 16)
500 sprintf(self_str, "%x", self);
501 tipc_nodeid2string(peer_str, peer_id);
502 if (strlen(peer_str) > 16)
503 sprintf(peer_str, "%x", peer);
504 }
505 /* Peer i/f name will be completed by reset/activate message */
506 snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
507 self_str, if_name, peer_str);
508
509 strcpy(l->if_name, if_name);
510 l->addr = peer;
511 l->peer_caps = peer_caps;
512 l->net = net;
513 l->in_session = false;
514 l->bearer_id = bearer_id;
515 l->tolerance = tolerance;
516 if (bc_rcvlink)
517 bc_rcvlink->tolerance = tolerance;
518 l->net_plane = net_plane;
519 l->advertised_mtu = mtu;
520 l->mtu = mtu;
521 l->priority = priority;
522 tipc_link_set_queue_limits(l, min_win, max_win);
523 l->ackers = 1;
524 l->bc_sndlink = bc_sndlink;
525 l->bc_rcvlink = bc_rcvlink;
526 l->inputq = inputq;
527 l->namedq = namedq;
528 l->state = LINK_RESETTING;
529 __skb_queue_head_init(&l->transmq);
530 __skb_queue_head_init(&l->backlogq);
531 __skb_queue_head_init(&l->deferdq);
532 __skb_queue_head_init(&l->failover_deferdq);
533 skb_queue_head_init(&l->wakeupq);
534 skb_queue_head_init(l->inputq);
535 return true;
536}
537
538/**
539 * tipc_link_bc_create - create new link to be used for broadcast
540 * @net: pointer to associated network namespace
541 * @mtu: mtu to be used initially if no peers
542 * @min_win: minimal send window to be used by link
543 * @max_win: maximal send window to be used by link
544 * @inputq: queue to put messages ready for delivery
545 * @namedq: queue to put binding table update messages ready for delivery
546 * @link: return value, pointer to put the created link
547 * @ownnode: identity of own node
548 * @peer: node id of peer node
549 * @peer_id: 128-bit ID of peer
550 * @peer_caps: bitmap describing peer node capabilities
551 * @bc_sndlink: the namespace global link used for broadcast sending
552 *
553 * Return: true if link was created, otherwise false
554 */
555bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer, u8 *peer_id,
556 int mtu, u32 min_win, u32 max_win, u16 peer_caps,
557 struct sk_buff_head *inputq,
558 struct sk_buff_head *namedq,
559 struct tipc_link *bc_sndlink,
560 struct tipc_link **link)
561{
562 struct tipc_link *l;
563
564 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, min_win,
565 max_win, 0, ownnode, peer, NULL, peer_caps,
566 bc_sndlink, NULL, inputq, namedq, link))
567 return false;
568
569 l = *link;
570 if (peer_id) {
571 char peer_str[NODE_ID_STR_LEN] = {0,};
572
573 tipc_nodeid2string(peer_str, peer_id);
574 if (strlen(peer_str) > 16)
575 sprintf(peer_str, "%x", peer);
576 /* Broadcast receiver link name: "broadcast-link:<peer>" */
577 snprintf(l->name, sizeof(l->name), "%s:%s", tipc_bclink_name,
578 peer_str);
579 } else {
580 strcpy(l->name, tipc_bclink_name);
581 }
582 trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
583 tipc_link_reset(l);
584 l->state = LINK_RESET;
585 l->ackers = 0;
586 l->bc_rcvlink = l;
587
588 /* Broadcast send link is always up */
589 if (link_is_bc_sndlink(l))
590 l->state = LINK_ESTABLISHED;
591
592 /* Disable replicast if even a single peer doesn't support it */
593 if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
594 tipc_bcast_toggle_rcast(net, false);
595
596 return true;
597}
598
599/**
600 * tipc_link_fsm_evt - link finite state machine
601 * @l: pointer to link
602 * @evt: state machine event to be processed
603 */
604int tipc_link_fsm_evt(struct tipc_link *l, int evt)
605{
606 int rc = 0;
607 int old_state = l->state;
608
609 switch (l->state) {
610 case LINK_RESETTING:
611 switch (evt) {
612 case LINK_PEER_RESET_EVT:
613 l->state = LINK_PEER_RESET;
614 break;
615 case LINK_RESET_EVT:
616 l->state = LINK_RESET;
617 break;
618 case LINK_FAILURE_EVT:
619 case LINK_FAILOVER_BEGIN_EVT:
620 case LINK_ESTABLISH_EVT:
621 case LINK_FAILOVER_END_EVT:
622 case LINK_SYNCH_BEGIN_EVT:
623 case LINK_SYNCH_END_EVT:
624 default:
625 goto illegal_evt;
626 }
627 break;
628 case LINK_RESET:
629 switch (evt) {
630 case LINK_PEER_RESET_EVT:
631 l->state = LINK_ESTABLISHING;
632 break;
633 case LINK_FAILOVER_BEGIN_EVT:
634 l->state = LINK_FAILINGOVER;
635 break;
636 case LINK_FAILURE_EVT:
637 case LINK_RESET_EVT:
638 case LINK_ESTABLISH_EVT:
639 case LINK_FAILOVER_END_EVT:
640 break;
641 case LINK_SYNCH_BEGIN_EVT:
642 case LINK_SYNCH_END_EVT:
643 default:
644 goto illegal_evt;
645 }
646 break;
647 case LINK_PEER_RESET:
648 switch (evt) {
649 case LINK_RESET_EVT:
650 l->state = LINK_ESTABLISHING;
651 break;
652 case LINK_PEER_RESET_EVT:
653 case LINK_ESTABLISH_EVT:
654 case LINK_FAILURE_EVT:
655 break;
656 case LINK_SYNCH_BEGIN_EVT:
657 case LINK_SYNCH_END_EVT:
658 case LINK_FAILOVER_BEGIN_EVT:
659 case LINK_FAILOVER_END_EVT:
660 default:
661 goto illegal_evt;
662 }
663 break;
664 case LINK_FAILINGOVER:
665 switch (evt) {
666 case LINK_FAILOVER_END_EVT:
667 l->state = LINK_RESET;
668 break;
669 case LINK_PEER_RESET_EVT:
670 case LINK_RESET_EVT:
671 case LINK_ESTABLISH_EVT:
672 case LINK_FAILURE_EVT:
673 break;
674 case LINK_FAILOVER_BEGIN_EVT:
675 case LINK_SYNCH_BEGIN_EVT:
676 case LINK_SYNCH_END_EVT:
677 default:
678 goto illegal_evt;
679 }
680 break;
681 case LINK_ESTABLISHING:
682 switch (evt) {
683 case LINK_ESTABLISH_EVT:
684 l->state = LINK_ESTABLISHED;
685 break;
686 case LINK_FAILOVER_BEGIN_EVT:
687 l->state = LINK_FAILINGOVER;
688 break;
689 case LINK_RESET_EVT:
690 l->state = LINK_RESET;
691 break;
692 case LINK_FAILURE_EVT:
693 case LINK_PEER_RESET_EVT:
694 case LINK_SYNCH_BEGIN_EVT:
695 case LINK_FAILOVER_END_EVT:
696 break;
697 case LINK_SYNCH_END_EVT:
698 default:
699 goto illegal_evt;
700 }
701 break;
702 case LINK_ESTABLISHED:
703 switch (evt) {
704 case LINK_PEER_RESET_EVT:
705 l->state = LINK_PEER_RESET;
706 rc |= TIPC_LINK_DOWN_EVT;
707 break;
708 case LINK_FAILURE_EVT:
709 l->state = LINK_RESETTING;
710 rc |= TIPC_LINK_DOWN_EVT;
711 break;
712 case LINK_RESET_EVT:
713 l->state = LINK_RESET;
714 break;
715 case LINK_ESTABLISH_EVT:
716 case LINK_SYNCH_END_EVT:
717 break;
718 case LINK_SYNCH_BEGIN_EVT:
719 l->state = LINK_SYNCHING;
720 break;
721 case LINK_FAILOVER_BEGIN_EVT:
722 case LINK_FAILOVER_END_EVT:
723 default:
724 goto illegal_evt;
725 }
726 break;
727 case LINK_SYNCHING:
728 switch (evt) {
729 case LINK_PEER_RESET_EVT:
730 l->state = LINK_PEER_RESET;
731 rc |= TIPC_LINK_DOWN_EVT;
732 break;
733 case LINK_FAILURE_EVT:
734 l->state = LINK_RESETTING;
735 rc |= TIPC_LINK_DOWN_EVT;
736 break;
737 case LINK_RESET_EVT:
738 l->state = LINK_RESET;
739 break;
740 case LINK_ESTABLISH_EVT:
741 case LINK_SYNCH_BEGIN_EVT:
742 break;
743 case LINK_SYNCH_END_EVT:
744 l->state = LINK_ESTABLISHED;
745 break;
746 case LINK_FAILOVER_BEGIN_EVT:
747 case LINK_FAILOVER_END_EVT:
748 default:
749 goto illegal_evt;
750 }
751 break;
752 default:
753 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
754 }
755 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
756 return rc;
757illegal_evt:
758 pr_err("Illegal FSM event %x in state %x on link %s\n",
759 evt, l->state, l->name);
760 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
761 return rc;
762}
763
764/* link_profile_stats - update statistical profiling of traffic
765 */
766static void link_profile_stats(struct tipc_link *l)
767{
768 struct sk_buff *skb;
769 struct tipc_msg *msg;
770 int length;
771
772 /* Update counters used in statistical profiling of send traffic */
773 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
774 l->stats.queue_sz_counts++;
775
776 skb = skb_peek(&l->transmq);
777 if (!skb)
778 return;
779 msg = buf_msg(skb);
780 length = msg_size(msg);
781
782 if (msg_user(msg) == MSG_FRAGMENTER) {
783 if (msg_type(msg) != FIRST_FRAGMENT)
784 return;
785 length = msg_size(msg_inner_hdr(msg));
786 }
787 l->stats.msg_lengths_total += length;
788 l->stats.msg_length_counts++;
789 if (length <= 64)
790 l->stats.msg_length_profile[0]++;
791 else if (length <= 256)
792 l->stats.msg_length_profile[1]++;
793 else if (length <= 1024)
794 l->stats.msg_length_profile[2]++;
795 else if (length <= 4096)
796 l->stats.msg_length_profile[3]++;
797 else if (length <= 16384)
798 l->stats.msg_length_profile[4]++;
799 else if (length <= 32768)
800 l->stats.msg_length_profile[5]++;
801 else
802 l->stats.msg_length_profile[6]++;
803}
804
805/**
806 * tipc_link_too_silent - check if link is "too silent"
807 * @l: tipc link to be checked
808 *
809 * Return: true if the link 'silent_intv_cnt' is about to reach the
810 * 'abort_limit' value, otherwise false
811 */
812bool tipc_link_too_silent(struct tipc_link *l)
813{
814 return (l->silent_intv_cnt + 2 > l->abort_limit);
815}
816
817/* tipc_link_timeout - perform periodic task as instructed from node timeout
818 */
819int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
820{
821 int mtyp = 0;
822 int rc = 0;
823 bool state = false;
824 bool probe = false;
825 bool setup = false;
826 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
827 u16 bc_acked = l->bc_rcvlink->acked;
828 struct tipc_mon_state *mstate = &l->mon_state;
829
830 trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
831 trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
832 switch (l->state) {
833 case LINK_ESTABLISHED:
834 case LINK_SYNCHING:
835 mtyp = STATE_MSG;
836 link_profile_stats(l);
837 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
838 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
839 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
840 state = bc_acked != bc_snt;
841 state |= l->bc_rcvlink->rcv_unacked;
842 state |= l->rcv_unacked;
843 state |= !skb_queue_empty(&l->transmq);
844 probe = mstate->probing;
845 probe |= l->silent_intv_cnt;
846 if (probe || mstate->monitoring)
847 l->silent_intv_cnt++;
848 probe |= !skb_queue_empty(&l->deferdq);
849 if (l->snd_nxt == l->checkpoint) {
850 tipc_link_update_cwin(l, 0, 0);
851 probe = true;
852 }
853 l->checkpoint = l->snd_nxt;
854 break;
855 case LINK_RESET:
856 setup = l->rst_cnt++ <= 4;
857 setup |= !(l->rst_cnt % 16);
858 mtyp = RESET_MSG;
859 break;
860 case LINK_ESTABLISHING:
861 setup = true;
862 mtyp = ACTIVATE_MSG;
863 break;
864 case LINK_PEER_RESET:
865 case LINK_RESETTING:
866 case LINK_FAILINGOVER:
867 break;
868 default:
869 break;
870 }
871
872 if (state || probe || setup)
873 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
874
875 return rc;
876}
877
878/**
879 * link_schedule_user - schedule a message sender for wakeup after congestion
880 * @l: congested link
881 * @hdr: header of message that is being sent
882 * Create pseudo msg to send back to user when congestion abates
883 */
884static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
885{
886 u32 dnode = tipc_own_addr(l->net);
887 u32 dport = msg_origport(hdr);
888 struct sk_buff *skb;
889
890 /* Create and schedule wakeup pseudo message */
891 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
892 dnode, l->addr, dport, 0, 0);
893 if (!skb)
894 return -ENOBUFS;
895 msg_set_dest_droppable(buf_msg(skb), true);
896 TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
897 skb_queue_tail(&l->wakeupq, skb);
898 l->stats.link_congs++;
899 trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
900 return -ELINKCONG;
901}
902
903/**
904 * link_prepare_wakeup - prepare users for wakeup after congestion
905 * @l: congested link
906 * Wake up a number of waiting users, as permitted by available space
907 * in the send queue
908 */
909static void link_prepare_wakeup(struct tipc_link *l)
910{
911 struct sk_buff_head *wakeupq = &l->wakeupq;
912 struct sk_buff_head *inputq = l->inputq;
913 struct sk_buff *skb, *tmp;
914 struct sk_buff_head tmpq;
915 int avail[5] = {0,};
916 int imp = 0;
917
918 __skb_queue_head_init(&tmpq);
919
920 for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
921 avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
922
923 skb_queue_walk_safe(wakeupq, skb, tmp) {
924 imp = TIPC_SKB_CB(skb)->chain_imp;
925 if (avail[imp] <= 0)
926 continue;
927 avail[imp]--;
928 __skb_unlink(skb, wakeupq);
929 __skb_queue_tail(&tmpq, skb);
930 }
931
932 spin_lock_bh(&inputq->lock);
933 skb_queue_splice_tail(&tmpq, inputq);
934 spin_unlock_bh(&inputq->lock);
935
936}
937
938/**
939 * tipc_link_set_skb_retransmit_time - set the time at which retransmission of
940 * the given skb should be next attempted
941 * @skb: skb to set a future retransmission time for
942 * @l: link the skb will be transmitted on
943 */
944static void tipc_link_set_skb_retransmit_time(struct sk_buff *skb,
945 struct tipc_link *l)
946{
947 if (link_is_bc_sndlink(l))
948 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
949 else
950 TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
951}
952
953void tipc_link_reset(struct tipc_link *l)
954{
955 struct sk_buff_head list;
956 u32 imp;
957
958 __skb_queue_head_init(&list);
959
960 l->in_session = false;
961 /* Force re-synch of peer session number before establishing */
962 l->peer_session--;
963 l->session++;
964 l->mtu = l->advertised_mtu;
965
966 spin_lock_bh(&l->wakeupq.lock);
967 skb_queue_splice_init(&l->wakeupq, &list);
968 spin_unlock_bh(&l->wakeupq.lock);
969
970 spin_lock_bh(&l->inputq->lock);
971 skb_queue_splice_init(&list, l->inputq);
972 spin_unlock_bh(&l->inputq->lock);
973
974 __skb_queue_purge(&l->transmq);
975 __skb_queue_purge(&l->deferdq);
976 __skb_queue_purge(&l->backlogq);
977 __skb_queue_purge(&l->failover_deferdq);
978 for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
979 l->backlog[imp].len = 0;
980 l->backlog[imp].target_bskb = NULL;
981 }
982 kfree_skb(l->reasm_buf);
983 kfree_skb(l->reasm_tnlmsg);
984 kfree_skb(l->failover_reasm_skb);
985 l->reasm_buf = NULL;
986 l->reasm_tnlmsg = NULL;
987 l->failover_reasm_skb = NULL;
988 l->rcv_unacked = 0;
989 l->snd_nxt = 1;
990 l->rcv_nxt = 1;
991 l->snd_nxt_state = 1;
992 l->rcv_nxt_state = 1;
993 l->acked = 0;
994 l->last_gap = 0;
995 kfree(l->last_ga);
996 l->last_ga = NULL;
997 l->silent_intv_cnt = 0;
998 l->rst_cnt = 0;
999 l->bc_peer_is_up = false;
1000 memset(&l->mon_state, 0, sizeof(l->mon_state));
1001 tipc_link_reset_stats(l);
1002}
1003
1004/**
1005 * tipc_link_xmit(): enqueue buffer list according to queue situation
1006 * @l: link to use
1007 * @list: chain of buffers containing message
1008 * @xmitq: returned list of packets to be sent by caller
1009 *
1010 * Consumes the buffer chain.
1011 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
1012 * Return: 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
1013 */
1014int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
1015 struct sk_buff_head *xmitq)
1016{
1017 struct sk_buff_head *backlogq = &l->backlogq;
1018 struct sk_buff_head *transmq = &l->transmq;
1019 struct sk_buff *skb, *_skb;
1020 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1021 u16 ack = l->rcv_nxt - 1;
1022 u16 seqno = l->snd_nxt;
1023 int pkt_cnt = skb_queue_len(list);
1024 unsigned int mss = tipc_link_mss(l);
1025 unsigned int cwin = l->window;
1026 unsigned int mtu = l->mtu;
1027 struct tipc_msg *hdr;
1028 bool new_bundle;
1029 int rc = 0;
1030 int imp;
1031
1032 if (pkt_cnt <= 0)
1033 return 0;
1034
1035 hdr = buf_msg(skb_peek(list));
1036 if (unlikely(msg_size(hdr) > mtu)) {
1037 pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
1038 skb_queue_len(list), msg_user(hdr),
1039 msg_type(hdr), msg_size(hdr), mtu);
1040 __skb_queue_purge(list);
1041 return -EMSGSIZE;
1042 }
1043
1044 imp = msg_importance(hdr);
1045 /* Allow oversubscription of one data msg per source at congestion */
1046 if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
1047 if (imp == TIPC_SYSTEM_IMPORTANCE) {
1048 pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
1049 return -ENOBUFS;
1050 }
1051 rc = link_schedule_user(l, hdr);
1052 }
1053
1054 if (pkt_cnt > 1) {
1055 l->stats.sent_fragmented++;
1056 l->stats.sent_fragments += pkt_cnt;
1057 }
1058
1059 /* Prepare each packet for sending, and add to relevant queue: */
1060 while ((skb = __skb_dequeue(list))) {
1061 if (likely(skb_queue_len(transmq) < cwin)) {
1062 hdr = buf_msg(skb);
1063 msg_set_seqno(hdr, seqno);
1064 msg_set_ack(hdr, ack);
1065 msg_set_bcast_ack(hdr, bc_ack);
1066 _skb = skb_clone(skb, GFP_ATOMIC);
1067 if (!_skb) {
1068 kfree_skb(skb);
1069 __skb_queue_purge(list);
1070 return -ENOBUFS;
1071 }
1072 __skb_queue_tail(transmq, skb);
1073 tipc_link_set_skb_retransmit_time(skb, l);
1074 __skb_queue_tail(xmitq, _skb);
1075 TIPC_SKB_CB(skb)->ackers = l->ackers;
1076 l->rcv_unacked = 0;
1077 l->stats.sent_pkts++;
1078 seqno++;
1079 continue;
1080 }
1081 if (tipc_msg_try_bundle(l->backlog[imp].target_bskb, &skb,
1082 mss, l->addr, &new_bundle)) {
1083 if (skb) {
1084 /* Keep a ref. to the skb for next try */
1085 l->backlog[imp].target_bskb = skb;
1086 l->backlog[imp].len++;
1087 __skb_queue_tail(backlogq, skb);
1088 } else {
1089 if (new_bundle) {
1090 l->stats.sent_bundles++;
1091 l->stats.sent_bundled++;
1092 }
1093 l->stats.sent_bundled++;
1094 }
1095 continue;
1096 }
1097 l->backlog[imp].target_bskb = NULL;
1098 l->backlog[imp].len += (1 + skb_queue_len(list));
1099 __skb_queue_tail(backlogq, skb);
1100 skb_queue_splice_tail_init(list, backlogq);
1101 }
1102 l->snd_nxt = seqno;
1103 return rc;
1104}
1105
1106static void tipc_link_update_cwin(struct tipc_link *l, int released,
1107 bool retransmitted)
1108{
1109 int bklog_len = skb_queue_len(&l->backlogq);
1110 struct sk_buff_head *txq = &l->transmq;
1111 int txq_len = skb_queue_len(txq);
1112 u16 cwin = l->window;
1113
1114 /* Enter fast recovery */
1115 if (unlikely(retransmitted)) {
1116 l->ssthresh = max_t(u16, l->window / 2, 300);
1117 l->window = min_t(u16, l->ssthresh, l->window);
1118 return;
1119 }
1120 /* Enter slow start */
1121 if (unlikely(!released)) {
1122 l->ssthresh = max_t(u16, l->window / 2, 300);
1123 l->window = l->min_win;
1124 return;
1125 }
1126 /* Don't increase window if no pressure on the transmit queue */
1127 if (txq_len + bklog_len < cwin)
1128 return;
1129
1130 /* Don't increase window if there are holes the transmit queue */
1131 if (txq_len && l->snd_nxt - buf_seqno(skb_peek(txq)) != txq_len)
1132 return;
1133
1134 l->cong_acks += released;
1135
1136 /* Slow start */
1137 if (cwin <= l->ssthresh) {
1138 l->window = min_t(u16, cwin + released, l->max_win);
1139 return;
1140 }
1141 /* Congestion avoidance */
1142 if (l->cong_acks < cwin)
1143 return;
1144 l->window = min_t(u16, ++cwin, l->max_win);
1145 l->cong_acks = 0;
1146}
1147
1148static void tipc_link_advance_backlog(struct tipc_link *l,
1149 struct sk_buff_head *xmitq)
1150{
1151 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1152 struct sk_buff_head *txq = &l->transmq;
1153 struct sk_buff *skb, *_skb;
1154 u16 ack = l->rcv_nxt - 1;
1155 u16 seqno = l->snd_nxt;
1156 struct tipc_msg *hdr;
1157 u16 cwin = l->window;
1158 u32 imp;
1159
1160 while (skb_queue_len(txq) < cwin) {
1161 skb = skb_peek(&l->backlogq);
1162 if (!skb)
1163 break;
1164 _skb = skb_clone(skb, GFP_ATOMIC);
1165 if (!_skb)
1166 break;
1167 __skb_dequeue(&l->backlogq);
1168 hdr = buf_msg(skb);
1169 imp = msg_importance(hdr);
1170 l->backlog[imp].len--;
1171 if (unlikely(skb == l->backlog[imp].target_bskb))
1172 l->backlog[imp].target_bskb = NULL;
1173 __skb_queue_tail(&l->transmq, skb);
1174 tipc_link_set_skb_retransmit_time(skb, l);
1175
1176 __skb_queue_tail(xmitq, _skb);
1177 TIPC_SKB_CB(skb)->ackers = l->ackers;
1178 msg_set_seqno(hdr, seqno);
1179 msg_set_ack(hdr, ack);
1180 msg_set_bcast_ack(hdr, bc_ack);
1181 l->rcv_unacked = 0;
1182 l->stats.sent_pkts++;
1183 seqno++;
1184 }
1185 l->snd_nxt = seqno;
1186}
1187
1188/**
1189 * link_retransmit_failure() - Detect repeated retransmit failures
1190 * @l: tipc link sender
1191 * @r: tipc link receiver (= l in case of unicast)
1192 * @rc: returned code
1193 *
1194 * Return: true if the repeated retransmit failures happens, otherwise
1195 * false
1196 */
1197static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1198 int *rc)
1199{
1200 struct sk_buff *skb = skb_peek(&l->transmq);
1201 struct tipc_msg *hdr;
1202
1203 if (!skb)
1204 return false;
1205
1206 if (!TIPC_SKB_CB(skb)->retr_cnt)
1207 return false;
1208
1209 if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1210 msecs_to_jiffies(r->tolerance * 10)))
1211 return false;
1212
1213 hdr = buf_msg(skb);
1214 if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1215 return false;
1216
1217 pr_warn("Retransmission failure on link <%s>\n", l->name);
1218 link_print(l, "State of link ");
1219 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1220 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1221 pr_info("sqno %u, prev: %x, dest: %x\n",
1222 msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1223 pr_info("retr_stamp %d, retr_cnt %d\n",
1224 jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1225 TIPC_SKB_CB(skb)->retr_cnt);
1226
1227 trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1228 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1229 trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1230
1231 if (link_is_bc_sndlink(l)) {
1232 r->state = LINK_RESET;
1233 *rc |= TIPC_LINK_DOWN_EVT;
1234 } else {
1235 *rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1236 }
1237
1238 return true;
1239}
1240
1241/* tipc_data_input - deliver data and name distr msgs to upper layer
1242 *
1243 * Consumes buffer if message is of right type
1244 * Node lock must be held
1245 */
1246static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1247 struct sk_buff_head *inputq)
1248{
1249 struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1250 struct tipc_msg *hdr = buf_msg(skb);
1251
1252 switch (msg_user(hdr)) {
1253 case TIPC_LOW_IMPORTANCE:
1254 case TIPC_MEDIUM_IMPORTANCE:
1255 case TIPC_HIGH_IMPORTANCE:
1256 case TIPC_CRITICAL_IMPORTANCE:
1257 if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1258 skb_queue_tail(mc_inputq, skb);
1259 return true;
1260 }
1261 fallthrough;
1262 case CONN_MANAGER:
1263 skb_queue_tail(inputq, skb);
1264 return true;
1265 case GROUP_PROTOCOL:
1266 skb_queue_tail(mc_inputq, skb);
1267 return true;
1268 case NAME_DISTRIBUTOR:
1269 l->bc_rcvlink->state = LINK_ESTABLISHED;
1270 skb_queue_tail(l->namedq, skb);
1271 return true;
1272 case MSG_BUNDLER:
1273 case TUNNEL_PROTOCOL:
1274 case MSG_FRAGMENTER:
1275 case BCAST_PROTOCOL:
1276 return false;
1277#ifdef CONFIG_TIPC_CRYPTO
1278 case MSG_CRYPTO:
1279 if (sysctl_tipc_key_exchange_enabled &&
1280 TIPC_SKB_CB(skb)->decrypted) {
1281 tipc_crypto_msg_rcv(l->net, skb);
1282 return true;
1283 }
1284 fallthrough;
1285#endif
1286 default:
1287 pr_warn("Dropping received illegal msg type\n");
1288 kfree_skb(skb);
1289 return true;
1290 }
1291}
1292
1293/* tipc_link_input - process packet that has passed link protocol check
1294 *
1295 * Consumes buffer
1296 */
1297static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1298 struct sk_buff_head *inputq,
1299 struct sk_buff **reasm_skb)
1300{
1301 struct tipc_msg *hdr = buf_msg(skb);
1302 struct sk_buff *iskb;
1303 struct sk_buff_head tmpq;
1304 int usr = msg_user(hdr);
1305 int pos = 0;
1306
1307 if (usr == MSG_BUNDLER) {
1308 skb_queue_head_init(&tmpq);
1309 l->stats.recv_bundles++;
1310 l->stats.recv_bundled += msg_msgcnt(hdr);
1311 while (tipc_msg_extract(skb, &iskb, &pos))
1312 tipc_data_input(l, iskb, &tmpq);
1313 tipc_skb_queue_splice_tail(&tmpq, inputq);
1314 return 0;
1315 } else if (usr == MSG_FRAGMENTER) {
1316 l->stats.recv_fragments++;
1317 if (tipc_buf_append(reasm_skb, &skb)) {
1318 l->stats.recv_fragmented++;
1319 tipc_data_input(l, skb, inputq);
1320 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1321 pr_warn_ratelimited("Unable to build fragment list\n");
1322 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1323 }
1324 return 0;
1325 } else if (usr == BCAST_PROTOCOL) {
1326 tipc_bcast_lock(l->net);
1327 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1328 tipc_bcast_unlock(l->net);
1329 }
1330
1331 kfree_skb(skb);
1332 return 0;
1333}
1334
1335/* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1336 * inner message along with the ones in the old link's
1337 * deferdq
1338 * @l: tunnel link
1339 * @skb: TUNNEL_PROTOCOL message
1340 * @inputq: queue to put messages ready for delivery
1341 */
1342static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1343 struct sk_buff_head *inputq)
1344{
1345 struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1346 struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1347 struct sk_buff_head *fdefq = &l->failover_deferdq;
1348 struct tipc_msg *hdr = buf_msg(skb);
1349 struct sk_buff *iskb;
1350 int ipos = 0;
1351 int rc = 0;
1352 u16 seqno;
1353
1354 if (msg_type(hdr) == SYNCH_MSG) {
1355 kfree_skb(skb);
1356 return 0;
1357 }
1358
1359 /* Not a fragment? */
1360 if (likely(!msg_nof_fragms(hdr))) {
1361 if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1362 pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1363 skb_queue_len(fdefq));
1364 return 0;
1365 }
1366 kfree_skb(skb);
1367 } else {
1368 /* Set fragment type for buf_append */
1369 if (msg_fragm_no(hdr) == 1)
1370 msg_set_type(hdr, FIRST_FRAGMENT);
1371 else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1372 msg_set_type(hdr, FRAGMENT);
1373 else
1374 msg_set_type(hdr, LAST_FRAGMENT);
1375
1376 if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1377 /* Successful but non-complete reassembly? */
1378 if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1379 return 0;
1380 pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1381 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1382 }
1383 iskb = skb;
1384 }
1385
1386 do {
1387 seqno = buf_seqno(iskb);
1388 if (unlikely(less(seqno, l->drop_point))) {
1389 kfree_skb(iskb);
1390 continue;
1391 }
1392 if (unlikely(seqno != l->drop_point)) {
1393 __tipc_skb_queue_sorted(fdefq, seqno, iskb);
1394 continue;
1395 }
1396
1397 l->drop_point++;
1398 if (!tipc_data_input(l, iskb, inputq))
1399 rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1400 if (unlikely(rc))
1401 break;
1402 } while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1403
1404 return rc;
1405}
1406
1407/**
1408 * tipc_get_gap_ack_blks - get Gap ACK blocks from PROTOCOL/STATE_MSG
1409 * @ga: returned pointer to the Gap ACK blocks if any
1410 * @l: the tipc link
1411 * @hdr: the PROTOCOL/STATE_MSG header
1412 * @uc: desired Gap ACK blocks type, i.e. unicast (= 1) or broadcast (= 0)
1413 *
1414 * Return: the total Gap ACK blocks size
1415 */
1416u16 tipc_get_gap_ack_blks(struct tipc_gap_ack_blks **ga, struct tipc_link *l,
1417 struct tipc_msg *hdr, bool uc)
1418{
1419 struct tipc_gap_ack_blks *p;
1420 u16 sz = 0;
1421
1422 /* Does peer support the Gap ACK blocks feature? */
1423 if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
1424 p = (struct tipc_gap_ack_blks *)msg_data(hdr);
1425 sz = ntohs(p->len);
1426 /* Sanity check */
1427 if (sz == struct_size(p, gacks, size_add(p->ugack_cnt, p->bgack_cnt))) {
1428 /* Good, check if the desired type exists */
1429 if ((uc && p->ugack_cnt) || (!uc && p->bgack_cnt))
1430 goto ok;
1431 /* Backward compatible: peer might not support bc, but uc? */
1432 } else if (uc && sz == struct_size(p, gacks, p->ugack_cnt)) {
1433 if (p->ugack_cnt) {
1434 p->bgack_cnt = 0;
1435 goto ok;
1436 }
1437 }
1438 }
1439 /* Other cases: ignore! */
1440 p = NULL;
1441
1442ok:
1443 *ga = p;
1444 return sz;
1445}
1446
1447static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
1448 struct tipc_link *l, u8 start_index)
1449{
1450 struct tipc_gap_ack *gacks = &ga->gacks[start_index];
1451 struct sk_buff *skb = skb_peek(&l->deferdq);
1452 u16 expect, seqno = 0;
1453 u8 n = 0;
1454
1455 if (!skb)
1456 return 0;
1457
1458 expect = buf_seqno(skb);
1459 skb_queue_walk(&l->deferdq, skb) {
1460 seqno = buf_seqno(skb);
1461 if (unlikely(more(seqno, expect))) {
1462 gacks[n].ack = htons(expect - 1);
1463 gacks[n].gap = htons(seqno - expect);
1464 if (++n >= MAX_GAP_ACK_BLKS / 2) {
1465 pr_info_ratelimited("Gacks on %s: %d, ql: %d!\n",
1466 l->name, n,
1467 skb_queue_len(&l->deferdq));
1468 return n;
1469 }
1470 } else if (unlikely(less(seqno, expect))) {
1471 pr_warn("Unexpected skb in deferdq!\n");
1472 continue;
1473 }
1474 expect = seqno + 1;
1475 }
1476
1477 /* last block */
1478 gacks[n].ack = htons(seqno);
1479 gacks[n].gap = 0;
1480 n++;
1481 return n;
1482}
1483
1484/* tipc_build_gap_ack_blks - build Gap ACK blocks
1485 * @l: tipc unicast link
1486 * @hdr: the tipc message buffer to store the Gap ACK blocks after built
1487 *
1488 * The function builds Gap ACK blocks for both the unicast & broadcast receiver
1489 * links of a certain peer, the buffer after built has the network data format
1490 * as found at the struct tipc_gap_ack_blks definition.
1491 *
1492 * returns the actual allocated memory size
1493 */
1494static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr)
1495{
1496 struct tipc_link *bcl = l->bc_rcvlink;
1497 struct tipc_gap_ack_blks *ga;
1498 u16 len;
1499
1500 ga = (struct tipc_gap_ack_blks *)msg_data(hdr);
1501
1502 /* Start with broadcast link first */
1503 tipc_bcast_lock(bcl->net);
1504 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1505 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1506 ga->bgack_cnt = __tipc_build_gap_ack_blks(ga, bcl, 0);
1507 tipc_bcast_unlock(bcl->net);
1508
1509 /* Now for unicast link, but an explicit NACK only (???) */
1510 ga->ugack_cnt = (msg_seq_gap(hdr)) ?
1511 __tipc_build_gap_ack_blks(ga, l, ga->bgack_cnt) : 0;
1512
1513 /* Total len */
1514 len = struct_size(ga, gacks, size_add(ga->bgack_cnt, ga->ugack_cnt));
1515 ga->len = htons(len);
1516 return len;
1517}
1518
1519/* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1520 * acked packets, also doing retransmissions if
1521 * gaps found
1522 * @l: tipc link with transmq queue to be advanced
1523 * @r: tipc link "receiver" i.e. in case of broadcast (= "l" if unicast)
1524 * @acked: seqno of last packet acked by peer without any gaps before
1525 * @gap: # of gap packets
1526 * @ga: buffer pointer to Gap ACK blocks from peer
1527 * @xmitq: queue for accumulating the retransmitted packets if any
1528 * @retransmitted: returned boolean value if a retransmission is really issued
1529 * @rc: returned code e.g. TIPC_LINK_DOWN_EVT if a repeated retransmit failures
1530 * happens (- unlikely case)
1531 *
1532 * Return: the number of packets released from the link transmq
1533 */
1534static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r,
1535 u16 acked, u16 gap,
1536 struct tipc_gap_ack_blks *ga,
1537 struct sk_buff_head *xmitq,
1538 bool *retransmitted, int *rc)
1539{
1540 struct tipc_gap_ack_blks *last_ga = r->last_ga, *this_ga = NULL;
1541 struct tipc_gap_ack *gacks = NULL;
1542 struct sk_buff *skb, *_skb, *tmp;
1543 struct tipc_msg *hdr;
1544 u32 qlen = skb_queue_len(&l->transmq);
1545 u16 nacked = acked, ngap = gap, gack_cnt = 0;
1546 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1547 u16 ack = l->rcv_nxt - 1;
1548 u16 seqno, n = 0;
1549 u16 end = r->acked, start = end, offset = r->last_gap;
1550 u16 si = (last_ga) ? last_ga->start_index : 0;
1551 bool is_uc = !link_is_bc_sndlink(l);
1552 bool bc_has_acked = false;
1553
1554 trace_tipc_link_retrans(r, acked + 1, acked + gap, &l->transmq);
1555
1556 /* Determine Gap ACK blocks if any for the particular link */
1557 if (ga && is_uc) {
1558 /* Get the Gap ACKs, uc part */
1559 gack_cnt = ga->ugack_cnt;
1560 gacks = &ga->gacks[ga->bgack_cnt];
1561 } else if (ga) {
1562 /* Copy the Gap ACKs, bc part, for later renewal if needed */
1563 this_ga = kmemdup(ga, struct_size(ga, gacks, ga->bgack_cnt),
1564 GFP_ATOMIC);
1565 if (likely(this_ga)) {
1566 this_ga->start_index = 0;
1567 /* Start with the bc Gap ACKs */
1568 gack_cnt = this_ga->bgack_cnt;
1569 gacks = &this_ga->gacks[0];
1570 } else {
1571 /* Hmm, we can get in trouble..., simply ignore it */
1572 pr_warn_ratelimited("Ignoring bc Gap ACKs, no memory\n");
1573 }
1574 }
1575
1576 /* Advance the link transmq */
1577 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1578 seqno = buf_seqno(skb);
1579
1580next_gap_ack:
1581 if (less_eq(seqno, nacked)) {
1582 if (is_uc)
1583 goto release;
1584 /* Skip packets peer has already acked */
1585 if (!more(seqno, r->acked))
1586 continue;
1587 /* Get the next of last Gap ACK blocks */
1588 while (more(seqno, end)) {
1589 if (!last_ga || si >= last_ga->bgack_cnt)
1590 break;
1591 start = end + offset + 1;
1592 end = ntohs(last_ga->gacks[si].ack);
1593 offset = ntohs(last_ga->gacks[si].gap);
1594 si++;
1595 WARN_ONCE(more(start, end) ||
1596 (!offset &&
1597 si < last_ga->bgack_cnt) ||
1598 si > MAX_GAP_ACK_BLKS,
1599 "Corrupted Gap ACK: %d %d %d %d %d\n",
1600 start, end, offset, si,
1601 last_ga->bgack_cnt);
1602 }
1603 /* Check against the last Gap ACK block */
1604 if (tipc_in_range(seqno, start, end))
1605 continue;
1606 /* Update/release the packet peer is acking */
1607 bc_has_acked = true;
1608 if (--TIPC_SKB_CB(skb)->ackers)
1609 continue;
1610release:
1611 /* release skb */
1612 __skb_unlink(skb, &l->transmq);
1613 kfree_skb(skb);
1614 } else if (less_eq(seqno, nacked + ngap)) {
1615 /* First gap: check if repeated retrans failures? */
1616 if (unlikely(seqno == acked + 1 &&
1617 link_retransmit_failure(l, r, rc))) {
1618 /* Ignore this bc Gap ACKs if any */
1619 kfree(this_ga);
1620 this_ga = NULL;
1621 break;
1622 }
1623 /* retransmit skb if unrestricted*/
1624 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1625 continue;
1626 tipc_link_set_skb_retransmit_time(skb, l);
1627 _skb = pskb_copy(skb, GFP_ATOMIC);
1628 if (!_skb)
1629 continue;
1630 hdr = buf_msg(_skb);
1631 msg_set_ack(hdr, ack);
1632 msg_set_bcast_ack(hdr, bc_ack);
1633 _skb->priority = TC_PRIO_CONTROL;
1634 __skb_queue_tail(xmitq, _skb);
1635 l->stats.retransmitted++;
1636 if (!is_uc)
1637 r->stats.retransmitted++;
1638 *retransmitted = true;
1639 /* Increase actual retrans counter & mark first time */
1640 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1641 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1642 } else {
1643 /* retry with Gap ACK blocks if any */
1644 if (n >= gack_cnt)
1645 break;
1646 nacked = ntohs(gacks[n].ack);
1647 ngap = ntohs(gacks[n].gap);
1648 n++;
1649 goto next_gap_ack;
1650 }
1651 }
1652
1653 /* Renew last Gap ACK blocks for bc if needed */
1654 if (bc_has_acked) {
1655 if (this_ga) {
1656 kfree(last_ga);
1657 r->last_ga = this_ga;
1658 r->last_gap = gap;
1659 } else if (last_ga) {
1660 if (less(acked, start)) {
1661 si--;
1662 offset = start - acked - 1;
1663 } else if (less(acked, end)) {
1664 acked = end;
1665 }
1666 if (si < last_ga->bgack_cnt) {
1667 last_ga->start_index = si;
1668 r->last_gap = offset;
1669 } else {
1670 kfree(last_ga);
1671 r->last_ga = NULL;
1672 r->last_gap = 0;
1673 }
1674 } else {
1675 r->last_gap = 0;
1676 }
1677 r->acked = acked;
1678 } else {
1679 kfree(this_ga);
1680 }
1681
1682 return qlen - skb_queue_len(&l->transmq);
1683}
1684
1685/* tipc_link_build_state_msg: prepare link state message for transmission
1686 *
1687 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1688 * risk of ack storms towards the sender
1689 */
1690int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1691{
1692 if (!l)
1693 return 0;
1694
1695 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1696 if (link_is_bc_rcvlink(l)) {
1697 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1698 return 0;
1699 l->rcv_unacked = 0;
1700
1701 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1702 l->snd_nxt = l->rcv_nxt;
1703 return TIPC_LINK_SND_STATE;
1704 }
1705 /* Unicast ACK */
1706 l->rcv_unacked = 0;
1707 l->stats.sent_acks++;
1708 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1709 return 0;
1710}
1711
1712/* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1713 */
1714void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1715{
1716 int mtyp = RESET_MSG;
1717 struct sk_buff *skb;
1718
1719 if (l->state == LINK_ESTABLISHING)
1720 mtyp = ACTIVATE_MSG;
1721
1722 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1723
1724 /* Inform peer that this endpoint is going down if applicable */
1725 skb = skb_peek_tail(xmitq);
1726 if (skb && (l->state == LINK_RESET))
1727 msg_set_peer_stopping(buf_msg(skb), 1);
1728}
1729
1730/* tipc_link_build_nack_msg: prepare link nack message for transmission
1731 * Note that sending of broadcast NACK is coordinated among nodes, to
1732 * reduce the risk of NACK storms towards the sender
1733 */
1734static int tipc_link_build_nack_msg(struct tipc_link *l,
1735 struct sk_buff_head *xmitq)
1736{
1737 u32 def_cnt = ++l->stats.deferred_recv;
1738 struct sk_buff_head *dfq = &l->deferdq;
1739 u32 defq_len = skb_queue_len(dfq);
1740 int match1, match2;
1741
1742 if (link_is_bc_rcvlink(l)) {
1743 match1 = def_cnt & 0xf;
1744 match2 = tipc_own_addr(l->net) & 0xf;
1745 if (match1 == match2)
1746 return TIPC_LINK_SND_STATE;
1747 return 0;
1748 }
1749
1750 if (defq_len >= 3 && !((defq_len - 3) % 16)) {
1751 u16 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1752
1753 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0,
1754 rcvgap, 0, 0, xmitq);
1755 }
1756 return 0;
1757}
1758
1759/* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1760 * @l: the link that should handle the message
1761 * @skb: TIPC packet
1762 * @xmitq: queue to place packets to be sent after this call
1763 */
1764int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1765 struct sk_buff_head *xmitq)
1766{
1767 struct sk_buff_head *defq = &l->deferdq;
1768 struct tipc_msg *hdr = buf_msg(skb);
1769 u16 seqno, rcv_nxt, win_lim;
1770 int released = 0;
1771 int rc = 0;
1772
1773 /* Verify and update link state */
1774 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1775 return tipc_link_proto_rcv(l, skb, xmitq);
1776
1777 /* Don't send probe at next timeout expiration */
1778 l->silent_intv_cnt = 0;
1779
1780 do {
1781 hdr = buf_msg(skb);
1782 seqno = msg_seqno(hdr);
1783 rcv_nxt = l->rcv_nxt;
1784 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1785
1786 if (unlikely(!tipc_link_is_up(l))) {
1787 if (l->state == LINK_ESTABLISHING)
1788 rc = TIPC_LINK_UP_EVT;
1789 kfree_skb(skb);
1790 break;
1791 }
1792
1793 /* Drop if outside receive window */
1794 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1795 l->stats.duplicates++;
1796 kfree_skb(skb);
1797 break;
1798 }
1799 released += tipc_link_advance_transmq(l, l, msg_ack(hdr), 0,
1800 NULL, NULL, NULL, NULL);
1801
1802 /* Defer delivery if sequence gap */
1803 if (unlikely(seqno != rcv_nxt)) {
1804 if (!__tipc_skb_queue_sorted(defq, seqno, skb))
1805 l->stats.duplicates++;
1806 rc |= tipc_link_build_nack_msg(l, xmitq);
1807 break;
1808 }
1809
1810 /* Deliver packet */
1811 l->rcv_nxt++;
1812 l->stats.recv_pkts++;
1813
1814 if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1815 rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1816 else if (!tipc_data_input(l, skb, l->inputq))
1817 rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1818 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1819 rc |= tipc_link_build_state_msg(l, xmitq);
1820 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1821 break;
1822 } while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1823
1824 /* Forward queues and wake up waiting users */
1825 if (released) {
1826 tipc_link_update_cwin(l, released, 0);
1827 tipc_link_advance_backlog(l, xmitq);
1828 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1829 link_prepare_wakeup(l);
1830 }
1831 return rc;
1832}
1833
1834static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1835 bool probe_reply, u16 rcvgap,
1836 int tolerance, int priority,
1837 struct sk_buff_head *xmitq)
1838{
1839 struct tipc_mon_state *mstate = &l->mon_state;
1840 struct sk_buff_head *dfq = &l->deferdq;
1841 struct tipc_link *bcl = l->bc_rcvlink;
1842 struct tipc_msg *hdr;
1843 struct sk_buff *skb;
1844 bool node_up = tipc_link_is_up(bcl);
1845 u16 glen = 0, bc_rcvgap = 0;
1846 int dlen = 0;
1847 void *data;
1848
1849 /* Don't send protocol message during reset or link failover */
1850 if (tipc_link_is_blocked(l))
1851 return;
1852
1853 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1854 return;
1855
1856 if ((probe || probe_reply) && !skb_queue_empty(dfq))
1857 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1858
1859 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1860 tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1861 l->addr, tipc_own_addr(l->net), 0, 0, 0);
1862 if (!skb)
1863 return;
1864
1865 hdr = buf_msg(skb);
1866 data = msg_data(hdr);
1867 msg_set_session(hdr, l->session);
1868 msg_set_bearer_id(hdr, l->bearer_id);
1869 msg_set_net_plane(hdr, l->net_plane);
1870 msg_set_next_sent(hdr, l->snd_nxt);
1871 msg_set_ack(hdr, l->rcv_nxt - 1);
1872 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1873 msg_set_bc_ack_invalid(hdr, !node_up);
1874 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1875 msg_set_link_tolerance(hdr, tolerance);
1876 msg_set_linkprio(hdr, priority);
1877 msg_set_redundant_link(hdr, node_up);
1878 msg_set_seq_gap(hdr, 0);
1879 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1880
1881 if (mtyp == STATE_MSG) {
1882 if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1883 msg_set_seqno(hdr, l->snd_nxt_state++);
1884 msg_set_seq_gap(hdr, rcvgap);
1885 bc_rcvgap = link_bc_rcv_gap(bcl);
1886 msg_set_bc_gap(hdr, bc_rcvgap);
1887 msg_set_probe(hdr, probe);
1888 msg_set_is_keepalive(hdr, probe || probe_reply);
1889 if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1890 glen = tipc_build_gap_ack_blks(l, hdr);
1891 tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1892 msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1893 skb_trim(skb, INT_H_SIZE + glen + dlen);
1894 l->stats.sent_states++;
1895 l->rcv_unacked = 0;
1896 } else {
1897 /* RESET_MSG or ACTIVATE_MSG */
1898 if (mtyp == ACTIVATE_MSG) {
1899 msg_set_dest_session_valid(hdr, 1);
1900 msg_set_dest_session(hdr, l->peer_session);
1901 }
1902 msg_set_max_pkt(hdr, l->advertised_mtu);
1903 strcpy(data, l->if_name);
1904 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1905 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1906 }
1907 if (probe)
1908 l->stats.sent_probes++;
1909 if (rcvgap)
1910 l->stats.sent_nacks++;
1911 if (bc_rcvgap)
1912 bcl->stats.sent_nacks++;
1913 skb->priority = TC_PRIO_CONTROL;
1914 __skb_queue_tail(xmitq, skb);
1915 trace_tipc_proto_build(skb, false, l->name);
1916}
1917
1918void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1919 struct sk_buff_head *xmitq)
1920{
1921 u32 onode = tipc_own_addr(l->net);
1922 struct tipc_msg *hdr, *ihdr;
1923 struct sk_buff_head tnlq;
1924 struct sk_buff *skb;
1925 u32 dnode = l->addr;
1926
1927 __skb_queue_head_init(&tnlq);
1928 skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1929 INT_H_SIZE, BASIC_H_SIZE,
1930 dnode, onode, 0, 0, 0);
1931 if (!skb) {
1932 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1933 return;
1934 }
1935
1936 hdr = buf_msg(skb);
1937 msg_set_msgcnt(hdr, 1);
1938 msg_set_bearer_id(hdr, l->peer_bearer_id);
1939
1940 ihdr = (struct tipc_msg *)msg_data(hdr);
1941 tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1942 BASIC_H_SIZE, dnode);
1943 msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1944 __skb_queue_tail(&tnlq, skb);
1945 tipc_link_xmit(l, &tnlq, xmitq);
1946}
1947
1948/* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1949 * with contents of the link's transmit and backlog queues.
1950 */
1951void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1952 int mtyp, struct sk_buff_head *xmitq)
1953{
1954 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1955 struct sk_buff *skb, *tnlskb;
1956 struct tipc_msg *hdr, tnlhdr;
1957 struct sk_buff_head *queue = &l->transmq;
1958 struct sk_buff_head tmpxq, tnlq, frags;
1959 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1960 bool pktcnt_need_update = false;
1961 u16 syncpt;
1962 int rc;
1963
1964 if (!tnl)
1965 return;
1966
1967 __skb_queue_head_init(&tnlq);
1968 /* Link Synching:
1969 * From now on, send only one single ("dummy") SYNCH message
1970 * to peer. The SYNCH message does not contain any data, just
1971 * a header conveying the synch point to the peer.
1972 */
1973 if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1974 tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1975 INT_H_SIZE, 0, l->addr,
1976 tipc_own_addr(l->net),
1977 0, 0, 0);
1978 if (!tnlskb) {
1979 pr_warn("%sunable to create dummy SYNCH_MSG\n",
1980 link_co_err);
1981 return;
1982 }
1983
1984 hdr = buf_msg(tnlskb);
1985 syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
1986 msg_set_syncpt(hdr, syncpt);
1987 msg_set_bearer_id(hdr, l->peer_bearer_id);
1988 __skb_queue_tail(&tnlq, tnlskb);
1989 tipc_link_xmit(tnl, &tnlq, xmitq);
1990 return;
1991 }
1992
1993 __skb_queue_head_init(&tmpxq);
1994 __skb_queue_head_init(&frags);
1995 /* At least one packet required for safe algorithm => add dummy */
1996 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1997 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1998 0, 0, TIPC_ERR_NO_PORT);
1999 if (!skb) {
2000 pr_warn("%sunable to create tunnel packet\n", link_co_err);
2001 return;
2002 }
2003 __skb_queue_tail(&tnlq, skb);
2004 tipc_link_xmit(l, &tnlq, &tmpxq);
2005 __skb_queue_purge(&tmpxq);
2006
2007 /* Initialize reusable tunnel packet header */
2008 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
2009 mtyp, INT_H_SIZE, l->addr);
2010 if (mtyp == SYNCH_MSG)
2011 pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
2012 else
2013 pktcnt = skb_queue_len(&l->transmq);
2014 pktcnt += skb_queue_len(&l->backlogq);
2015 msg_set_msgcnt(&tnlhdr, pktcnt);
2016 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
2017tnl:
2018 /* Wrap each packet into a tunnel packet */
2019 skb_queue_walk(queue, skb) {
2020 hdr = buf_msg(skb);
2021 if (queue == &l->backlogq)
2022 msg_set_seqno(hdr, seqno++);
2023 pktlen = msg_size(hdr);
2024
2025 /* Tunnel link MTU is not large enough? This could be
2026 * due to:
2027 * 1) Link MTU has just changed or set differently;
2028 * 2) Or FAILOVER on the top of a SYNCH message
2029 *
2030 * The 2nd case should not happen if peer supports
2031 * TIPC_TUNNEL_ENHANCED
2032 */
2033 if (pktlen > tnl->mtu - INT_H_SIZE) {
2034 if (mtyp == FAILOVER_MSG &&
2035 (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
2036 rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
2037 &frags);
2038 if (rc) {
2039 pr_warn("%sunable to frag msg: rc %d\n",
2040 link_co_err, rc);
2041 return;
2042 }
2043 pktcnt += skb_queue_len(&frags) - 1;
2044 pktcnt_need_update = true;
2045 skb_queue_splice_tail_init(&frags, &tnlq);
2046 continue;
2047 }
2048 /* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
2049 * => Just warn it and return!
2050 */
2051 pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
2052 link_co_err, msg_user(hdr),
2053 msg_type(hdr), msg_size(hdr));
2054 return;
2055 }
2056
2057 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
2058 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
2059 if (!tnlskb) {
2060 pr_warn("%sunable to send packet\n", link_co_err);
2061 return;
2062 }
2063 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
2064 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
2065 __skb_queue_tail(&tnlq, tnlskb);
2066 }
2067 if (queue != &l->backlogq) {
2068 queue = &l->backlogq;
2069 goto tnl;
2070 }
2071
2072 if (pktcnt_need_update)
2073 skb_queue_walk(&tnlq, skb) {
2074 hdr = buf_msg(skb);
2075 msg_set_msgcnt(hdr, pktcnt);
2076 }
2077
2078 tipc_link_xmit(tnl, &tnlq, xmitq);
2079
2080 if (mtyp == FAILOVER_MSG) {
2081 tnl->drop_point = l->rcv_nxt;
2082 tnl->failover_reasm_skb = l->reasm_buf;
2083 l->reasm_buf = NULL;
2084
2085 /* Failover the link's deferdq */
2086 if (unlikely(!skb_queue_empty(fdefq))) {
2087 pr_warn("Link failover deferdq not empty: %d!\n",
2088 skb_queue_len(fdefq));
2089 __skb_queue_purge(fdefq);
2090 }
2091 skb_queue_splice_init(&l->deferdq, fdefq);
2092 }
2093}
2094
2095/**
2096 * tipc_link_failover_prepare() - prepare tnl for link failover
2097 *
2098 * This is a special version of the precursor - tipc_link_tnl_prepare(),
2099 * see the tipc_node_link_failover() for details
2100 *
2101 * @l: failover link
2102 * @tnl: tunnel link
2103 * @xmitq: queue for messages to be xmited
2104 */
2105void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
2106 struct sk_buff_head *xmitq)
2107{
2108 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
2109
2110 tipc_link_create_dummy_tnl_msg(tnl, xmitq);
2111
2112 /* This failover link endpoint was never established before,
2113 * so it has not received anything from peer.
2114 * Otherwise, it must be a normal failover situation or the
2115 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
2116 * would have to start over from scratch instead.
2117 */
2118 tnl->drop_point = 1;
2119 tnl->failover_reasm_skb = NULL;
2120
2121 /* Initiate the link's failover deferdq */
2122 if (unlikely(!skb_queue_empty(fdefq))) {
2123 pr_warn("Link failover deferdq not empty: %d!\n",
2124 skb_queue_len(fdefq));
2125 __skb_queue_purge(fdefq);
2126 }
2127}
2128
2129/* tipc_link_validate_msg(): validate message against current link state
2130 * Returns true if message should be accepted, otherwise false
2131 */
2132bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
2133{
2134 u16 curr_session = l->peer_session;
2135 u16 session = msg_session(hdr);
2136 int mtyp = msg_type(hdr);
2137
2138 if (msg_user(hdr) != LINK_PROTOCOL)
2139 return true;
2140
2141 switch (mtyp) {
2142 case RESET_MSG:
2143 if (!l->in_session)
2144 return true;
2145 /* Accept only RESET with new session number */
2146 return more(session, curr_session);
2147 case ACTIVATE_MSG:
2148 if (!l->in_session)
2149 return true;
2150 /* Accept only ACTIVATE with new or current session number */
2151 return !less(session, curr_session);
2152 case STATE_MSG:
2153 /* Accept only STATE with current session number */
2154 if (!l->in_session)
2155 return false;
2156 if (session != curr_session)
2157 return false;
2158 /* Extra sanity check */
2159 if (!tipc_link_is_up(l) && msg_ack(hdr))
2160 return false;
2161 if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
2162 return true;
2163 /* Accept only STATE with new sequence number */
2164 return !less(msg_seqno(hdr), l->rcv_nxt_state);
2165 default:
2166 return false;
2167 }
2168}
2169
2170/* tipc_link_proto_rcv(): receive link level protocol message :
2171 * Note that network plane id propagates through the network, and may
2172 * change at any time. The node with lowest numerical id determines
2173 * network plane
2174 */
2175static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
2176 struct sk_buff_head *xmitq)
2177{
2178 struct tipc_msg *hdr = buf_msg(skb);
2179 struct tipc_gap_ack_blks *ga = NULL;
2180 bool reply = msg_probe(hdr), retransmitted = false;
2181 u32 dlen = msg_data_sz(hdr), glen = 0, msg_max;
2182 u16 peers_snd_nxt = msg_next_sent(hdr);
2183 u16 peers_tol = msg_link_tolerance(hdr);
2184 u16 peers_prio = msg_linkprio(hdr);
2185 u16 gap = msg_seq_gap(hdr);
2186 u16 ack = msg_ack(hdr);
2187 u16 rcv_nxt = l->rcv_nxt;
2188 u16 rcvgap = 0;
2189 int mtyp = msg_type(hdr);
2190 int rc = 0, released;
2191 char *if_name;
2192 void *data;
2193
2194 trace_tipc_proto_rcv(skb, false, l->name);
2195
2196 if (dlen > U16_MAX)
2197 goto exit;
2198
2199 if (tipc_link_is_blocked(l) || !xmitq)
2200 goto exit;
2201
2202 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
2203 l->net_plane = msg_net_plane(hdr);
2204
2205 if (skb_linearize(skb))
2206 goto exit;
2207
2208 hdr = buf_msg(skb);
2209 data = msg_data(hdr);
2210
2211 if (!tipc_link_validate_msg(l, hdr)) {
2212 trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
2213 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
2214 goto exit;
2215 }
2216
2217 switch (mtyp) {
2218 case RESET_MSG:
2219 case ACTIVATE_MSG:
2220 msg_max = msg_max_pkt(hdr);
2221 if (msg_max < tipc_bearer_min_mtu(l->net, l->bearer_id))
2222 break;
2223 /* Complete own link name with peer's interface name */
2224 if_name = strrchr(l->name, ':') + 1;
2225 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
2226 break;
2227 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
2228 break;
2229 strncpy(if_name, data, TIPC_MAX_IF_NAME);
2230
2231 /* Update own tolerance if peer indicates a non-zero value */
2232 if (tipc_in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2233 l->tolerance = peers_tol;
2234 l->bc_rcvlink->tolerance = peers_tol;
2235 }
2236 /* Update own priority if peer's priority is higher */
2237 if (tipc_in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
2238 l->priority = peers_prio;
2239
2240 /* If peer is going down we want full re-establish cycle */
2241 if (msg_peer_stopping(hdr)) {
2242 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2243 break;
2244 }
2245
2246 /* If this endpoint was re-created while peer was ESTABLISHING
2247 * it doesn't know current session number. Force re-synch.
2248 */
2249 if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2250 l->session != msg_dest_session(hdr)) {
2251 if (less(l->session, msg_dest_session(hdr)))
2252 l->session = msg_dest_session(hdr) + 1;
2253 break;
2254 }
2255
2256 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2257 if (mtyp == RESET_MSG || !tipc_link_is_up(l))
2258 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2259
2260 /* ACTIVATE_MSG takes up link if it was already locally reset */
2261 if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2262 rc = TIPC_LINK_UP_EVT;
2263
2264 l->peer_session = msg_session(hdr);
2265 l->in_session = true;
2266 l->peer_bearer_id = msg_bearer_id(hdr);
2267 if (l->mtu > msg_max)
2268 l->mtu = msg_max;
2269 break;
2270
2271 case STATE_MSG:
2272 /* Validate Gap ACK blocks, drop if invalid */
2273 glen = tipc_get_gap_ack_blks(&ga, l, hdr, true);
2274 if (glen > dlen)
2275 break;
2276
2277 l->rcv_nxt_state = msg_seqno(hdr) + 1;
2278
2279 /* Update own tolerance if peer indicates a non-zero value */
2280 if (tipc_in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2281 l->tolerance = peers_tol;
2282 l->bc_rcvlink->tolerance = peers_tol;
2283 }
2284 /* Update own prio if peer indicates a different value */
2285 if ((peers_prio != l->priority) &&
2286 tipc_in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2287 l->priority = peers_prio;
2288 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2289 }
2290
2291 l->silent_intv_cnt = 0;
2292 l->stats.recv_states++;
2293 if (msg_probe(hdr))
2294 l->stats.recv_probes++;
2295
2296 if (!tipc_link_is_up(l)) {
2297 if (l->state == LINK_ESTABLISHING)
2298 rc = TIPC_LINK_UP_EVT;
2299 break;
2300 }
2301
2302 tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2303 &l->mon_state, l->bearer_id);
2304
2305 /* Send NACK if peer has sent pkts we haven't received yet */
2306 if ((reply || msg_is_keepalive(hdr)) &&
2307 more(peers_snd_nxt, rcv_nxt) &&
2308 !tipc_link_is_synching(l) &&
2309 skb_queue_empty(&l->deferdq))
2310 rcvgap = peers_snd_nxt - l->rcv_nxt;
2311 if (rcvgap || reply)
2312 tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2313 rcvgap, 0, 0, xmitq);
2314
2315 released = tipc_link_advance_transmq(l, l, ack, gap, ga, xmitq,
2316 &retransmitted, &rc);
2317 if (gap)
2318 l->stats.recv_nacks++;
2319 if (released || retransmitted)
2320 tipc_link_update_cwin(l, released, retransmitted);
2321 if (released)
2322 tipc_link_advance_backlog(l, xmitq);
2323 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2324 link_prepare_wakeup(l);
2325 }
2326exit:
2327 kfree_skb(skb);
2328 return rc;
2329}
2330
2331/* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2332 */
2333static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2334 u16 peers_snd_nxt,
2335 struct sk_buff_head *xmitq)
2336{
2337 struct sk_buff *skb;
2338 struct tipc_msg *hdr;
2339 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2340 u16 ack = l->rcv_nxt - 1;
2341 u16 gap_to = peers_snd_nxt - 1;
2342
2343 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2344 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2345 if (!skb)
2346 return false;
2347 hdr = buf_msg(skb);
2348 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2349 msg_set_bcast_ack(hdr, ack);
2350 msg_set_bcgap_after(hdr, ack);
2351 if (dfrd_skb)
2352 gap_to = buf_seqno(dfrd_skb) - 1;
2353 msg_set_bcgap_to(hdr, gap_to);
2354 msg_set_non_seq(hdr, bcast);
2355 __skb_queue_tail(xmitq, skb);
2356 return true;
2357}
2358
2359/* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2360 *
2361 * Give a newly added peer node the sequence number where it should
2362 * start receiving and acking broadcast packets.
2363 */
2364static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2365 struct sk_buff_head *xmitq)
2366{
2367 struct sk_buff_head list;
2368
2369 __skb_queue_head_init(&list);
2370 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2371 return;
2372 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2373 tipc_link_xmit(l, &list, xmitq);
2374}
2375
2376/* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2377 */
2378void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2379{
2380 int mtyp = msg_type(hdr);
2381 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2382
2383 if (tipc_link_is_up(l))
2384 return;
2385
2386 if (msg_user(hdr) == BCAST_PROTOCOL) {
2387 l->rcv_nxt = peers_snd_nxt;
2388 l->state = LINK_ESTABLISHED;
2389 return;
2390 }
2391
2392 if (l->peer_caps & TIPC_BCAST_SYNCH)
2393 return;
2394
2395 if (msg_peer_node_is_up(hdr))
2396 return;
2397
2398 /* Compatibility: accept older, less safe initial synch data */
2399 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2400 l->rcv_nxt = peers_snd_nxt;
2401}
2402
2403/* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2404 */
2405int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2406 struct sk_buff_head *xmitq)
2407{
2408 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2409 int rc = 0;
2410
2411 if (!tipc_link_is_up(l))
2412 return rc;
2413
2414 if (!msg_peer_node_is_up(hdr))
2415 return rc;
2416
2417 /* Open when peer acknowledges our bcast init msg (pkt #1) */
2418 if (msg_ack(hdr))
2419 l->bc_peer_is_up = true;
2420
2421 if (!l->bc_peer_is_up)
2422 return rc;
2423
2424 /* Ignore if peers_snd_nxt goes beyond receive window */
2425 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2426 return rc;
2427
2428 l->snd_nxt = peers_snd_nxt;
2429 if (link_bc_rcv_gap(l))
2430 rc |= TIPC_LINK_SND_STATE;
2431
2432 /* Return now if sender supports nack via STATE messages */
2433 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2434 return rc;
2435
2436 /* Otherwise, be backwards compatible */
2437
2438 if (!more(peers_snd_nxt, l->rcv_nxt)) {
2439 l->nack_state = BC_NACK_SND_CONDITIONAL;
2440 return 0;
2441 }
2442
2443 /* Don't NACK if one was recently sent or peeked */
2444 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2445 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2446 return 0;
2447 }
2448
2449 /* Conditionally delay NACK sending until next synch rcv */
2450 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2451 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2452 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2453 return 0;
2454 }
2455
2456 /* Send NACK now but suppress next one */
2457 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2458 l->nack_state = BC_NACK_SND_SUPPRESS;
2459 return 0;
2460}
2461
2462int tipc_link_bc_ack_rcv(struct tipc_link *r, u16 acked, u16 gap,
2463 struct tipc_gap_ack_blks *ga,
2464 struct sk_buff_head *xmitq,
2465 struct sk_buff_head *retrq)
2466{
2467 struct tipc_link *l = r->bc_sndlink;
2468 bool unused = false;
2469 int rc = 0;
2470
2471 if (!tipc_link_is_up(r) || !r->bc_peer_is_up)
2472 return 0;
2473
2474 if (gap) {
2475 l->stats.recv_nacks++;
2476 r->stats.recv_nacks++;
2477 }
2478
2479 if (less(acked, r->acked) || (acked == r->acked && !gap && !ga))
2480 return 0;
2481
2482 trace_tipc_link_bc_ack(r, acked, gap, &l->transmq);
2483 tipc_link_advance_transmq(l, r, acked, gap, ga, retrq, &unused, &rc);
2484
2485 tipc_link_advance_backlog(l, xmitq);
2486 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2487 link_prepare_wakeup(l);
2488
2489 return rc;
2490}
2491
2492/* tipc_link_bc_nack_rcv(): receive broadcast nack message
2493 * This function is here for backwards compatibility, since
2494 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2495 */
2496int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2497 struct sk_buff_head *xmitq)
2498{
2499 struct tipc_msg *hdr = buf_msg(skb);
2500 u32 dnode = msg_destnode(hdr);
2501 int mtyp = msg_type(hdr);
2502 u16 acked = msg_bcast_ack(hdr);
2503 u16 from = acked + 1;
2504 u16 to = msg_bcgap_to(hdr);
2505 u16 peers_snd_nxt = to + 1;
2506 int rc = 0;
2507
2508 kfree_skb(skb);
2509
2510 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2511 return 0;
2512
2513 if (mtyp != STATE_MSG)
2514 return 0;
2515
2516 if (dnode == tipc_own_addr(l->net)) {
2517 rc = tipc_link_bc_ack_rcv(l, acked, to - acked, NULL, xmitq,
2518 xmitq);
2519 l->stats.recv_nacks++;
2520 return rc;
2521 }
2522
2523 /* Msg for other node => suppress own NACK at next sync if applicable */
2524 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2525 l->nack_state = BC_NACK_SND_SUPPRESS;
2526
2527 return 0;
2528}
2529
2530void tipc_link_set_queue_limits(struct tipc_link *l, u32 min_win, u32 max_win)
2531{
2532 int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2533
2534 l->min_win = min_win;
2535 l->ssthresh = max_win;
2536 l->max_win = max_win;
2537 l->window = min_win;
2538 l->backlog[TIPC_LOW_IMPORTANCE].limit = min_win * 2;
2539 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = min_win * 4;
2540 l->backlog[TIPC_HIGH_IMPORTANCE].limit = min_win * 6;
2541 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = min_win * 8;
2542 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
2543}
2544
2545/**
2546 * tipc_link_reset_stats - reset link statistics
2547 * @l: pointer to link
2548 */
2549void tipc_link_reset_stats(struct tipc_link *l)
2550{
2551 memset(&l->stats, 0, sizeof(l->stats));
2552}
2553
2554static void link_print(struct tipc_link *l, const char *str)
2555{
2556 struct sk_buff *hskb = skb_peek(&l->transmq);
2557 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2558 u16 tail = l->snd_nxt - 1;
2559
2560 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2561 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2562 skb_queue_len(&l->transmq), head, tail,
2563 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2564}
2565
2566/* Parse and validate nested (link) properties valid for media, bearer and link
2567 */
2568int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2569{
2570 int err;
2571
2572 err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2573 tipc_nl_prop_policy, NULL);
2574 if (err)
2575 return err;
2576
2577 if (props[TIPC_NLA_PROP_PRIO]) {
2578 u32 prio;
2579
2580 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2581 if (prio > TIPC_MAX_LINK_PRI)
2582 return -EINVAL;
2583 }
2584
2585 if (props[TIPC_NLA_PROP_TOL]) {
2586 u32 tol;
2587
2588 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2589 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2590 return -EINVAL;
2591 }
2592
2593 if (props[TIPC_NLA_PROP_WIN]) {
2594 u32 max_win;
2595
2596 max_win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2597 if (max_win < TIPC_DEF_LINK_WIN || max_win > TIPC_MAX_LINK_WIN)
2598 return -EINVAL;
2599 }
2600
2601 return 0;
2602}
2603
2604static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2605{
2606 int i;
2607 struct nlattr *stats;
2608
2609 struct nla_map {
2610 u32 key;
2611 u32 val;
2612 };
2613
2614 struct nla_map map[] = {
2615 {TIPC_NLA_STATS_RX_INFO, 0},
2616 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2617 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2618 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2619 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2620 {TIPC_NLA_STATS_TX_INFO, 0},
2621 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2622 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2623 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2624 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2625 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2626 s->msg_length_counts : 1},
2627 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2628 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2629 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2630 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2631 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2632 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2633 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2634 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2635 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2636 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
2637 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2638 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2639 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2640 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
2641 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2642 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2643 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2644 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2645 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2646 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2647 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2648 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2649 (s->accu_queue_sz / s->queue_sz_counts) : 0}
2650 };
2651
2652 stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2653 if (!stats)
2654 return -EMSGSIZE;
2655
2656 for (i = 0; i < ARRAY_SIZE(map); i++)
2657 if (nla_put_u32(skb, map[i].key, map[i].val))
2658 goto msg_full;
2659
2660 nla_nest_end(skb, stats);
2661
2662 return 0;
2663msg_full:
2664 nla_nest_cancel(skb, stats);
2665
2666 return -EMSGSIZE;
2667}
2668
2669/* Caller should hold appropriate locks to protect the link */
2670int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2671 struct tipc_link *link, int nlflags)
2672{
2673 u32 self = tipc_own_addr(net);
2674 struct nlattr *attrs;
2675 struct nlattr *prop;
2676 void *hdr;
2677 int err;
2678
2679 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2680 nlflags, TIPC_NL_LINK_GET);
2681 if (!hdr)
2682 return -EMSGSIZE;
2683
2684 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2685 if (!attrs)
2686 goto msg_full;
2687
2688 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2689 goto attr_msg_full;
2690 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2691 goto attr_msg_full;
2692 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2693 goto attr_msg_full;
2694 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2695 goto attr_msg_full;
2696 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2697 goto attr_msg_full;
2698
2699 if (tipc_link_is_up(link))
2700 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2701 goto attr_msg_full;
2702 if (link->active)
2703 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2704 goto attr_msg_full;
2705
2706 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2707 if (!prop)
2708 goto attr_msg_full;
2709 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2710 goto prop_msg_full;
2711 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2712 goto prop_msg_full;
2713 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2714 link->window))
2715 goto prop_msg_full;
2716 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2717 goto prop_msg_full;
2718 nla_nest_end(msg->skb, prop);
2719
2720 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2721 if (err)
2722 goto attr_msg_full;
2723
2724 nla_nest_end(msg->skb, attrs);
2725 genlmsg_end(msg->skb, hdr);
2726
2727 return 0;
2728
2729prop_msg_full:
2730 nla_nest_cancel(msg->skb, prop);
2731attr_msg_full:
2732 nla_nest_cancel(msg->skb, attrs);
2733msg_full:
2734 genlmsg_cancel(msg->skb, hdr);
2735
2736 return -EMSGSIZE;
2737}
2738
2739static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2740 struct tipc_stats *stats)
2741{
2742 int i;
2743 struct nlattr *nest;
2744
2745 struct nla_map {
2746 __u32 key;
2747 __u32 val;
2748 };
2749
2750 struct nla_map map[] = {
2751 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2752 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2753 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2754 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2755 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2756 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2757 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2758 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2759 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2760 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2761 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2762 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2763 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2764 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2765 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2766 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2767 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2768 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2769 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2770 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2771 };
2772
2773 nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2774 if (!nest)
2775 return -EMSGSIZE;
2776
2777 for (i = 0; i < ARRAY_SIZE(map); i++)
2778 if (nla_put_u32(skb, map[i].key, map[i].val))
2779 goto msg_full;
2780
2781 nla_nest_end(skb, nest);
2782
2783 return 0;
2784msg_full:
2785 nla_nest_cancel(skb, nest);
2786
2787 return -EMSGSIZE;
2788}
2789
2790int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg,
2791 struct tipc_link *bcl)
2792{
2793 int err;
2794 void *hdr;
2795 struct nlattr *attrs;
2796 struct nlattr *prop;
2797 u32 bc_mode = tipc_bcast_get_mode(net);
2798 u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2799
2800 if (!bcl)
2801 return 0;
2802
2803 tipc_bcast_lock(net);
2804
2805 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2806 NLM_F_MULTI, TIPC_NL_LINK_GET);
2807 if (!hdr) {
2808 tipc_bcast_unlock(net);
2809 return -EMSGSIZE;
2810 }
2811
2812 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2813 if (!attrs)
2814 goto msg_full;
2815
2816 /* The broadcast link is always up */
2817 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2818 goto attr_msg_full;
2819
2820 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2821 goto attr_msg_full;
2822 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2823 goto attr_msg_full;
2824 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2825 goto attr_msg_full;
2826 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2827 goto attr_msg_full;
2828
2829 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2830 if (!prop)
2831 goto attr_msg_full;
2832 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->max_win))
2833 goto prop_msg_full;
2834 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2835 goto prop_msg_full;
2836 if (bc_mode & BCLINK_MODE_SEL)
2837 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2838 bc_ratio))
2839 goto prop_msg_full;
2840 nla_nest_end(msg->skb, prop);
2841
2842 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2843 if (err)
2844 goto attr_msg_full;
2845
2846 tipc_bcast_unlock(net);
2847 nla_nest_end(msg->skb, attrs);
2848 genlmsg_end(msg->skb, hdr);
2849
2850 return 0;
2851
2852prop_msg_full:
2853 nla_nest_cancel(msg->skb, prop);
2854attr_msg_full:
2855 nla_nest_cancel(msg->skb, attrs);
2856msg_full:
2857 tipc_bcast_unlock(net);
2858 genlmsg_cancel(msg->skb, hdr);
2859
2860 return -EMSGSIZE;
2861}
2862
2863void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2864 struct sk_buff_head *xmitq)
2865{
2866 l->tolerance = tol;
2867 if (l->bc_rcvlink)
2868 l->bc_rcvlink->tolerance = tol;
2869 if (tipc_link_is_up(l))
2870 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2871}
2872
2873void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2874 struct sk_buff_head *xmitq)
2875{
2876 l->priority = prio;
2877 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2878}
2879
2880void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2881{
2882 l->abort_limit = limit;
2883}
2884
2885/**
2886 * tipc_link_dump - dump TIPC link data
2887 * @l: tipc link to be dumped
2888 * @dqueues: bitmask to decide if any link queue to be dumped?
2889 * - TIPC_DUMP_NONE: don't dump link queues
2890 * - TIPC_DUMP_TRANSMQ: dump link transmq queue
2891 * - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2892 * - TIPC_DUMP_DEFERDQ: dump link deferd queue
2893 * - TIPC_DUMP_INPUTQ: dump link input queue
2894 * - TIPC_DUMP_WAKEUP: dump link wakeup queue
2895 * - TIPC_DUMP_ALL: dump all the link queues above
2896 * @buf: returned buffer of dump data in format
2897 */
2898int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2899{
2900 int i = 0;
2901 size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2902 struct sk_buff_head *list;
2903 struct sk_buff *hskb, *tskb;
2904 u32 len;
2905
2906 if (!l) {
2907 i += scnprintf(buf, sz, "link data: (null)\n");
2908 return i;
2909 }
2910
2911 i += scnprintf(buf, sz, "link data: %x", l->addr);
2912 i += scnprintf(buf + i, sz - i, " %x", l->state);
2913 i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2914 i += scnprintf(buf + i, sz - i, " %u", l->session);
2915 i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2916 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2917 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2918 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2919 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2920 i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2921 i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2922 i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2923 i += scnprintf(buf + i, sz - i, " %u", 0);
2924 i += scnprintf(buf + i, sz - i, " %u", 0);
2925 i += scnprintf(buf + i, sz - i, " %u", l->acked);
2926
2927 list = &l->transmq;
2928 len = skb_queue_len(list);
2929 hskb = skb_peek(list);
2930 tskb = skb_peek_tail(list);
2931 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2932 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2933 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2934
2935 list = &l->deferdq;
2936 len = skb_queue_len(list);
2937 hskb = skb_peek(list);
2938 tskb = skb_peek_tail(list);
2939 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2940 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2941 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2942
2943 list = &l->backlogq;
2944 len = skb_queue_len(list);
2945 hskb = skb_peek(list);
2946 tskb = skb_peek_tail(list);
2947 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2948 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2949 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2950
2951 list = l->inputq;
2952 len = skb_queue_len(list);
2953 hskb = skb_peek(list);
2954 tskb = skb_peek_tail(list);
2955 i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2956 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2957 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2958
2959 if (dqueues & TIPC_DUMP_TRANSMQ) {
2960 i += scnprintf(buf + i, sz - i, "transmq: ");
2961 i += tipc_list_dump(&l->transmq, false, buf + i);
2962 }
2963 if (dqueues & TIPC_DUMP_BACKLOGQ) {
2964 i += scnprintf(buf + i, sz - i,
2965 "backlogq: <%u %u %u %u %u>, ",
2966 l->backlog[TIPC_LOW_IMPORTANCE].len,
2967 l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2968 l->backlog[TIPC_HIGH_IMPORTANCE].len,
2969 l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2970 l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2971 i += tipc_list_dump(&l->backlogq, false, buf + i);
2972 }
2973 if (dqueues & TIPC_DUMP_DEFERDQ) {
2974 i += scnprintf(buf + i, sz - i, "deferdq: ");
2975 i += tipc_list_dump(&l->deferdq, false, buf + i);
2976 }
2977 if (dqueues & TIPC_DUMP_INPUTQ) {
2978 i += scnprintf(buf + i, sz - i, "inputq: ");
2979 i += tipc_list_dump(l->inputq, false, buf + i);
2980 }
2981 if (dqueues & TIPC_DUMP_WAKEUP) {
2982 i += scnprintf(buf + i, sz - i, "wakeup: ");
2983 i += tipc_list_dump(&l->wakeupq, false, buf + i);
2984 }
2985
2986 return i;
2987}