]> git.karo-electronics.de Git - karo-tx-linux.git/blob - net/tipc/link.c
Merge branch 'tools' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux...
[karo-tx-linux.git] / net / tipc / link.c
1 /*
2  * net/tipc/link.c: TIPC link code
3  *
4  * Copyright (c) 1996-2007, Ericsson AB
5  * Copyright (c) 2004-2007, 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
81 struct 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
88 static void link_handle_out_of_seq_msg(struct link *l_ptr,
89                                        struct sk_buff *buf);
90 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf);
91 static int  link_recv_changeover_msg(struct link **l_ptr, struct sk_buff **buf);
92 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance);
93 static int  link_send_sections_long(struct port *sender,
94                                     struct iovec const *msg_sect,
95                                     u32 num_sect, u32 destnode);
96 static void link_check_defragm_bufs(struct link *l_ptr);
97 static void link_state_event(struct link *l_ptr, u32 event);
98 static void link_reset_statistics(struct link *l_ptr);
99 static void link_print(struct link *l_ptr, const char *str);
100 static void link_start(struct link *l_ptr);
101 static int link_send_long_buf(struct link *l_ptr, struct sk_buff *buf);
102
103 /*
104  *  Simple link routines
105  */
106
107 static unsigned int align(unsigned int i)
108 {
109         return (i + 3) & ~3u;
110 }
111
112 static void link_init_max_pkt(struct link *l_ptr)
113 {
114         u32 max_pkt;
115
116         max_pkt = (l_ptr->b_ptr->publ.mtu & ~3);
117         if (max_pkt > MAX_MSG_SIZE)
118                 max_pkt = MAX_MSG_SIZE;
119
120         l_ptr->max_pkt_target = max_pkt;
121         if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
122                 l_ptr->max_pkt = l_ptr->max_pkt_target;
123         else
124                 l_ptr->max_pkt = MAX_PKT_DEFAULT;
125
126         l_ptr->max_pkt_probes = 0;
127 }
128
129 static u32 link_next_sent(struct link *l_ptr)
130 {
131         if (l_ptr->next_out)
132                 return msg_seqno(buf_msg(l_ptr->next_out));
133         return mod(l_ptr->next_out_no);
134 }
135
136 static u32 link_last_sent(struct link *l_ptr)
137 {
138         return mod(link_next_sent(l_ptr) - 1);
139 }
140
141 /*
142  *  Simple non-static link routines (i.e. referenced outside this file)
143  */
144
145 int tipc_link_is_up(struct link *l_ptr)
146 {
147         if (!l_ptr)
148                 return 0;
149         return link_working_working(l_ptr) || link_working_unknown(l_ptr);
150 }
151
152 int tipc_link_is_active(struct link *l_ptr)
153 {
154         return  (l_ptr->owner->active_links[0] == l_ptr) ||
155                 (l_ptr->owner->active_links[1] == l_ptr);
156 }
157
158 /**
159  * link_name_validate - validate & (optionally) deconstruct link name
160  * @name - ptr to link name string
161  * @name_parts - ptr to area for link name components (or NULL if not needed)
162  *
163  * Returns 1 if link name is valid, otherwise 0.
164  */
165
166 static int link_name_validate(const char *name, struct link_name *name_parts)
167 {
168         char name_copy[TIPC_MAX_LINK_NAME];
169         char *addr_local;
170         char *if_local;
171         char *addr_peer;
172         char *if_peer;
173         char dummy;
174         u32 z_local, c_local, n_local;
175         u32 z_peer, c_peer, n_peer;
176         u32 if_local_len;
177         u32 if_peer_len;
178
179         /* copy link name & ensure length is OK */
180
181         name_copy[TIPC_MAX_LINK_NAME - 1] = 0;
182         /* need above in case non-Posix strncpy() doesn't pad with nulls */
183         strncpy(name_copy, name, TIPC_MAX_LINK_NAME);
184         if (name_copy[TIPC_MAX_LINK_NAME - 1] != 0)
185                 return 0;
186
187         /* ensure all component parts of link name are present */
188
189         addr_local = name_copy;
190         if_local = strchr(addr_local, ':');
191         if (if_local == NULL)
192                 return 0;
193         *(if_local++) = 0;
194         addr_peer = strchr(if_local, '-');
195         if (addr_peer == NULL)
196                 return 0;
197         *(addr_peer++) = 0;
198         if_local_len = addr_peer - if_local;
199         if_peer = strchr(addr_peer, ':');
200         if (if_peer == NULL)
201                 return 0;
202         *(if_peer++) = 0;
203         if_peer_len = strlen(if_peer) + 1;
204
205         /* validate component parts of link name */
206
207         if ((sscanf(addr_local, "%u.%u.%u%c",
208                     &z_local, &c_local, &n_local, &dummy) != 3) ||
209             (sscanf(addr_peer, "%u.%u.%u%c",
210                     &z_peer, &c_peer, &n_peer, &dummy) != 3) ||
211             (z_local > 255) || (c_local > 4095) || (n_local > 4095) ||
212             (z_peer  > 255) || (c_peer  > 4095) || (n_peer  > 4095) ||
213             (if_local_len <= 1) || (if_local_len > TIPC_MAX_IF_NAME) ||
214             (if_peer_len  <= 1) || (if_peer_len  > TIPC_MAX_IF_NAME) ||
215             (strspn(if_local, tipc_alphabet) != (if_local_len - 1)) ||
216             (strspn(if_peer, tipc_alphabet) != (if_peer_len - 1)))
217                 return 0;
218
219         /* return link name components, if necessary */
220
221         if (name_parts) {
222                 name_parts->addr_local = tipc_addr(z_local, c_local, n_local);
223                 strcpy(name_parts->if_local, if_local);
224                 name_parts->addr_peer = tipc_addr(z_peer, c_peer, n_peer);
225                 strcpy(name_parts->if_peer, if_peer);
226         }
227         return 1;
228 }
229
230 /**
231  * link_timeout - handle expiration of link timer
232  * @l_ptr: pointer to link
233  *
234  * This routine must not grab "tipc_net_lock" to avoid a potential deadlock conflict
235  * with tipc_link_delete().  (There is no risk that the node will be deleted by
236  * another thread because tipc_link_delete() always cancels the link timer before
237  * tipc_node_delete() is called.)
238  */
239
240 static void link_timeout(struct link *l_ptr)
241 {
242         tipc_node_lock(l_ptr->owner);
243
244         /* update counters used in statistical profiling of send traffic */
245
246         l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
247         l_ptr->stats.queue_sz_counts++;
248
249         if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
250                 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
251
252         if (l_ptr->first_out) {
253                 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
254                 u32 length = msg_size(msg);
255
256                 if ((msg_user(msg) == MSG_FRAGMENTER) &&
257                     (msg_type(msg) == FIRST_FRAGMENT)) {
258                         length = msg_size(msg_get_wrapped(msg));
259                 }
260                 if (length) {
261                         l_ptr->stats.msg_lengths_total += length;
262                         l_ptr->stats.msg_length_counts++;
263                         if (length <= 64)
264                                 l_ptr->stats.msg_length_profile[0]++;
265                         else if (length <= 256)
266                                 l_ptr->stats.msg_length_profile[1]++;
267                         else if (length <= 1024)
268                                 l_ptr->stats.msg_length_profile[2]++;
269                         else if (length <= 4096)
270                                 l_ptr->stats.msg_length_profile[3]++;
271                         else if (length <= 16384)
272                                 l_ptr->stats.msg_length_profile[4]++;
273                         else if (length <= 32768)
274                                 l_ptr->stats.msg_length_profile[5]++;
275                         else
276                                 l_ptr->stats.msg_length_profile[6]++;
277                 }
278         }
279
280         /* do all other link processing performed on a periodic basis */
281
282         link_check_defragm_bufs(l_ptr);
283
284         link_state_event(l_ptr, TIMEOUT_EVT);
285
286         if (l_ptr->next_out)
287                 tipc_link_push_queue(l_ptr);
288
289         tipc_node_unlock(l_ptr->owner);
290 }
291
292 static void link_set_timer(struct link *l_ptr, u32 time)
293 {
294         k_start_timer(&l_ptr->timer, time);
295 }
296
297 /**
298  * tipc_link_create - create a new link
299  * @b_ptr: pointer to associated bearer
300  * @peer: network address of node at other end of link
301  * @media_addr: media address to use when sending messages over link
302  *
303  * Returns pointer to link.
304  */
305
306 struct link *tipc_link_create(struct bearer *b_ptr, const u32 peer,
307                               const struct tipc_media_addr *media_addr)
308 {
309         struct link *l_ptr;
310         struct tipc_msg *msg;
311         char *if_name;
312
313         l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
314         if (!l_ptr) {
315                 warn("Link creation failed, no memory\n");
316                 return NULL;
317         }
318
319         l_ptr->addr = peer;
320         if_name = strchr(b_ptr->publ.name, ':') + 1;
321         sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:",
322                 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
323                 tipc_node(tipc_own_addr),
324                 if_name,
325                 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
326                 /* note: peer i/f is appended to link name by reset/activate */
327         memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
328         l_ptr->checkpoint = 1;
329         l_ptr->b_ptr = b_ptr;
330         link_set_supervision_props(l_ptr, b_ptr->media->tolerance);
331         l_ptr->state = RESET_UNKNOWN;
332
333         l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
334         msg = l_ptr->pmsg;
335         tipc_msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
336         msg_set_size(msg, sizeof(l_ptr->proto_msg));
337         msg_set_session(msg, (tipc_random & 0xffff));
338         msg_set_bearer_id(msg, b_ptr->identity);
339         strcpy((char *)msg_data(msg), if_name);
340
341         l_ptr->priority = b_ptr->priority;
342         tipc_link_set_queue_limits(l_ptr, b_ptr->media->window);
343
344         link_init_max_pkt(l_ptr);
345
346         l_ptr->next_out_no = 1;
347         INIT_LIST_HEAD(&l_ptr->waiting_ports);
348
349         link_reset_statistics(l_ptr);
350
351         l_ptr->owner = tipc_node_attach_link(l_ptr);
352         if (!l_ptr->owner) {
353                 kfree(l_ptr);
354                 return NULL;
355         }
356
357         k_init_timer(&l_ptr->timer, (Handler)link_timeout, (unsigned long)l_ptr);
358         list_add_tail(&l_ptr->link_list, &b_ptr->links);
359         tipc_k_signal((Handler)link_start, (unsigned long)l_ptr);
360
361         return l_ptr;
362 }
363
364 /**
365  * tipc_link_delete - delete a link
366  * @l_ptr: pointer to link
367  *
368  * Note: 'tipc_net_lock' is write_locked, bearer is locked.
369  * This routine must not grab the node lock until after link timer cancellation
370  * to avoid a potential deadlock situation.
371  */
372
373 void tipc_link_delete(struct link *l_ptr)
374 {
375         if (!l_ptr) {
376                 err("Attempt to delete non-existent link\n");
377                 return;
378         }
379
380         k_cancel_timer(&l_ptr->timer);
381
382         tipc_node_lock(l_ptr->owner);
383         tipc_link_reset(l_ptr);
384         tipc_node_detach_link(l_ptr->owner, l_ptr);
385         tipc_link_stop(l_ptr);
386         list_del_init(&l_ptr->link_list);
387         tipc_node_unlock(l_ptr->owner);
388         k_term_timer(&l_ptr->timer);
389         kfree(l_ptr);
390 }
391
392 static void link_start(struct link *l_ptr)
393 {
394         link_state_event(l_ptr, STARTING_EVT);
395 }
396
397 /**
398  * link_schedule_port - schedule port for deferred sending
399  * @l_ptr: pointer to link
400  * @origport: reference to sending port
401  * @sz: amount of data to be sent
402  *
403  * Schedules port for renewed sending of messages after link congestion
404  * has abated.
405  */
406
407 static int link_schedule_port(struct link *l_ptr, u32 origport, u32 sz)
408 {
409         struct port *p_ptr;
410
411         spin_lock_bh(&tipc_port_list_lock);
412         p_ptr = tipc_port_lock(origport);
413         if (p_ptr) {
414                 if (!p_ptr->wakeup)
415                         goto exit;
416                 if (!list_empty(&p_ptr->wait_list))
417                         goto exit;
418                 p_ptr->publ.congested = 1;
419                 p_ptr->waiting_pkts = 1 + ((sz - 1) / l_ptr->max_pkt);
420                 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
421                 l_ptr->stats.link_congs++;
422 exit:
423                 tipc_port_unlock(p_ptr);
424         }
425         spin_unlock_bh(&tipc_port_list_lock);
426         return -ELINKCONG;
427 }
428
429 void tipc_link_wakeup_ports(struct link *l_ptr, int all)
430 {
431         struct port *p_ptr;
432         struct port *temp_p_ptr;
433         int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
434
435         if (all)
436                 win = 100000;
437         if (win <= 0)
438                 return;
439         if (!spin_trylock_bh(&tipc_port_list_lock))
440                 return;
441         if (link_congested(l_ptr))
442                 goto exit;
443         list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
444                                  wait_list) {
445                 if (win <= 0)
446                         break;
447                 list_del_init(&p_ptr->wait_list);
448                 spin_lock_bh(p_ptr->publ.lock);
449                 p_ptr->publ.congested = 0;
450                 p_ptr->wakeup(&p_ptr->publ);
451                 win -= p_ptr->waiting_pkts;
452                 spin_unlock_bh(p_ptr->publ.lock);
453         }
454
455 exit:
456         spin_unlock_bh(&tipc_port_list_lock);
457 }
458
459 /**
460  * link_release_outqueue - purge link's outbound message queue
461  * @l_ptr: pointer to link
462  */
463
464 static void link_release_outqueue(struct link *l_ptr)
465 {
466         struct sk_buff *buf = l_ptr->first_out;
467         struct sk_buff *next;
468
469         while (buf) {
470                 next = buf->next;
471                 buf_discard(buf);
472                 buf = next;
473         }
474         l_ptr->first_out = NULL;
475         l_ptr->out_queue_size = 0;
476 }
477
478 /**
479  * tipc_link_reset_fragments - purge link's inbound message fragments queue
480  * @l_ptr: pointer to link
481  */
482
483 void tipc_link_reset_fragments(struct link *l_ptr)
484 {
485         struct sk_buff *buf = l_ptr->defragm_buf;
486         struct sk_buff *next;
487
488         while (buf) {
489                 next = buf->next;
490                 buf_discard(buf);
491                 buf = next;
492         }
493         l_ptr->defragm_buf = NULL;
494 }
495
496 /**
497  * tipc_link_stop - purge all inbound and outbound messages associated with link
498  * @l_ptr: pointer to link
499  */
500
501 void tipc_link_stop(struct link *l_ptr)
502 {
503         struct sk_buff *buf;
504         struct sk_buff *next;
505
506         buf = l_ptr->oldest_deferred_in;
507         while (buf) {
508                 next = buf->next;
509                 buf_discard(buf);
510                 buf = next;
511         }
512
513         buf = l_ptr->first_out;
514         while (buf) {
515                 next = buf->next;
516                 buf_discard(buf);
517                 buf = next;
518         }
519
520         tipc_link_reset_fragments(l_ptr);
521
522         buf_discard(l_ptr->proto_msg_queue);
523         l_ptr->proto_msg_queue = NULL;
524 }
525
526 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
527 #define link_send_event(fcn, l_ptr, up) do { } while (0)
528
529 void tipc_link_reset(struct link *l_ptr)
530 {
531         struct sk_buff *buf;
532         u32 prev_state = l_ptr->state;
533         u32 checkpoint = l_ptr->next_in_no;
534         int was_active_link = tipc_link_is_active(l_ptr);
535
536         msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
537
538         /* Link is down, accept any session */
539         l_ptr->peer_session = INVALID_SESSION;
540
541         /* Prepare for max packet size negotiation */
542         link_init_max_pkt(l_ptr);
543
544         l_ptr->state = RESET_UNKNOWN;
545
546         if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
547                 return;
548
549         tipc_node_link_down(l_ptr->owner, l_ptr);
550         tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
551
552         if (was_active_link && tipc_node_has_active_links(l_ptr->owner) &&
553             l_ptr->owner->permit_changeover) {
554                 l_ptr->reset_checkpoint = checkpoint;
555                 l_ptr->exp_msg_count = START_CHANGEOVER;
556         }
557
558         /* Clean up all queues: */
559
560         link_release_outqueue(l_ptr);
561         buf_discard(l_ptr->proto_msg_queue);
562         l_ptr->proto_msg_queue = NULL;
563         buf = l_ptr->oldest_deferred_in;
564         while (buf) {
565                 struct sk_buff *next = buf->next;
566                 buf_discard(buf);
567                 buf = next;
568         }
569         if (!list_empty(&l_ptr->waiting_ports))
570                 tipc_link_wakeup_ports(l_ptr, 1);
571
572         l_ptr->retransm_queue_head = 0;
573         l_ptr->retransm_queue_size = 0;
574         l_ptr->last_out = NULL;
575         l_ptr->first_out = NULL;
576         l_ptr->next_out = NULL;
577         l_ptr->unacked_window = 0;
578         l_ptr->checkpoint = 1;
579         l_ptr->next_out_no = 1;
580         l_ptr->deferred_inqueue_sz = 0;
581         l_ptr->oldest_deferred_in = NULL;
582         l_ptr->newest_deferred_in = NULL;
583         l_ptr->fsm_msg_cnt = 0;
584         l_ptr->stale_count = 0;
585         link_reset_statistics(l_ptr);
586
587         link_send_event(tipc_cfg_link_event, l_ptr, 0);
588         if (!in_own_cluster(l_ptr->addr))
589                 link_send_event(tipc_disc_link_event, l_ptr, 0);
590 }
591
592
593 static void link_activate(struct link *l_ptr)
594 {
595         l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
596         tipc_node_link_up(l_ptr->owner, l_ptr);
597         tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
598         link_send_event(tipc_cfg_link_event, l_ptr, 1);
599         if (!in_own_cluster(l_ptr->addr))
600                 link_send_event(tipc_disc_link_event, l_ptr, 1);
601 }
602
603 /**
604  * link_state_event - link finite state machine
605  * @l_ptr: pointer to link
606  * @event: state machine event to process
607  */
608
609 static void link_state_event(struct link *l_ptr, unsigned event)
610 {
611         struct link *other;
612         u32 cont_intv = l_ptr->continuity_interval;
613
614         if (!l_ptr->started && (event != STARTING_EVT))
615                 return;         /* Not yet. */
616
617         if (link_blocked(l_ptr)) {
618                 if (event == TIMEOUT_EVT)
619                         link_set_timer(l_ptr, cont_intv);
620                 return;   /* Changeover going on */
621         }
622
623         switch (l_ptr->state) {
624         case WORKING_WORKING:
625                 switch (event) {
626                 case TRAFFIC_MSG_EVT:
627                 case ACTIVATE_MSG:
628                         break;
629                 case TIMEOUT_EVT:
630                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
631                                 l_ptr->checkpoint = l_ptr->next_in_no;
632                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
633                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
634                                                                  0, 0, 0, 0, 0);
635                                         l_ptr->fsm_msg_cnt++;
636                                 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
637                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
638                                                                  1, 0, 0, 0, 0);
639                                         l_ptr->fsm_msg_cnt++;
640                                 }
641                                 link_set_timer(l_ptr, cont_intv);
642                                 break;
643                         }
644                         l_ptr->state = WORKING_UNKNOWN;
645                         l_ptr->fsm_msg_cnt = 0;
646                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
647                         l_ptr->fsm_msg_cnt++;
648                         link_set_timer(l_ptr, cont_intv / 4);
649                         break;
650                 case RESET_MSG:
651                         info("Resetting link <%s>, requested by peer\n",
652                              l_ptr->name);
653                         tipc_link_reset(l_ptr);
654                         l_ptr->state = RESET_RESET;
655                         l_ptr->fsm_msg_cnt = 0;
656                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
657                         l_ptr->fsm_msg_cnt++;
658                         link_set_timer(l_ptr, cont_intv);
659                         break;
660                 default:
661                         err("Unknown link event %u in WW state\n", event);
662                 }
663                 break;
664         case WORKING_UNKNOWN:
665                 switch (event) {
666                 case TRAFFIC_MSG_EVT:
667                 case ACTIVATE_MSG:
668                         l_ptr->state = WORKING_WORKING;
669                         l_ptr->fsm_msg_cnt = 0;
670                         link_set_timer(l_ptr, cont_intv);
671                         break;
672                 case RESET_MSG:
673                         info("Resetting link <%s>, requested by peer "
674                              "while probing\n", l_ptr->name);
675                         tipc_link_reset(l_ptr);
676                         l_ptr->state = RESET_RESET;
677                         l_ptr->fsm_msg_cnt = 0;
678                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
679                         l_ptr->fsm_msg_cnt++;
680                         link_set_timer(l_ptr, cont_intv);
681                         break;
682                 case TIMEOUT_EVT:
683                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
684                                 l_ptr->state = WORKING_WORKING;
685                                 l_ptr->fsm_msg_cnt = 0;
686                                 l_ptr->checkpoint = l_ptr->next_in_no;
687                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
688                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
689                                                                  0, 0, 0, 0, 0);
690                                         l_ptr->fsm_msg_cnt++;
691                                 }
692                                 link_set_timer(l_ptr, cont_intv);
693                         } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
694                                 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
695                                                          1, 0, 0, 0, 0);
696                                 l_ptr->fsm_msg_cnt++;
697                                 link_set_timer(l_ptr, cont_intv / 4);
698                         } else {        /* Link has failed */
699                                 warn("Resetting link <%s>, peer not responding\n",
700                                      l_ptr->name);
701                                 tipc_link_reset(l_ptr);
702                                 l_ptr->state = RESET_UNKNOWN;
703                                 l_ptr->fsm_msg_cnt = 0;
704                                 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
705                                                          0, 0, 0, 0, 0);
706                                 l_ptr->fsm_msg_cnt++;
707                                 link_set_timer(l_ptr, cont_intv);
708                         }
709                         break;
710                 default:
711                         err("Unknown link event %u in WU state\n", event);
712                 }
713                 break;
714         case RESET_UNKNOWN:
715                 switch (event) {
716                 case TRAFFIC_MSG_EVT:
717                         break;
718                 case ACTIVATE_MSG:
719                         other = l_ptr->owner->active_links[0];
720                         if (other && link_working_unknown(other))
721                                 break;
722                         l_ptr->state = WORKING_WORKING;
723                         l_ptr->fsm_msg_cnt = 0;
724                         link_activate(l_ptr);
725                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
726                         l_ptr->fsm_msg_cnt++;
727                         link_set_timer(l_ptr, cont_intv);
728                         break;
729                 case RESET_MSG:
730                         l_ptr->state = RESET_RESET;
731                         l_ptr->fsm_msg_cnt = 0;
732                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
733                         l_ptr->fsm_msg_cnt++;
734                         link_set_timer(l_ptr, cont_intv);
735                         break;
736                 case STARTING_EVT:
737                         l_ptr->started = 1;
738                         /* fall through */
739                 case TIMEOUT_EVT:
740                         tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
741                         l_ptr->fsm_msg_cnt++;
742                         link_set_timer(l_ptr, cont_intv);
743                         break;
744                 default:
745                         err("Unknown link event %u in RU state\n", event);
746                 }
747                 break;
748         case RESET_RESET:
749                 switch (event) {
750                 case TRAFFIC_MSG_EVT:
751                 case ACTIVATE_MSG:
752                         other = l_ptr->owner->active_links[0];
753                         if (other && link_working_unknown(other))
754                                 break;
755                         l_ptr->state = WORKING_WORKING;
756                         l_ptr->fsm_msg_cnt = 0;
757                         link_activate(l_ptr);
758                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
759                         l_ptr->fsm_msg_cnt++;
760                         link_set_timer(l_ptr, cont_intv);
761                         break;
762                 case RESET_MSG:
763                         break;
764                 case TIMEOUT_EVT:
765                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
766                         l_ptr->fsm_msg_cnt++;
767                         link_set_timer(l_ptr, cont_intv);
768                         break;
769                 default:
770                         err("Unknown link event %u in RR state\n", event);
771                 }
772                 break;
773         default:
774                 err("Unknown link state %u/%u\n", l_ptr->state, event);
775         }
776 }
777
778 /*
779  * link_bundle_buf(): Append contents of a buffer to
780  * the tail of an existing one.
781  */
782
783 static int link_bundle_buf(struct link *l_ptr,
784                            struct sk_buff *bundler,
785                            struct sk_buff *buf)
786 {
787         struct tipc_msg *bundler_msg = buf_msg(bundler);
788         struct tipc_msg *msg = buf_msg(buf);
789         u32 size = msg_size(msg);
790         u32 bundle_size = msg_size(bundler_msg);
791         u32 to_pos = align(bundle_size);
792         u32 pad = to_pos - bundle_size;
793
794         if (msg_user(bundler_msg) != MSG_BUNDLER)
795                 return 0;
796         if (msg_type(bundler_msg) != OPEN_MSG)
797                 return 0;
798         if (skb_tailroom(bundler) < (pad + size))
799                 return 0;
800         if (l_ptr->max_pkt < (to_pos + size))
801                 return 0;
802
803         skb_put(bundler, pad + size);
804         skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
805         msg_set_size(bundler_msg, to_pos + size);
806         msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
807         buf_discard(buf);
808         l_ptr->stats.sent_bundled++;
809         return 1;
810 }
811
812 static void link_add_to_outqueue(struct link *l_ptr,
813                                  struct sk_buff *buf,
814                                  struct tipc_msg *msg)
815 {
816         u32 ack = mod(l_ptr->next_in_no - 1);
817         u32 seqno = mod(l_ptr->next_out_no++);
818
819         msg_set_word(msg, 2, ((ack << 16) | seqno));
820         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
821         buf->next = NULL;
822         if (l_ptr->first_out) {
823                 l_ptr->last_out->next = buf;
824                 l_ptr->last_out = buf;
825         } else
826                 l_ptr->first_out = l_ptr->last_out = buf;
827         l_ptr->out_queue_size++;
828 }
829
830 /*
831  * tipc_link_send_buf() is the 'full path' for messages, called from
832  * inside TIPC when the 'fast path' in tipc_send_buf
833  * has failed, and from link_send()
834  */
835
836 int tipc_link_send_buf(struct link *l_ptr, struct sk_buff *buf)
837 {
838         struct tipc_msg *msg = buf_msg(buf);
839         u32 size = msg_size(msg);
840         u32 dsz = msg_data_sz(msg);
841         u32 queue_size = l_ptr->out_queue_size;
842         u32 imp = tipc_msg_tot_importance(msg);
843         u32 queue_limit = l_ptr->queue_limit[imp];
844         u32 max_packet = l_ptr->max_pkt;
845
846         msg_set_prevnode(msg, tipc_own_addr);   /* If routed message */
847
848         /* Match msg importance against queue limits: */
849
850         if (unlikely(queue_size >= queue_limit)) {
851                 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
852                         return link_schedule_port(l_ptr, msg_origport(msg),
853                                                   size);
854                 }
855                 buf_discard(buf);
856                 if (imp > CONN_MANAGER) {
857                         warn("Resetting link <%s>, send queue full", l_ptr->name);
858                         tipc_link_reset(l_ptr);
859                 }
860                 return dsz;
861         }
862
863         /* Fragmentation needed ? */
864
865         if (size > max_packet)
866                 return link_send_long_buf(l_ptr, buf);
867
868         /* Packet can be queued or sent: */
869
870         if (queue_size > l_ptr->stats.max_queue_sz)
871                 l_ptr->stats.max_queue_sz = queue_size;
872
873         if (likely(!tipc_bearer_congested(l_ptr->b_ptr, l_ptr) &&
874                    !link_congested(l_ptr))) {
875                 link_add_to_outqueue(l_ptr, buf, msg);
876
877                 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr))) {
878                         l_ptr->unacked_window = 0;
879                 } else {
880                         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
881                         l_ptr->stats.bearer_congs++;
882                         l_ptr->next_out = buf;
883                 }
884                 return dsz;
885         }
886         /* Congestion: can message be bundled ?: */
887
888         if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
889             (msg_user(msg) != MSG_FRAGMENTER)) {
890
891                 /* Try adding message to an existing bundle */
892
893                 if (l_ptr->next_out &&
894                     link_bundle_buf(l_ptr, l_ptr->last_out, buf)) {
895                         tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
896                         return dsz;
897                 }
898
899                 /* Try creating a new bundle */
900
901                 if (size <= max_packet * 2 / 3) {
902                         struct sk_buff *bundler = tipc_buf_acquire(max_packet);
903                         struct tipc_msg bundler_hdr;
904
905                         if (bundler) {
906                                 tipc_msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
907                                          INT_H_SIZE, l_ptr->addr);
908                                 skb_copy_to_linear_data(bundler, &bundler_hdr,
909                                                         INT_H_SIZE);
910                                 skb_trim(bundler, INT_H_SIZE);
911                                 link_bundle_buf(l_ptr, bundler, buf);
912                                 buf = bundler;
913                                 msg = buf_msg(buf);
914                                 l_ptr->stats.sent_bundles++;
915                         }
916                 }
917         }
918         if (!l_ptr->next_out)
919                 l_ptr->next_out = buf;
920         link_add_to_outqueue(l_ptr, buf, msg);
921         tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
922         return dsz;
923 }
924
925 /*
926  * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
927  * not been selected yet, and the the owner node is not locked
928  * Called by TIPC internal users, e.g. the name distributor
929  */
930
931 int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
932 {
933         struct link *l_ptr;
934         struct tipc_node *n_ptr;
935         int res = -ELINKCONG;
936
937         read_lock_bh(&tipc_net_lock);
938         n_ptr = tipc_node_find(dest);
939         if (n_ptr) {
940                 tipc_node_lock(n_ptr);
941                 l_ptr = n_ptr->active_links[selector & 1];
942                 if (l_ptr)
943                         res = tipc_link_send_buf(l_ptr, buf);
944                 else
945                         buf_discard(buf);
946                 tipc_node_unlock(n_ptr);
947         } else {
948                 buf_discard(buf);
949         }
950         read_unlock_bh(&tipc_net_lock);
951         return res;
952 }
953
954 /*
955  * link_send_buf_fast: Entry for data messages where the
956  * destination link is known and the header is complete,
957  * inclusive total message length. Very time critical.
958  * Link is locked. Returns user data length.
959  */
960
961 static int link_send_buf_fast(struct link *l_ptr, struct sk_buff *buf,
962                               u32 *used_max_pkt)
963 {
964         struct tipc_msg *msg = buf_msg(buf);
965         int res = msg_data_sz(msg);
966
967         if (likely(!link_congested(l_ptr))) {
968                 if (likely(msg_size(msg) <= l_ptr->max_pkt)) {
969                         if (likely(list_empty(&l_ptr->b_ptr->cong_links))) {
970                                 link_add_to_outqueue(l_ptr, buf, msg);
971                                 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf,
972                                                             &l_ptr->media_addr))) {
973                                         l_ptr->unacked_window = 0;
974                                         return res;
975                                 }
976                                 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
977                                 l_ptr->stats.bearer_congs++;
978                                 l_ptr->next_out = buf;
979                                 return res;
980                         }
981                 } else
982                         *used_max_pkt = l_ptr->max_pkt;
983         }
984         return tipc_link_send_buf(l_ptr, buf);  /* All other cases */
985 }
986
987 /*
988  * tipc_send_buf_fast: Entry for data messages where the
989  * destination node is known and the header is complete,
990  * inclusive total message length.
991  * Returns user data length.
992  */
993 int tipc_send_buf_fast(struct sk_buff *buf, u32 destnode)
994 {
995         struct link *l_ptr;
996         struct tipc_node *n_ptr;
997         int res;
998         u32 selector = msg_origport(buf_msg(buf)) & 1;
999         u32 dummy;
1000
1001         if (destnode == tipc_own_addr)
1002                 return tipc_port_recv_msg(buf);
1003
1004         read_lock_bh(&tipc_net_lock);
1005         n_ptr = tipc_node_find(destnode);
1006         if (likely(n_ptr)) {
1007                 tipc_node_lock(n_ptr);
1008                 l_ptr = n_ptr->active_links[selector];
1009                 if (likely(l_ptr)) {
1010                         res = link_send_buf_fast(l_ptr, buf, &dummy);
1011                         tipc_node_unlock(n_ptr);
1012                         read_unlock_bh(&tipc_net_lock);
1013                         return res;
1014                 }
1015                 tipc_node_unlock(n_ptr);
1016         }
1017         read_unlock_bh(&tipc_net_lock);
1018         res = msg_data_sz(buf_msg(buf));
1019         tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1020         return res;
1021 }
1022
1023
1024 /*
1025  * tipc_link_send_sections_fast: Entry for messages where the
1026  * destination processor is known and the header is complete,
1027  * except for total message length.
1028  * Returns user data length or errno.
1029  */
1030 int tipc_link_send_sections_fast(struct port *sender,
1031                                  struct iovec const *msg_sect,
1032                                  const u32 num_sect,
1033                                  u32 destaddr)
1034 {
1035         struct tipc_msg *hdr = &sender->publ.phdr;
1036         struct link *l_ptr;
1037         struct sk_buff *buf;
1038         struct tipc_node *node;
1039         int res;
1040         u32 selector = msg_origport(hdr) & 1;
1041
1042 again:
1043         /*
1044          * Try building message using port's max_pkt hint.
1045          * (Must not hold any locks while building message.)
1046          */
1047
1048         res = tipc_msg_build(hdr, msg_sect, num_sect, sender->publ.max_pkt,
1049                         !sender->user_port, &buf);
1050
1051         read_lock_bh(&tipc_net_lock);
1052         node = tipc_node_find(destaddr);
1053         if (likely(node)) {
1054                 tipc_node_lock(node);
1055                 l_ptr = node->active_links[selector];
1056                 if (likely(l_ptr)) {
1057                         if (likely(buf)) {
1058                                 res = link_send_buf_fast(l_ptr, buf,
1059                                                          &sender->publ.max_pkt);
1060                                 if (unlikely(res < 0))
1061                                         buf_discard(buf);
1062 exit:
1063                                 tipc_node_unlock(node);
1064                                 read_unlock_bh(&tipc_net_lock);
1065                                 return res;
1066                         }
1067
1068                         /* Exit if build request was invalid */
1069
1070                         if (unlikely(res < 0))
1071                                 goto exit;
1072
1073                         /* Exit if link (or bearer) is congested */
1074
1075                         if (link_congested(l_ptr) ||
1076                             !list_empty(&l_ptr->b_ptr->cong_links)) {
1077                                 res = link_schedule_port(l_ptr,
1078                                                          sender->publ.ref, res);
1079                                 goto exit;
1080                         }
1081
1082                         /*
1083                          * Message size exceeds max_pkt hint; update hint,
1084                          * then re-try fast path or fragment the message
1085                          */
1086
1087                         sender->publ.max_pkt = l_ptr->max_pkt;
1088                         tipc_node_unlock(node);
1089                         read_unlock_bh(&tipc_net_lock);
1090
1091
1092                         if ((msg_hdr_sz(hdr) + res) <= sender->publ.max_pkt)
1093                                 goto again;
1094
1095                         return link_send_sections_long(sender, msg_sect,
1096                                                        num_sect, destaddr);
1097                 }
1098                 tipc_node_unlock(node);
1099         }
1100         read_unlock_bh(&tipc_net_lock);
1101
1102         /* Couldn't find a link to the destination node */
1103
1104         if (buf)
1105                 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1106         if (res >= 0)
1107                 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1108                                                  TIPC_ERR_NO_NODE);
1109         return res;
1110 }
1111
1112 /*
1113  * link_send_sections_long(): Entry for long messages where the
1114  * destination node is known and the header is complete,
1115  * inclusive total message length.
1116  * Link and bearer congestion status have been checked to be ok,
1117  * and are ignored if they change.
1118  *
1119  * Note that fragments do not use the full link MTU so that they won't have
1120  * to undergo refragmentation if link changeover causes them to be sent
1121  * over another link with an additional tunnel header added as prefix.
1122  * (Refragmentation will still occur if the other link has a smaller MTU.)
1123  *
1124  * Returns user data length or errno.
1125  */
1126 static int link_send_sections_long(struct port *sender,
1127                                    struct iovec const *msg_sect,
1128                                    u32 num_sect,
1129                                    u32 destaddr)
1130 {
1131         struct link *l_ptr;
1132         struct tipc_node *node;
1133         struct tipc_msg *hdr = &sender->publ.phdr;
1134         u32 dsz = msg_data_sz(hdr);
1135         u32 max_pkt, fragm_sz, rest;
1136         struct tipc_msg fragm_hdr;
1137         struct sk_buff *buf, *buf_chain, *prev;
1138         u32 fragm_crs, fragm_rest, hsz, sect_rest;
1139         const unchar *sect_crs;
1140         int curr_sect;
1141         u32 fragm_no;
1142
1143 again:
1144         fragm_no = 1;
1145         max_pkt = sender->publ.max_pkt - INT_H_SIZE;
1146                 /* leave room for tunnel header in case of link changeover */
1147         fragm_sz = max_pkt - INT_H_SIZE;
1148                 /* leave room for fragmentation header in each fragment */
1149         rest = dsz;
1150         fragm_crs = 0;
1151         fragm_rest = 0;
1152         sect_rest = 0;
1153         sect_crs = NULL;
1154         curr_sect = -1;
1155
1156         /* Prepare reusable fragment header: */
1157
1158         tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1159                  INT_H_SIZE, msg_destnode(hdr));
1160         msg_set_link_selector(&fragm_hdr, sender->publ.ref);
1161         msg_set_size(&fragm_hdr, max_pkt);
1162         msg_set_fragm_no(&fragm_hdr, 1);
1163
1164         /* Prepare header of first fragment: */
1165
1166         buf_chain = buf = tipc_buf_acquire(max_pkt);
1167         if (!buf)
1168                 return -ENOMEM;
1169         buf->next = NULL;
1170         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1171         hsz = msg_hdr_sz(hdr);
1172         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1173
1174         /* Chop up message: */
1175
1176         fragm_crs = INT_H_SIZE + hsz;
1177         fragm_rest = fragm_sz - hsz;
1178
1179         do {            /* For all sections */
1180                 u32 sz;
1181
1182                 if (!sect_rest) {
1183                         sect_rest = msg_sect[++curr_sect].iov_len;
1184                         sect_crs = (const unchar *)msg_sect[curr_sect].iov_base;
1185                 }
1186
1187                 if (sect_rest < fragm_rest)
1188                         sz = sect_rest;
1189                 else
1190                         sz = fragm_rest;
1191
1192                 if (likely(!sender->user_port)) {
1193                         if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1194 error:
1195                                 for (; buf_chain; buf_chain = buf) {
1196                                         buf = buf_chain->next;
1197                                         buf_discard(buf_chain);
1198                                 }
1199                                 return -EFAULT;
1200                         }
1201                 } else
1202                         skb_copy_to_linear_data_offset(buf, fragm_crs,
1203                                                        sect_crs, sz);
1204                 sect_crs += sz;
1205                 sect_rest -= sz;
1206                 fragm_crs += sz;
1207                 fragm_rest -= sz;
1208                 rest -= sz;
1209
1210                 if (!fragm_rest && rest) {
1211
1212                         /* Initiate new fragment: */
1213                         if (rest <= fragm_sz) {
1214                                 fragm_sz = rest;
1215                                 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
1216                         } else {
1217                                 msg_set_type(&fragm_hdr, FRAGMENT);
1218                         }
1219                         msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1220                         msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1221                         prev = buf;
1222                         buf = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
1223                         if (!buf)
1224                                 goto error;
1225
1226                         buf->next = NULL;
1227                         prev->next = buf;
1228                         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1229                         fragm_crs = INT_H_SIZE;
1230                         fragm_rest = fragm_sz;
1231                 }
1232         } while (rest > 0);
1233
1234         /*
1235          * Now we have a buffer chain. Select a link and check
1236          * that packet size is still OK
1237          */
1238         node = tipc_node_find(destaddr);
1239         if (likely(node)) {
1240                 tipc_node_lock(node);
1241                 l_ptr = node->active_links[sender->publ.ref & 1];
1242                 if (!l_ptr) {
1243                         tipc_node_unlock(node);
1244                         goto reject;
1245                 }
1246                 if (l_ptr->max_pkt < max_pkt) {
1247                         sender->publ.max_pkt = l_ptr->max_pkt;
1248                         tipc_node_unlock(node);
1249                         for (; buf_chain; buf_chain = buf) {
1250                                 buf = buf_chain->next;
1251                                 buf_discard(buf_chain);
1252                         }
1253                         goto again;
1254                 }
1255         } else {
1256 reject:
1257                 for (; buf_chain; buf_chain = buf) {
1258                         buf = buf_chain->next;
1259                         buf_discard(buf_chain);
1260                 }
1261                 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1262                                                  TIPC_ERR_NO_NODE);
1263         }
1264
1265         /* Append whole chain to send queue: */
1266
1267         buf = buf_chain;
1268         l_ptr->long_msg_seq_no = mod(l_ptr->long_msg_seq_no + 1);
1269         if (!l_ptr->next_out)
1270                 l_ptr->next_out = buf_chain;
1271         l_ptr->stats.sent_fragmented++;
1272         while (buf) {
1273                 struct sk_buff *next = buf->next;
1274                 struct tipc_msg *msg = buf_msg(buf);
1275
1276                 l_ptr->stats.sent_fragments++;
1277                 msg_set_long_msgno(msg, l_ptr->long_msg_seq_no);
1278                 link_add_to_outqueue(l_ptr, buf, msg);
1279                 buf = next;
1280         }
1281
1282         /* Send it, if possible: */
1283
1284         tipc_link_push_queue(l_ptr);
1285         tipc_node_unlock(node);
1286         return dsz;
1287 }
1288
1289 /*
1290  * tipc_link_push_packet: Push one unsent packet to the media
1291  */
1292 u32 tipc_link_push_packet(struct link *l_ptr)
1293 {
1294         struct sk_buff *buf = l_ptr->first_out;
1295         u32 r_q_size = l_ptr->retransm_queue_size;
1296         u32 r_q_head = l_ptr->retransm_queue_head;
1297
1298         /* Step to position where retransmission failed, if any,    */
1299         /* consider that buffers may have been released in meantime */
1300
1301         if (r_q_size && buf) {
1302                 u32 last = lesser(mod(r_q_head + r_q_size),
1303                                   link_last_sent(l_ptr));
1304                 u32 first = msg_seqno(buf_msg(buf));
1305
1306                 while (buf && less(first, r_q_head)) {
1307                         first = mod(first + 1);
1308                         buf = buf->next;
1309                 }
1310                 l_ptr->retransm_queue_head = r_q_head = first;
1311                 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1312         }
1313
1314         /* Continue retransmission now, if there is anything: */
1315
1316         if (r_q_size && buf) {
1317                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1318                 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1319                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1320                         l_ptr->retransm_queue_head = mod(++r_q_head);
1321                         l_ptr->retransm_queue_size = --r_q_size;
1322                         l_ptr->stats.retransmitted++;
1323                         return 0;
1324                 } else {
1325                         l_ptr->stats.bearer_congs++;
1326                         return PUSH_FAILED;
1327                 }
1328         }
1329
1330         /* Send deferred protocol message, if any: */
1331
1332         buf = l_ptr->proto_msg_queue;
1333         if (buf) {
1334                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1335                 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1336                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1337                         l_ptr->unacked_window = 0;
1338                         buf_discard(buf);
1339                         l_ptr->proto_msg_queue = NULL;
1340                         return 0;
1341                 } else {
1342                         l_ptr->stats.bearer_congs++;
1343                         return PUSH_FAILED;
1344                 }
1345         }
1346
1347         /* Send one deferred data message, if send window not full: */
1348
1349         buf = l_ptr->next_out;
1350         if (buf) {
1351                 struct tipc_msg *msg = buf_msg(buf);
1352                 u32 next = msg_seqno(msg);
1353                 u32 first = msg_seqno(buf_msg(l_ptr->first_out));
1354
1355                 if (mod(next - first) < l_ptr->queue_limit[0]) {
1356                         msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1357                         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1358                         if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1359                                 if (msg_user(msg) == MSG_BUNDLER)
1360                                         msg_set_type(msg, CLOSED_MSG);
1361                                 l_ptr->next_out = buf->next;
1362                                 return 0;
1363                         } else {
1364                                 l_ptr->stats.bearer_congs++;
1365                                 return PUSH_FAILED;
1366                         }
1367                 }
1368         }
1369         return PUSH_FINISHED;
1370 }
1371
1372 /*
1373  * push_queue(): push out the unsent messages of a link where
1374  *               congestion has abated. Node is locked
1375  */
1376 void tipc_link_push_queue(struct link *l_ptr)
1377 {
1378         u32 res;
1379
1380         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr))
1381                 return;
1382
1383         do {
1384                 res = tipc_link_push_packet(l_ptr);
1385         } while (!res);
1386
1387         if (res == PUSH_FAILED)
1388                 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1389 }
1390
1391 static void link_reset_all(unsigned long addr)
1392 {
1393         struct tipc_node *n_ptr;
1394         char addr_string[16];
1395         u32 i;
1396
1397         read_lock_bh(&tipc_net_lock);
1398         n_ptr = tipc_node_find((u32)addr);
1399         if (!n_ptr) {
1400                 read_unlock_bh(&tipc_net_lock);
1401                 return; /* node no longer exists */
1402         }
1403
1404         tipc_node_lock(n_ptr);
1405
1406         warn("Resetting all links to %s\n",
1407              tipc_addr_string_fill(addr_string, n_ptr->addr));
1408
1409         for (i = 0; i < MAX_BEARERS; i++) {
1410                 if (n_ptr->links[i]) {
1411                         link_print(n_ptr->links[i], "Resetting link\n");
1412                         tipc_link_reset(n_ptr->links[i]);
1413                 }
1414         }
1415
1416         tipc_node_unlock(n_ptr);
1417         read_unlock_bh(&tipc_net_lock);
1418 }
1419
1420 static void link_retransmit_failure(struct link *l_ptr, struct sk_buff *buf)
1421 {
1422         struct tipc_msg *msg = buf_msg(buf);
1423
1424         warn("Retransmission failure on link <%s>\n", l_ptr->name);
1425
1426         if (l_ptr->addr) {
1427
1428                 /* Handle failure on standard link */
1429
1430                 link_print(l_ptr, "Resetting link\n");
1431                 tipc_link_reset(l_ptr);
1432
1433         } else {
1434
1435                 /* Handle failure on broadcast link */
1436
1437                 struct tipc_node *n_ptr;
1438                 char addr_string[16];
1439
1440                 info("Msg seq number: %u,  ", msg_seqno(msg));
1441                 info("Outstanding acks: %lu\n",
1442                      (unsigned long) TIPC_SKB_CB(buf)->handle);
1443
1444                 n_ptr = l_ptr->owner->next;
1445                 tipc_node_lock(n_ptr);
1446
1447                 tipc_addr_string_fill(addr_string, n_ptr->addr);
1448                 info("Multicast link info for %s\n", addr_string);
1449                 info("Supported: %d,  ", n_ptr->bclink.supported);
1450                 info("Acked: %u\n", n_ptr->bclink.acked);
1451                 info("Last in: %u,  ", n_ptr->bclink.last_in);
1452                 info("Gap after: %u,  ", n_ptr->bclink.gap_after);
1453                 info("Gap to: %u\n", n_ptr->bclink.gap_to);
1454                 info("Nack sync: %u\n\n", n_ptr->bclink.nack_sync);
1455
1456                 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1457
1458                 tipc_node_unlock(n_ptr);
1459
1460                 l_ptr->stale_count = 0;
1461         }
1462 }
1463
1464 void tipc_link_retransmit(struct link *l_ptr, struct sk_buff *buf,
1465                           u32 retransmits)
1466 {
1467         struct tipc_msg *msg;
1468
1469         if (!buf)
1470                 return;
1471
1472         msg = buf_msg(buf);
1473
1474         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1475                 if (l_ptr->retransm_queue_size == 0) {
1476                         l_ptr->retransm_queue_head = msg_seqno(msg);
1477                         l_ptr->retransm_queue_size = retransmits;
1478                 } else {
1479                         err("Unexpected retransmit on link %s (qsize=%d)\n",
1480                             l_ptr->name, l_ptr->retransm_queue_size);
1481                 }
1482                 return;
1483         } else {
1484                 /* Detect repeated retransmit failures on uncongested bearer */
1485
1486                 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1487                         if (++l_ptr->stale_count > 100) {
1488                                 link_retransmit_failure(l_ptr, buf);
1489                                 return;
1490                         }
1491                 } else {
1492                         l_ptr->last_retransmitted = msg_seqno(msg);
1493                         l_ptr->stale_count = 1;
1494                 }
1495         }
1496
1497         while (retransmits && (buf != l_ptr->next_out) && buf) {
1498                 msg = buf_msg(buf);
1499                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1500                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1501                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1502                         buf = buf->next;
1503                         retransmits--;
1504                         l_ptr->stats.retransmitted++;
1505                 } else {
1506                         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1507                         l_ptr->stats.bearer_congs++;
1508                         l_ptr->retransm_queue_head = msg_seqno(buf_msg(buf));
1509                         l_ptr->retransm_queue_size = retransmits;
1510                         return;
1511                 }
1512         }
1513
1514         l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1515 }
1516
1517 /**
1518  * link_insert_deferred_queue - insert deferred messages back into receive chain
1519  */
1520
1521 static struct sk_buff *link_insert_deferred_queue(struct link *l_ptr,
1522                                                   struct sk_buff *buf)
1523 {
1524         u32 seq_no;
1525
1526         if (l_ptr->oldest_deferred_in == NULL)
1527                 return buf;
1528
1529         seq_no = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1530         if (seq_no == mod(l_ptr->next_in_no)) {
1531                 l_ptr->newest_deferred_in->next = buf;
1532                 buf = l_ptr->oldest_deferred_in;
1533                 l_ptr->oldest_deferred_in = NULL;
1534                 l_ptr->deferred_inqueue_sz = 0;
1535         }
1536         return buf;
1537 }
1538
1539 /**
1540  * link_recv_buf_validate - validate basic format of received message
1541  *
1542  * This routine ensures a TIPC message has an acceptable header, and at least
1543  * as much data as the header indicates it should.  The routine also ensures
1544  * that the entire message header is stored in the main fragment of the message
1545  * buffer, to simplify future access to message header fields.
1546  *
1547  * Note: Having extra info present in the message header or data areas is OK.
1548  * TIPC will ignore the excess, under the assumption that it is optional info
1549  * introduced by a later release of the protocol.
1550  */
1551
1552 static int link_recv_buf_validate(struct sk_buff *buf)
1553 {
1554         static u32 min_data_hdr_size[8] = {
1555                 SHORT_H_SIZE, MCAST_H_SIZE, LONG_H_SIZE, DIR_MSG_H_SIZE,
1556                 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1557                 };
1558
1559         struct tipc_msg *msg;
1560         u32 tipc_hdr[2];
1561         u32 size;
1562         u32 hdr_size;
1563         u32 min_hdr_size;
1564
1565         if (unlikely(buf->len < MIN_H_SIZE))
1566                 return 0;
1567
1568         msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1569         if (msg == NULL)
1570                 return 0;
1571
1572         if (unlikely(msg_version(msg) != TIPC_VERSION))
1573                 return 0;
1574
1575         size = msg_size(msg);
1576         hdr_size = msg_hdr_sz(msg);
1577         min_hdr_size = msg_isdata(msg) ?
1578                 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1579
1580         if (unlikely((hdr_size < min_hdr_size) ||
1581                      (size < hdr_size) ||
1582                      (buf->len < size) ||
1583                      (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1584                 return 0;
1585
1586         return pskb_may_pull(buf, hdr_size);
1587 }
1588
1589 /**
1590  * tipc_recv_msg - process TIPC messages arriving from off-node
1591  * @head: pointer to message buffer chain
1592  * @tb_ptr: pointer to bearer message arrived on
1593  *
1594  * Invoked with no locks held.  Bearer pointer must point to a valid bearer
1595  * structure (i.e. cannot be NULL), but bearer can be inactive.
1596  */
1597
1598 void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *tb_ptr)
1599 {
1600         read_lock_bh(&tipc_net_lock);
1601         while (head) {
1602                 struct bearer *b_ptr = (struct bearer *)tb_ptr;
1603                 struct tipc_node *n_ptr;
1604                 struct link *l_ptr;
1605                 struct sk_buff *crs;
1606                 struct sk_buff *buf = head;
1607                 struct tipc_msg *msg;
1608                 u32 seq_no;
1609                 u32 ackd;
1610                 u32 released = 0;
1611                 int type;
1612
1613                 head = head->next;
1614
1615                 /* Ensure bearer is still enabled */
1616
1617                 if (unlikely(!b_ptr->active))
1618                         goto cont;
1619
1620                 /* Ensure message is well-formed */
1621
1622                 if (unlikely(!link_recv_buf_validate(buf)))
1623                         goto cont;
1624
1625                 /* Ensure message data is a single contiguous unit */
1626
1627                 if (unlikely(buf_linearize(buf)))
1628                         goto cont;
1629
1630                 /* Handle arrival of a non-unicast link message */
1631
1632                 msg = buf_msg(buf);
1633
1634                 if (unlikely(msg_non_seq(msg))) {
1635                         if (msg_user(msg) ==  LINK_CONFIG)
1636                                 tipc_disc_recv_msg(buf, b_ptr);
1637                         else
1638                                 tipc_bclink_recv_pkt(buf);
1639                         continue;
1640                 }
1641
1642                 if (unlikely(!msg_short(msg) &&
1643                              (msg_destnode(msg) != tipc_own_addr)))
1644                         goto cont;
1645
1646                 /* Discard non-routeable messages destined for another node */
1647
1648                 if (unlikely(!msg_isdata(msg) &&
1649                              (msg_destnode(msg) != tipc_own_addr))) {
1650                         if ((msg_user(msg) != CONN_MANAGER) &&
1651                             (msg_user(msg) != MSG_FRAGMENTER))
1652                                 goto cont;
1653                 }
1654
1655                 /* Locate neighboring node that sent message */
1656
1657                 n_ptr = tipc_node_find(msg_prevnode(msg));
1658                 if (unlikely(!n_ptr))
1659                         goto cont;
1660                 tipc_node_lock(n_ptr);
1661
1662                 /* Don't talk to neighbor during cleanup after last session */
1663
1664                 if (n_ptr->cleanup_required) {
1665                         tipc_node_unlock(n_ptr);
1666                         goto cont;
1667                 }
1668
1669                 /* Locate unicast link endpoint that should handle message */
1670
1671                 l_ptr = n_ptr->links[b_ptr->identity];
1672                 if (unlikely(!l_ptr)) {
1673                         tipc_node_unlock(n_ptr);
1674                         goto cont;
1675                 }
1676
1677                 /* Validate message sequence number info */
1678
1679                 seq_no = msg_seqno(msg);
1680                 ackd = msg_ack(msg);
1681
1682                 /* Release acked messages */
1683
1684                 if (less(n_ptr->bclink.acked, msg_bcast_ack(msg))) {
1685                         if (tipc_node_is_up(n_ptr) && n_ptr->bclink.supported)
1686                                 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1687                 }
1688
1689                 crs = l_ptr->first_out;
1690                 while ((crs != l_ptr->next_out) &&
1691                        less_eq(msg_seqno(buf_msg(crs)), ackd)) {
1692                         struct sk_buff *next = crs->next;
1693
1694                         buf_discard(crs);
1695                         crs = next;
1696                         released++;
1697                 }
1698                 if (released) {
1699                         l_ptr->first_out = crs;
1700                         l_ptr->out_queue_size -= released;
1701                 }
1702
1703                 /* Try sending any messages link endpoint has pending */
1704
1705                 if (unlikely(l_ptr->next_out))
1706                         tipc_link_push_queue(l_ptr);
1707                 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1708                         tipc_link_wakeup_ports(l_ptr, 0);
1709                 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1710                         l_ptr->stats.sent_acks++;
1711                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1712                 }
1713
1714                 /* Now (finally!) process the incoming message */
1715
1716 protocol_check:
1717                 if (likely(link_working_working(l_ptr))) {
1718                         if (likely(seq_no == mod(l_ptr->next_in_no))) {
1719                                 l_ptr->next_in_no++;
1720                                 if (unlikely(l_ptr->oldest_deferred_in))
1721                                         head = link_insert_deferred_queue(l_ptr,
1722                                                                           head);
1723                                 if (likely(msg_is_dest(msg, tipc_own_addr))) {
1724 deliver:
1725                                         if (likely(msg_isdata(msg))) {
1726                                                 tipc_node_unlock(n_ptr);
1727                                                 tipc_port_recv_msg(buf);
1728                                                 continue;
1729                                         }
1730                                         switch (msg_user(msg)) {
1731                                         case MSG_BUNDLER:
1732                                                 l_ptr->stats.recv_bundles++;
1733                                                 l_ptr->stats.recv_bundled +=
1734                                                         msg_msgcnt(msg);
1735                                                 tipc_node_unlock(n_ptr);
1736                                                 tipc_link_recv_bundle(buf);
1737                                                 continue;
1738                                         case ROUTE_DISTRIBUTOR:
1739                                                 tipc_node_unlock(n_ptr);
1740                                                 buf_discard(buf);
1741                                                 continue;
1742                                         case NAME_DISTRIBUTOR:
1743                                                 tipc_node_unlock(n_ptr);
1744                                                 tipc_named_recv(buf);
1745                                                 continue;
1746                                         case CONN_MANAGER:
1747                                                 tipc_node_unlock(n_ptr);
1748                                                 tipc_port_recv_proto_msg(buf);
1749                                                 continue;
1750                                         case MSG_FRAGMENTER:
1751                                                 l_ptr->stats.recv_fragments++;
1752                                                 if (tipc_link_recv_fragment(&l_ptr->defragm_buf,
1753                                                                             &buf, &msg)) {
1754                                                         l_ptr->stats.recv_fragmented++;
1755                                                         goto deliver;
1756                                                 }
1757                                                 break;
1758                                         case CHANGEOVER_PROTOCOL:
1759                                                 type = msg_type(msg);
1760                                                 if (link_recv_changeover_msg(&l_ptr, &buf)) {
1761                                                         msg = buf_msg(buf);
1762                                                         seq_no = msg_seqno(msg);
1763                                                         if (type == ORIGINAL_MSG)
1764                                                                 goto deliver;
1765                                                         goto protocol_check;
1766                                                 }
1767                                                 break;
1768                                         }
1769                                 }
1770                                 tipc_node_unlock(n_ptr);
1771                                 tipc_net_route_msg(buf);
1772                                 continue;
1773                         }
1774                         link_handle_out_of_seq_msg(l_ptr, buf);
1775                         head = link_insert_deferred_queue(l_ptr, head);
1776                         tipc_node_unlock(n_ptr);
1777                         continue;
1778                 }
1779
1780                 if (msg_user(msg) == LINK_PROTOCOL) {
1781                         link_recv_proto_msg(l_ptr, buf);
1782                         head = link_insert_deferred_queue(l_ptr, head);
1783                         tipc_node_unlock(n_ptr);
1784                         continue;
1785                 }
1786                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1787
1788                 if (link_working_working(l_ptr)) {
1789                         /* Re-insert in front of queue */
1790                         buf->next = head;
1791                         head = buf;
1792                         tipc_node_unlock(n_ptr);
1793                         continue;
1794                 }
1795                 tipc_node_unlock(n_ptr);
1796 cont:
1797                 buf_discard(buf);
1798         }
1799         read_unlock_bh(&tipc_net_lock);
1800 }
1801
1802 /*
1803  * link_defer_buf(): Sort a received out-of-sequence packet
1804  *                   into the deferred reception queue.
1805  * Returns the increase of the queue length,i.e. 0 or 1
1806  */
1807
1808 u32 tipc_link_defer_pkt(struct sk_buff **head,
1809                         struct sk_buff **tail,
1810                         struct sk_buff *buf)
1811 {
1812         struct sk_buff *prev = NULL;
1813         struct sk_buff *crs = *head;
1814         u32 seq_no = msg_seqno(buf_msg(buf));
1815
1816         buf->next = NULL;
1817
1818         /* Empty queue ? */
1819         if (*head == NULL) {
1820                 *head = *tail = buf;
1821                 return 1;
1822         }
1823
1824         /* Last ? */
1825         if (less(msg_seqno(buf_msg(*tail)), seq_no)) {
1826                 (*tail)->next = buf;
1827                 *tail = buf;
1828                 return 1;
1829         }
1830
1831         /* Scan through queue and sort it in */
1832         do {
1833                 struct tipc_msg *msg = buf_msg(crs);
1834
1835                 if (less(seq_no, msg_seqno(msg))) {
1836                         buf->next = crs;
1837                         if (prev)
1838                                 prev->next = buf;
1839                         else
1840                                 *head = buf;
1841                         return 1;
1842                 }
1843                 if (seq_no == msg_seqno(msg))
1844                         break;
1845                 prev = crs;
1846                 crs = crs->next;
1847         } while (crs);
1848
1849         /* Message is a duplicate of an existing message */
1850
1851         buf_discard(buf);
1852         return 0;
1853 }
1854
1855 /**
1856  * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
1857  */
1858
1859 static void link_handle_out_of_seq_msg(struct link *l_ptr,
1860                                        struct sk_buff *buf)
1861 {
1862         u32 seq_no = msg_seqno(buf_msg(buf));
1863
1864         if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
1865                 link_recv_proto_msg(l_ptr, buf);
1866                 return;
1867         }
1868
1869         /* Record OOS packet arrival (force mismatch on next timeout) */
1870
1871         l_ptr->checkpoint--;
1872
1873         /*
1874          * Discard packet if a duplicate; otherwise add it to deferred queue
1875          * and notify peer of gap as per protocol specification
1876          */
1877
1878         if (less(seq_no, mod(l_ptr->next_in_no))) {
1879                 l_ptr->stats.duplicates++;
1880                 buf_discard(buf);
1881                 return;
1882         }
1883
1884         if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
1885                                 &l_ptr->newest_deferred_in, buf)) {
1886                 l_ptr->deferred_inqueue_sz++;
1887                 l_ptr->stats.deferred_recv++;
1888                 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
1889                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1890         } else
1891                 l_ptr->stats.duplicates++;
1892 }
1893
1894 /*
1895  * Send protocol message to the other endpoint.
1896  */
1897 void tipc_link_send_proto_msg(struct link *l_ptr, u32 msg_typ, int probe_msg,
1898                               u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
1899 {
1900         struct sk_buff *buf = NULL;
1901         struct tipc_msg *msg = l_ptr->pmsg;
1902         u32 msg_size = sizeof(l_ptr->proto_msg);
1903
1904         if (link_blocked(l_ptr))
1905                 return;
1906         msg_set_type(msg, msg_typ);
1907         msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
1908         msg_set_bcast_ack(msg, mod(l_ptr->owner->bclink.last_in));
1909         msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
1910
1911         if (msg_typ == STATE_MSG) {
1912                 u32 next_sent = mod(l_ptr->next_out_no);
1913
1914                 if (!tipc_link_is_up(l_ptr))
1915                         return;
1916                 if (l_ptr->next_out)
1917                         next_sent = msg_seqno(buf_msg(l_ptr->next_out));
1918                 msg_set_next_sent(msg, next_sent);
1919                 if (l_ptr->oldest_deferred_in) {
1920                         u32 rec = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1921                         gap = mod(rec - mod(l_ptr->next_in_no));
1922                 }
1923                 msg_set_seq_gap(msg, gap);
1924                 if (gap)
1925                         l_ptr->stats.sent_nacks++;
1926                 msg_set_link_tolerance(msg, tolerance);
1927                 msg_set_linkprio(msg, priority);
1928                 msg_set_max_pkt(msg, ack_mtu);
1929                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1930                 msg_set_probe(msg, probe_msg != 0);
1931                 if (probe_msg) {
1932                         u32 mtu = l_ptr->max_pkt;
1933
1934                         if ((mtu < l_ptr->max_pkt_target) &&
1935                             link_working_working(l_ptr) &&
1936                             l_ptr->fsm_msg_cnt) {
1937                                 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1938                                 if (l_ptr->max_pkt_probes == 10) {
1939                                         l_ptr->max_pkt_target = (msg_size - 4);
1940                                         l_ptr->max_pkt_probes = 0;
1941                                         msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1942                                 }
1943                                 l_ptr->max_pkt_probes++;
1944                         }
1945
1946                         l_ptr->stats.sent_probes++;
1947                 }
1948                 l_ptr->stats.sent_states++;
1949         } else {                /* RESET_MSG or ACTIVATE_MSG */
1950                 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
1951                 msg_set_seq_gap(msg, 0);
1952                 msg_set_next_sent(msg, 1);
1953                 msg_set_link_tolerance(msg, l_ptr->tolerance);
1954                 msg_set_linkprio(msg, l_ptr->priority);
1955                 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
1956         }
1957
1958         if (tipc_node_has_redundant_links(l_ptr->owner))
1959                 msg_set_redundant_link(msg);
1960         else
1961                 msg_clear_redundant_link(msg);
1962         msg_set_linkprio(msg, l_ptr->priority);
1963
1964         /* Ensure sequence number will not fit : */
1965
1966         msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
1967
1968         /* Congestion? */
1969
1970         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1971                 if (!l_ptr->proto_msg_queue) {
1972                         l_ptr->proto_msg_queue =
1973                                 tipc_buf_acquire(sizeof(l_ptr->proto_msg));
1974                 }
1975                 buf = l_ptr->proto_msg_queue;
1976                 if (!buf)
1977                         return;
1978                 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1979                 return;
1980         }
1981         msg_set_timestamp(msg, jiffies_to_msecs(jiffies));
1982
1983         /* Message can be sent */
1984
1985         buf = tipc_buf_acquire(msg_size);
1986         if (!buf)
1987                 return;
1988
1989         skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1990         msg_set_size(buf_msg(buf), msg_size);
1991
1992         if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1993                 l_ptr->unacked_window = 0;
1994                 buf_discard(buf);
1995                 return;
1996         }
1997
1998         /* New congestion */
1999         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
2000         l_ptr->proto_msg_queue = buf;
2001         l_ptr->stats.bearer_congs++;
2002 }
2003
2004 /*
2005  * Receive protocol message :
2006  * Note that network plane id propagates through the network, and may
2007  * change at any time. The node with lowest address rules
2008  */
2009
2010 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf)
2011 {
2012         u32 rec_gap = 0;
2013         u32 max_pkt_info;
2014         u32 max_pkt_ack;
2015         u32 msg_tol;
2016         struct tipc_msg *msg = buf_msg(buf);
2017
2018         if (link_blocked(l_ptr))
2019                 goto exit;
2020
2021         /* record unnumbered packet arrival (force mismatch on next timeout) */
2022
2023         l_ptr->checkpoint--;
2024
2025         if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
2026                 if (tipc_own_addr > msg_prevnode(msg))
2027                         l_ptr->b_ptr->net_plane = msg_net_plane(msg);
2028
2029         l_ptr->owner->permit_changeover = msg_redundant_link(msg);
2030
2031         switch (msg_type(msg)) {
2032
2033         case RESET_MSG:
2034                 if (!link_working_unknown(l_ptr) &&
2035                     (l_ptr->peer_session != INVALID_SESSION)) {
2036                         if (msg_session(msg) == l_ptr->peer_session)
2037                                 break; /* duplicate: ignore */
2038                 }
2039                 /* fall thru' */
2040         case ACTIVATE_MSG:
2041                 /* Update link settings according other endpoint's values */
2042
2043                 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2044
2045                 msg_tol = msg_link_tolerance(msg);
2046                 if (msg_tol > l_ptr->tolerance)
2047                         link_set_supervision_props(l_ptr, msg_tol);
2048
2049                 if (msg_linkprio(msg) > l_ptr->priority)
2050                         l_ptr->priority = msg_linkprio(msg);
2051
2052                 max_pkt_info = msg_max_pkt(msg);
2053                 if (max_pkt_info) {
2054                         if (max_pkt_info < l_ptr->max_pkt_target)
2055                                 l_ptr->max_pkt_target = max_pkt_info;
2056                         if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2057                                 l_ptr->max_pkt = l_ptr->max_pkt_target;
2058                 } else {
2059                         l_ptr->max_pkt = l_ptr->max_pkt_target;
2060                 }
2061                 l_ptr->owner->bclink.supported = (max_pkt_info != 0);
2062
2063                 link_state_event(l_ptr, msg_type(msg));
2064
2065                 l_ptr->peer_session = msg_session(msg);
2066                 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2067
2068                 /* Synchronize broadcast sequence numbers */
2069                 if (!tipc_node_has_redundant_links(l_ptr->owner))
2070                         l_ptr->owner->bclink.last_in = mod(msg_last_bcast(msg));
2071                 break;
2072         case STATE_MSG:
2073
2074                 msg_tol = msg_link_tolerance(msg);
2075                 if (msg_tol)
2076                         link_set_supervision_props(l_ptr, msg_tol);
2077
2078                 if (msg_linkprio(msg) &&
2079                     (msg_linkprio(msg) != l_ptr->priority)) {
2080                         warn("Resetting link <%s>, priority change %u->%u\n",
2081                              l_ptr->name, l_ptr->priority, msg_linkprio(msg));
2082                         l_ptr->priority = msg_linkprio(msg);
2083                         tipc_link_reset(l_ptr); /* Enforce change to take effect */
2084                         break;
2085                 }
2086                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2087                 l_ptr->stats.recv_states++;
2088                 if (link_reset_unknown(l_ptr))
2089                         break;
2090
2091                 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2092                         rec_gap = mod(msg_next_sent(msg) -
2093                                       mod(l_ptr->next_in_no));
2094                 }
2095
2096                 max_pkt_ack = msg_max_pkt(msg);
2097                 if (max_pkt_ack > l_ptr->max_pkt) {
2098                         l_ptr->max_pkt = max_pkt_ack;
2099                         l_ptr->max_pkt_probes = 0;
2100                 }
2101
2102                 max_pkt_ack = 0;
2103                 if (msg_probe(msg)) {
2104                         l_ptr->stats.recv_probes++;
2105                         if (msg_size(msg) > sizeof(l_ptr->proto_msg))
2106                                 max_pkt_ack = msg_size(msg);
2107                 }
2108
2109                 /* Protocol message before retransmits, reduce loss risk */
2110
2111                 tipc_bclink_check_gap(l_ptr->owner, msg_last_bcast(msg));
2112
2113                 if (rec_gap || (msg_probe(msg))) {
2114                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2115                                                  0, rec_gap, 0, 0, max_pkt_ack);
2116                 }
2117                 if (msg_seq_gap(msg)) {
2118                         l_ptr->stats.recv_nacks++;
2119                         tipc_link_retransmit(l_ptr, l_ptr->first_out,
2120                                              msg_seq_gap(msg));
2121                 }
2122                 break;
2123         }
2124 exit:
2125         buf_discard(buf);
2126 }
2127
2128
2129 /*
2130  * tipc_link_tunnel(): Send one message via a link belonging to
2131  * another bearer. Owner node is locked.
2132  */
2133 static void tipc_link_tunnel(struct link *l_ptr,
2134                              struct tipc_msg *tunnel_hdr,
2135                              struct tipc_msg  *msg,
2136                              u32 selector)
2137 {
2138         struct link *tunnel;
2139         struct sk_buff *buf;
2140         u32 length = msg_size(msg);
2141
2142         tunnel = l_ptr->owner->active_links[selector & 1];
2143         if (!tipc_link_is_up(tunnel)) {
2144                 warn("Link changeover error, "
2145                      "tunnel link no longer available\n");
2146                 return;
2147         }
2148         msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2149         buf = tipc_buf_acquire(length + INT_H_SIZE);
2150         if (!buf) {
2151                 warn("Link changeover error, "
2152                      "unable to send tunnel msg\n");
2153                 return;
2154         }
2155         skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
2156         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
2157         tipc_link_send_buf(tunnel, buf);
2158 }
2159
2160
2161
2162 /*
2163  * changeover(): Send whole message queue via the remaining link
2164  *               Owner node is locked.
2165  */
2166
2167 void tipc_link_changeover(struct link *l_ptr)
2168 {
2169         u32 msgcount = l_ptr->out_queue_size;
2170         struct sk_buff *crs = l_ptr->first_out;
2171         struct link *tunnel = l_ptr->owner->active_links[0];
2172         struct tipc_msg tunnel_hdr;
2173         int split_bundles;
2174
2175         if (!tunnel)
2176                 return;
2177
2178         if (!l_ptr->owner->permit_changeover) {
2179                 warn("Link changeover error, "
2180                      "peer did not permit changeover\n");
2181                 return;
2182         }
2183
2184         tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2185                  ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
2186         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2187         msg_set_msgcnt(&tunnel_hdr, msgcount);
2188
2189         if (!l_ptr->first_out) {
2190                 struct sk_buff *buf;
2191
2192                 buf = tipc_buf_acquire(INT_H_SIZE);
2193                 if (buf) {
2194                         skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2195                         msg_set_size(&tunnel_hdr, INT_H_SIZE);
2196                         tipc_link_send_buf(tunnel, buf);
2197                 } else {
2198                         warn("Link changeover error, "
2199                              "unable to send changeover msg\n");
2200                 }
2201                 return;
2202         }
2203
2204         split_bundles = (l_ptr->owner->active_links[0] !=
2205                          l_ptr->owner->active_links[1]);
2206
2207         while (crs) {
2208                 struct tipc_msg *msg = buf_msg(crs);
2209
2210                 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2211                         struct tipc_msg *m = msg_get_wrapped(msg);
2212                         unchar *pos = (unchar *)m;
2213
2214                         msgcount = msg_msgcnt(msg);
2215                         while (msgcount--) {
2216                                 msg_set_seqno(m, msg_seqno(msg));
2217                                 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2218                                                  msg_link_selector(m));
2219                                 pos += align(msg_size(m));
2220                                 m = (struct tipc_msg *)pos;
2221                         }
2222                 } else {
2223                         tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2224                                          msg_link_selector(msg));
2225                 }
2226                 crs = crs->next;
2227         }
2228 }
2229
2230 void tipc_link_send_duplicate(struct link *l_ptr, struct link *tunnel)
2231 {
2232         struct sk_buff *iter;
2233         struct tipc_msg tunnel_hdr;
2234
2235         tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2236                  DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
2237         msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2238         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2239         iter = l_ptr->first_out;
2240         while (iter) {
2241                 struct sk_buff *outbuf;
2242                 struct tipc_msg *msg = buf_msg(iter);
2243                 u32 length = msg_size(msg);
2244
2245                 if (msg_user(msg) == MSG_BUNDLER)
2246                         msg_set_type(msg, CLOSED_MSG);
2247                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));   /* Update */
2248                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2249                 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2250                 outbuf = tipc_buf_acquire(length + INT_H_SIZE);
2251                 if (outbuf == NULL) {
2252                         warn("Link changeover error, "
2253                              "unable to send duplicate msg\n");
2254                         return;
2255                 }
2256                 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2257                 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2258                                                length);
2259                 tipc_link_send_buf(tunnel, outbuf);
2260                 if (!tipc_link_is_up(l_ptr))
2261                         return;
2262                 iter = iter->next;
2263         }
2264 }
2265
2266
2267
2268 /**
2269  * buf_extract - extracts embedded TIPC message from another message
2270  * @skb: encapsulating message buffer
2271  * @from_pos: offset to extract from
2272  *
2273  * Returns a new message buffer containing an embedded message.  The
2274  * encapsulating message itself is left unchanged.
2275  */
2276
2277 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2278 {
2279         struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2280         u32 size = msg_size(msg);
2281         struct sk_buff *eb;
2282
2283         eb = tipc_buf_acquire(size);
2284         if (eb)
2285                 skb_copy_to_linear_data(eb, msg, size);
2286         return eb;
2287 }
2288
2289 /*
2290  *  link_recv_changeover_msg(): Receive tunneled packet sent
2291  *  via other link. Node is locked. Return extracted buffer.
2292  */
2293
2294 static int link_recv_changeover_msg(struct link **l_ptr,
2295                                     struct sk_buff **buf)
2296 {
2297         struct sk_buff *tunnel_buf = *buf;
2298         struct link *dest_link;
2299         struct tipc_msg *msg;
2300         struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2301         u32 msg_typ = msg_type(tunnel_msg);
2302         u32 msg_count = msg_msgcnt(tunnel_msg);
2303
2304         dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2305         if (!dest_link)
2306                 goto exit;
2307         if (dest_link == *l_ptr) {
2308                 err("Unexpected changeover message on link <%s>\n",
2309                     (*l_ptr)->name);
2310                 goto exit;
2311         }
2312         *l_ptr = dest_link;
2313         msg = msg_get_wrapped(tunnel_msg);
2314
2315         if (msg_typ == DUPLICATE_MSG) {
2316                 if (less(msg_seqno(msg), mod(dest_link->next_in_no)))
2317                         goto exit;
2318                 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2319                 if (*buf == NULL) {
2320                         warn("Link changeover error, duplicate msg dropped\n");
2321                         goto exit;
2322                 }
2323                 buf_discard(tunnel_buf);
2324                 return 1;
2325         }
2326
2327         /* First original message ?: */
2328
2329         if (tipc_link_is_up(dest_link)) {
2330                 info("Resetting link <%s>, changeover initiated by peer\n",
2331                      dest_link->name);
2332                 tipc_link_reset(dest_link);
2333                 dest_link->exp_msg_count = msg_count;
2334                 if (!msg_count)
2335                         goto exit;
2336         } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2337                 dest_link->exp_msg_count = msg_count;
2338                 if (!msg_count)
2339                         goto exit;
2340         }
2341
2342         /* Receive original message */
2343
2344         if (dest_link->exp_msg_count == 0) {
2345                 warn("Link switchover error, "
2346                      "got too many tunnelled messages\n");
2347                 goto exit;
2348         }
2349         dest_link->exp_msg_count--;
2350         if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2351                 goto exit;
2352         } else {
2353                 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2354                 if (*buf != NULL) {
2355                         buf_discard(tunnel_buf);
2356                         return 1;
2357                 } else {
2358                         warn("Link changeover error, original msg dropped\n");
2359                 }
2360         }
2361 exit:
2362         *buf = NULL;
2363         buf_discard(tunnel_buf);
2364         return 0;
2365 }
2366
2367 /*
2368  *  Bundler functionality:
2369  */
2370 void tipc_link_recv_bundle(struct sk_buff *buf)
2371 {
2372         u32 msgcount = msg_msgcnt(buf_msg(buf));
2373         u32 pos = INT_H_SIZE;
2374         struct sk_buff *obuf;
2375
2376         while (msgcount--) {
2377                 obuf = buf_extract(buf, pos);
2378                 if (obuf == NULL) {
2379                         warn("Link unable to unbundle message(s)\n");
2380                         break;
2381                 }
2382                 pos += align(msg_size(buf_msg(obuf)));
2383                 tipc_net_route_msg(obuf);
2384         }
2385         buf_discard(buf);
2386 }
2387
2388 /*
2389  *  Fragmentation/defragmentation:
2390  */
2391
2392
2393 /*
2394  * link_send_long_buf: Entry for buffers needing fragmentation.
2395  * The buffer is complete, inclusive total message length.
2396  * Returns user data length.
2397  */
2398 static int link_send_long_buf(struct link *l_ptr, struct sk_buff *buf)
2399 {
2400         struct tipc_msg *inmsg = buf_msg(buf);
2401         struct tipc_msg fragm_hdr;
2402         u32 insize = msg_size(inmsg);
2403         u32 dsz = msg_data_sz(inmsg);
2404         unchar *crs = buf->data;
2405         u32 rest = insize;
2406         u32 pack_sz = l_ptr->max_pkt;
2407         u32 fragm_sz = pack_sz - INT_H_SIZE;
2408         u32 fragm_no = 1;
2409         u32 destaddr;
2410
2411         if (msg_short(inmsg))
2412                 destaddr = l_ptr->addr;
2413         else
2414                 destaddr = msg_destnode(inmsg);
2415
2416         if (msg_routed(inmsg))
2417                 msg_set_prevnode(inmsg, tipc_own_addr);
2418
2419         /* Prepare reusable fragment header: */
2420
2421         tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2422                  INT_H_SIZE, destaddr);
2423         msg_set_link_selector(&fragm_hdr, msg_link_selector(inmsg));
2424         msg_set_long_msgno(&fragm_hdr, mod(l_ptr->long_msg_seq_no++));
2425         msg_set_fragm_no(&fragm_hdr, fragm_no);
2426         l_ptr->stats.sent_fragmented++;
2427
2428         /* Chop up message: */
2429
2430         while (rest > 0) {
2431                 struct sk_buff *fragm;
2432
2433                 if (rest <= fragm_sz) {
2434                         fragm_sz = rest;
2435                         msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2436                 }
2437                 fragm = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
2438                 if (fragm == NULL) {
2439                         warn("Link unable to fragment message\n");
2440                         dsz = -ENOMEM;
2441                         goto exit;
2442                 }
2443                 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2444                 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2445                 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2446                                                fragm_sz);
2447                 /*  Send queued messages first, if any: */
2448
2449                 l_ptr->stats.sent_fragments++;
2450                 tipc_link_send_buf(l_ptr, fragm);
2451                 if (!tipc_link_is_up(l_ptr))
2452                         return dsz;
2453                 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
2454                 rest -= fragm_sz;
2455                 crs += fragm_sz;
2456                 msg_set_type(&fragm_hdr, FRAGMENT);
2457         }
2458 exit:
2459         buf_discard(buf);
2460         return dsz;
2461 }
2462
2463 /*
2464  * A pending message being re-assembled must store certain values
2465  * to handle subsequent fragments correctly. The following functions
2466  * help storing these values in unused, available fields in the
2467  * pending message. This makes dynamic memory allocation unecessary.
2468  */
2469
2470 static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2471 {
2472         msg_set_seqno(buf_msg(buf), seqno);
2473 }
2474
2475 static u32 get_fragm_size(struct sk_buff *buf)
2476 {
2477         return msg_ack(buf_msg(buf));
2478 }
2479
2480 static void set_fragm_size(struct sk_buff *buf, u32 sz)
2481 {
2482         msg_set_ack(buf_msg(buf), sz);
2483 }
2484
2485 static u32 get_expected_frags(struct sk_buff *buf)
2486 {
2487         return msg_bcast_ack(buf_msg(buf));
2488 }
2489
2490 static void set_expected_frags(struct sk_buff *buf, u32 exp)
2491 {
2492         msg_set_bcast_ack(buf_msg(buf), exp);
2493 }
2494
2495 static u32 get_timer_cnt(struct sk_buff *buf)
2496 {
2497         return msg_reroute_cnt(buf_msg(buf));
2498 }
2499
2500 static void incr_timer_cnt(struct sk_buff *buf)
2501 {
2502         msg_incr_reroute_cnt(buf_msg(buf));
2503 }
2504
2505 /*
2506  * tipc_link_recv_fragment(): Called with node lock on. Returns
2507  * the reassembled buffer if message is complete.
2508  */
2509 int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2510                             struct tipc_msg **m)
2511 {
2512         struct sk_buff *prev = NULL;
2513         struct sk_buff *fbuf = *fb;
2514         struct tipc_msg *fragm = buf_msg(fbuf);
2515         struct sk_buff *pbuf = *pending;
2516         u32 long_msg_seq_no = msg_long_msgno(fragm);
2517
2518         *fb = NULL;
2519
2520         /* Is there an incomplete message waiting for this fragment? */
2521
2522         while (pbuf && ((msg_seqno(buf_msg(pbuf)) != long_msg_seq_no) ||
2523                         (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2524                 prev = pbuf;
2525                 pbuf = pbuf->next;
2526         }
2527
2528         if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2529                 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2530                 u32 msg_sz = msg_size(imsg);
2531                 u32 fragm_sz = msg_data_sz(fragm);
2532                 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2533                 u32 max =  TIPC_MAX_USER_MSG_SIZE + LONG_H_SIZE;
2534                 if (msg_type(imsg) == TIPC_MCAST_MSG)
2535                         max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2536                 if (msg_size(imsg) > max) {
2537                         buf_discard(fbuf);
2538                         return 0;
2539                 }
2540                 pbuf = tipc_buf_acquire(msg_size(imsg));
2541                 if (pbuf != NULL) {
2542                         pbuf->next = *pending;
2543                         *pending = pbuf;
2544                         skb_copy_to_linear_data(pbuf, imsg,
2545                                                 msg_data_sz(fragm));
2546                         /*  Prepare buffer for subsequent fragments. */
2547
2548                         set_long_msg_seqno(pbuf, long_msg_seq_no);
2549                         set_fragm_size(pbuf, fragm_sz);
2550                         set_expected_frags(pbuf, exp_fragm_cnt - 1);
2551                 } else {
2552                         warn("Link unable to reassemble fragmented message\n");
2553                 }
2554                 buf_discard(fbuf);
2555                 return 0;
2556         } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2557                 u32 dsz = msg_data_sz(fragm);
2558                 u32 fsz = get_fragm_size(pbuf);
2559                 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2560                 u32 exp_frags = get_expected_frags(pbuf) - 1;
2561                 skb_copy_to_linear_data_offset(pbuf, crs,
2562                                                msg_data(fragm), dsz);
2563                 buf_discard(fbuf);
2564
2565                 /* Is message complete? */
2566
2567                 if (exp_frags == 0) {
2568                         if (prev)
2569                                 prev->next = pbuf->next;
2570                         else
2571                                 *pending = pbuf->next;
2572                         msg_reset_reroute_cnt(buf_msg(pbuf));
2573                         *fb = pbuf;
2574                         *m = buf_msg(pbuf);
2575                         return 1;
2576                 }
2577                 set_expected_frags(pbuf, exp_frags);
2578                 return 0;
2579         }
2580         buf_discard(fbuf);
2581         return 0;
2582 }
2583
2584 /**
2585  * link_check_defragm_bufs - flush stale incoming message fragments
2586  * @l_ptr: pointer to link
2587  */
2588
2589 static void link_check_defragm_bufs(struct link *l_ptr)
2590 {
2591         struct sk_buff *prev = NULL;
2592         struct sk_buff *next = NULL;
2593         struct sk_buff *buf = l_ptr->defragm_buf;
2594
2595         if (!buf)
2596                 return;
2597         if (!link_working_working(l_ptr))
2598                 return;
2599         while (buf) {
2600                 u32 cnt = get_timer_cnt(buf);
2601
2602                 next = buf->next;
2603                 if (cnt < 4) {
2604                         incr_timer_cnt(buf);
2605                         prev = buf;
2606                 } else {
2607                         if (prev)
2608                                 prev->next = buf->next;
2609                         else
2610                                 l_ptr->defragm_buf = buf->next;
2611                         buf_discard(buf);
2612                 }
2613                 buf = next;
2614         }
2615 }
2616
2617
2618
2619 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance)
2620 {
2621         l_ptr->tolerance = tolerance;
2622         l_ptr->continuity_interval =
2623                 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2624         l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2625 }
2626
2627
2628 void tipc_link_set_queue_limits(struct link *l_ptr, u32 window)
2629 {
2630         /* Data messages from this node, inclusive FIRST_FRAGM */
2631         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2632         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2633         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2634         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2635         /* Transiting data messages,inclusive FIRST_FRAGM */
2636         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2637         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2638         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2639         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2640         l_ptr->queue_limit[CONN_MANAGER] = 1200;
2641         l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2642         l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2643         /* FRAGMENT and LAST_FRAGMENT packets */
2644         l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2645 }
2646
2647 /**
2648  * link_find_link - locate link by name
2649  * @name - ptr to link name string
2650  * @node - ptr to area to be filled with ptr to associated node
2651  *
2652  * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2653  * this also prevents link deletion.
2654  *
2655  * Returns pointer to link (or 0 if invalid link name).
2656  */
2657
2658 static struct link *link_find_link(const char *name, struct tipc_node **node)
2659 {
2660         struct link_name link_name_parts;
2661         struct bearer *b_ptr;
2662         struct link *l_ptr;
2663
2664         if (!link_name_validate(name, &link_name_parts))
2665                 return NULL;
2666
2667         b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2668         if (!b_ptr)
2669                 return NULL;
2670
2671         *node = tipc_node_find(link_name_parts.addr_peer);
2672         if (!*node)
2673                 return NULL;
2674
2675         l_ptr = (*node)->links[b_ptr->identity];
2676         if (!l_ptr || strcmp(l_ptr->name, name))
2677                 return NULL;
2678
2679         return l_ptr;
2680 }
2681
2682 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2683                                      u16 cmd)
2684 {
2685         struct tipc_link_config *args;
2686         u32 new_value;
2687         struct link *l_ptr;
2688         struct tipc_node *node;
2689         int res;
2690
2691         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2692                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2693
2694         args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2695         new_value = ntohl(args->value);
2696
2697         if (!strcmp(args->name, tipc_bclink_name)) {
2698                 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2699                     (tipc_bclink_set_queue_limits(new_value) == 0))
2700                         return tipc_cfg_reply_none();
2701                 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2702                                                    " (cannot change setting on broadcast link)");
2703         }
2704
2705         read_lock_bh(&tipc_net_lock);
2706         l_ptr = link_find_link(args->name, &node);
2707         if (!l_ptr) {
2708                 read_unlock_bh(&tipc_net_lock);
2709                 return tipc_cfg_reply_error_string("link not found");
2710         }
2711
2712         tipc_node_lock(node);
2713         res = -EINVAL;
2714         switch (cmd) {
2715         case TIPC_CMD_SET_LINK_TOL:
2716                 if ((new_value >= TIPC_MIN_LINK_TOL) &&
2717                     (new_value <= TIPC_MAX_LINK_TOL)) {
2718                         link_set_supervision_props(l_ptr, new_value);
2719                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2720                                                  0, 0, new_value, 0, 0);
2721                         res = 0;
2722                 }
2723                 break;
2724         case TIPC_CMD_SET_LINK_PRI:
2725                 if ((new_value >= TIPC_MIN_LINK_PRI) &&
2726                     (new_value <= TIPC_MAX_LINK_PRI)) {
2727                         l_ptr->priority = new_value;
2728                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2729                                                  0, 0, 0, new_value, 0);
2730                         res = 0;
2731                 }
2732                 break;
2733         case TIPC_CMD_SET_LINK_WINDOW:
2734                 if ((new_value >= TIPC_MIN_LINK_WIN) &&
2735                     (new_value <= TIPC_MAX_LINK_WIN)) {
2736                         tipc_link_set_queue_limits(l_ptr, new_value);
2737                         res = 0;
2738                 }
2739                 break;
2740         }
2741         tipc_node_unlock(node);
2742
2743         read_unlock_bh(&tipc_net_lock);
2744         if (res)
2745                 return tipc_cfg_reply_error_string("cannot change link setting");
2746
2747         return tipc_cfg_reply_none();
2748 }
2749
2750 /**
2751  * link_reset_statistics - reset link statistics
2752  * @l_ptr: pointer to link
2753  */
2754
2755 static void link_reset_statistics(struct link *l_ptr)
2756 {
2757         memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2758         l_ptr->stats.sent_info = l_ptr->next_out_no;
2759         l_ptr->stats.recv_info = l_ptr->next_in_no;
2760 }
2761
2762 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2763 {
2764         char *link_name;
2765         struct link *l_ptr;
2766         struct tipc_node *node;
2767
2768         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2769                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2770
2771         link_name = (char *)TLV_DATA(req_tlv_area);
2772         if (!strcmp(link_name, tipc_bclink_name)) {
2773                 if (tipc_bclink_reset_stats())
2774                         return tipc_cfg_reply_error_string("link not found");
2775                 return tipc_cfg_reply_none();
2776         }
2777
2778         read_lock_bh(&tipc_net_lock);
2779         l_ptr = link_find_link(link_name, &node);
2780         if (!l_ptr) {
2781                 read_unlock_bh(&tipc_net_lock);
2782                 return tipc_cfg_reply_error_string("link not found");
2783         }
2784
2785         tipc_node_lock(node);
2786         link_reset_statistics(l_ptr);
2787         tipc_node_unlock(node);
2788         read_unlock_bh(&tipc_net_lock);
2789         return tipc_cfg_reply_none();
2790 }
2791
2792 /**
2793  * percent - convert count to a percentage of total (rounding up or down)
2794  */
2795
2796 static u32 percent(u32 count, u32 total)
2797 {
2798         return (count * 100 + (total / 2)) / total;
2799 }
2800
2801 /**
2802  * tipc_link_stats - print link statistics
2803  * @name: link name
2804  * @buf: print buffer area
2805  * @buf_size: size of print buffer area
2806  *
2807  * Returns length of print buffer data string (or 0 if error)
2808  */
2809
2810 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
2811 {
2812         struct print_buf pb;
2813         struct link *l_ptr;
2814         struct tipc_node *node;
2815         char *status;
2816         u32 profile_total = 0;
2817
2818         if (!strcmp(name, tipc_bclink_name))
2819                 return tipc_bclink_stats(buf, buf_size);
2820
2821         tipc_printbuf_init(&pb, buf, buf_size);
2822
2823         read_lock_bh(&tipc_net_lock);
2824         l_ptr = link_find_link(name, &node);
2825         if (!l_ptr) {
2826                 read_unlock_bh(&tipc_net_lock);
2827                 return 0;
2828         }
2829         tipc_node_lock(node);
2830
2831         if (tipc_link_is_active(l_ptr))
2832                 status = "ACTIVE";
2833         else if (tipc_link_is_up(l_ptr))
2834                 status = "STANDBY";
2835         else
2836                 status = "DEFUNCT";
2837         tipc_printf(&pb, "Link <%s>\n"
2838                          "  %s  MTU:%u  Priority:%u  Tolerance:%u ms"
2839                          "  Window:%u packets\n",
2840                     l_ptr->name, status, l_ptr->max_pkt,
2841                     l_ptr->priority, l_ptr->tolerance, l_ptr->queue_limit[0]);
2842         tipc_printf(&pb, "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
2843                     l_ptr->next_in_no - l_ptr->stats.recv_info,
2844                     l_ptr->stats.recv_fragments,
2845                     l_ptr->stats.recv_fragmented,
2846                     l_ptr->stats.recv_bundles,
2847                     l_ptr->stats.recv_bundled);
2848         tipc_printf(&pb, "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
2849                     l_ptr->next_out_no - l_ptr->stats.sent_info,
2850                     l_ptr->stats.sent_fragments,
2851                     l_ptr->stats.sent_fragmented,
2852                     l_ptr->stats.sent_bundles,
2853                     l_ptr->stats.sent_bundled);
2854         profile_total = l_ptr->stats.msg_length_counts;
2855         if (!profile_total)
2856                 profile_total = 1;
2857         tipc_printf(&pb, "  TX profile sample:%u packets  average:%u octets\n"
2858                          "  0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
2859                          "-16354:%u%% -32768:%u%% -66000:%u%%\n",
2860                     l_ptr->stats.msg_length_counts,
2861                     l_ptr->stats.msg_lengths_total / profile_total,
2862                     percent(l_ptr->stats.msg_length_profile[0], profile_total),
2863                     percent(l_ptr->stats.msg_length_profile[1], profile_total),
2864                     percent(l_ptr->stats.msg_length_profile[2], profile_total),
2865                     percent(l_ptr->stats.msg_length_profile[3], profile_total),
2866                     percent(l_ptr->stats.msg_length_profile[4], profile_total),
2867                     percent(l_ptr->stats.msg_length_profile[5], profile_total),
2868                     percent(l_ptr->stats.msg_length_profile[6], profile_total));
2869         tipc_printf(&pb, "  RX states:%u probes:%u naks:%u defs:%u dups:%u\n",
2870                     l_ptr->stats.recv_states,
2871                     l_ptr->stats.recv_probes,
2872                     l_ptr->stats.recv_nacks,
2873                     l_ptr->stats.deferred_recv,
2874                     l_ptr->stats.duplicates);
2875         tipc_printf(&pb, "  TX states:%u probes:%u naks:%u acks:%u dups:%u\n",
2876                     l_ptr->stats.sent_states,
2877                     l_ptr->stats.sent_probes,
2878                     l_ptr->stats.sent_nacks,
2879                     l_ptr->stats.sent_acks,
2880                     l_ptr->stats.retransmitted);
2881         tipc_printf(&pb, "  Congestion bearer:%u link:%u  Send queue max:%u avg:%u\n",
2882                     l_ptr->stats.bearer_congs,
2883                     l_ptr->stats.link_congs,
2884                     l_ptr->stats.max_queue_sz,
2885                     l_ptr->stats.queue_sz_counts
2886                     ? (l_ptr->stats.accu_queue_sz / l_ptr->stats.queue_sz_counts)
2887                     : 0);
2888
2889         tipc_node_unlock(node);
2890         read_unlock_bh(&tipc_net_lock);
2891         return tipc_printbuf_validate(&pb);
2892 }
2893
2894 #define MAX_LINK_STATS_INFO 2000
2895
2896 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
2897 {
2898         struct sk_buff *buf;
2899         struct tlv_desc *rep_tlv;
2900         int str_len;
2901
2902         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2903                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2904
2905         buf = tipc_cfg_reply_alloc(TLV_SPACE(MAX_LINK_STATS_INFO));
2906         if (!buf)
2907                 return NULL;
2908
2909         rep_tlv = (struct tlv_desc *)buf->data;
2910
2911         str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
2912                                   (char *)TLV_DATA(rep_tlv), MAX_LINK_STATS_INFO);
2913         if (!str_len) {
2914                 buf_discard(buf);
2915                 return tipc_cfg_reply_error_string("link not found");
2916         }
2917
2918         skb_put(buf, TLV_SPACE(str_len));
2919         TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2920
2921         return buf;
2922 }
2923
2924 /**
2925  * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
2926  * @dest: network address of destination node
2927  * @selector: used to select from set of active links
2928  *
2929  * If no active link can be found, uses default maximum packet size.
2930  */
2931
2932 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
2933 {
2934         struct tipc_node *n_ptr;
2935         struct link *l_ptr;
2936         u32 res = MAX_PKT_DEFAULT;
2937
2938         if (dest == tipc_own_addr)
2939                 return MAX_MSG_SIZE;
2940
2941         read_lock_bh(&tipc_net_lock);
2942         n_ptr = tipc_node_find(dest);
2943         if (n_ptr) {
2944                 tipc_node_lock(n_ptr);
2945                 l_ptr = n_ptr->active_links[selector & 1];
2946                 if (l_ptr)
2947                         res = l_ptr->max_pkt;
2948                 tipc_node_unlock(n_ptr);
2949         }
2950         read_unlock_bh(&tipc_net_lock);
2951         return res;
2952 }
2953
2954 static void link_print(struct link *l_ptr, const char *str)
2955 {
2956         char print_area[256];
2957         struct print_buf pb;
2958         struct print_buf *buf = &pb;
2959
2960         tipc_printbuf_init(buf, print_area, sizeof(print_area));
2961
2962         tipc_printf(buf, str);
2963         tipc_printf(buf, "Link %x<%s>:",
2964                     l_ptr->addr, l_ptr->b_ptr->publ.name);
2965
2966 #ifdef CONFIG_TIPC_DEBUG
2967         if (link_reset_reset(l_ptr) || link_reset_unknown(l_ptr))
2968                 goto print_state;
2969
2970         tipc_printf(buf, ": NXO(%u):", mod(l_ptr->next_out_no));
2971         tipc_printf(buf, "NXI(%u):", mod(l_ptr->next_in_no));
2972         tipc_printf(buf, "SQUE");
2973         if (l_ptr->first_out) {
2974                 tipc_printf(buf, "[%u..", msg_seqno(buf_msg(l_ptr->first_out)));
2975                 if (l_ptr->next_out)
2976                         tipc_printf(buf, "%u..",
2977                                     msg_seqno(buf_msg(l_ptr->next_out)));
2978                 tipc_printf(buf, "%u]", msg_seqno(buf_msg(l_ptr->last_out)));
2979                 if ((mod(msg_seqno(buf_msg(l_ptr->last_out)) -
2980                          msg_seqno(buf_msg(l_ptr->first_out)))
2981                      != (l_ptr->out_queue_size - 1)) ||
2982                     (l_ptr->last_out->next != NULL)) {
2983                         tipc_printf(buf, "\nSend queue inconsistency\n");
2984                         tipc_printf(buf, "first_out= %x ", l_ptr->first_out);
2985                         tipc_printf(buf, "next_out= %x ", l_ptr->next_out);
2986                         tipc_printf(buf, "last_out= %x ", l_ptr->last_out);
2987                 }
2988         } else
2989                 tipc_printf(buf, "[]");
2990         tipc_printf(buf, "SQSIZ(%u)", l_ptr->out_queue_size);
2991         if (l_ptr->oldest_deferred_in) {
2992                 u32 o = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
2993                 u32 n = msg_seqno(buf_msg(l_ptr->newest_deferred_in));
2994                 tipc_printf(buf, ":RQUE[%u..%u]", o, n);
2995                 if (l_ptr->deferred_inqueue_sz != mod((n + 1) - o)) {
2996                         tipc_printf(buf, ":RQSIZ(%u)",
2997                                     l_ptr->deferred_inqueue_sz);
2998                 }
2999         }
3000 print_state:
3001 #endif
3002
3003         if (link_working_unknown(l_ptr))
3004                 tipc_printf(buf, ":WU");
3005         else if (link_reset_reset(l_ptr))
3006                 tipc_printf(buf, ":RR");
3007         else if (link_reset_unknown(l_ptr))
3008                 tipc_printf(buf, ":RU");
3009         else if (link_working_working(l_ptr))
3010                 tipc_printf(buf, ":WW");
3011         tipc_printf(buf, "\n");
3012
3013         tipc_printbuf_validate(buf);
3014         info("%s", print_area);
3015 }
3016