Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
[linux-drm-fsl-dcu.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 "dbg.h"
39 #include "link.h"
40 #include "net.h"
41 #include "node.h"
42 #include "port.h"
43 #include "addr.h"
44 #include "node_subscr.h"
45 #include "name_distr.h"
46 #include "bearer.h"
47 #include "name_table.h"
48 #include "discover.h"
49 #include "config.h"
50 #include "bcast.h"
51
52
53 /*
54  * Out-of-range value for link session numbers
55  */
56
57 #define INVALID_SESSION 0x10000
58
59 /*
60  * Limit for deferred reception queue:
61  */
62
63 #define DEF_QUEUE_LIMIT 256u
64
65 /*
66  * Link state events:
67  */
68
69 #define  STARTING_EVT    856384768      /* link processing trigger */
70 #define  TRAFFIC_MSG_EVT 560815u        /* rx'd ??? */
71 #define  TIMEOUT_EVT     560817u        /* link timer expired */
72
73 /*
74  * The following two 'message types' is really just implementation
75  * data conveniently stored in the message header.
76  * They must not be considered part of the protocol
77  */
78 #define OPEN_MSG   0
79 #define CLOSED_MSG 1
80
81 /*
82  * State value stored in 'exp_msg_count'
83  */
84
85 #define START_CHANGEOVER 100000u
86
87 /**
88  * struct link_name - deconstructed link name
89  * @addr_local: network address of node at this end
90  * @if_local: name of interface at this end
91  * @addr_peer: network address of node at far end
92  * @if_peer: name of interface at far end
93  */
94
95 struct link_name {
96         u32 addr_local;
97         char if_local[TIPC_MAX_IF_NAME];
98         u32 addr_peer;
99         char if_peer[TIPC_MAX_IF_NAME];
100 };
101
102 #if 0
103
104 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
105
106 /**
107  * struct link_event - link up/down event notification
108  */
109
110 struct link_event {
111         u32 addr;
112         int up;
113         void (*fcn)(u32, char *, int);
114         char name[TIPC_MAX_LINK_NAME];
115 };
116
117 #endif
118
119 static void link_handle_out_of_seq_msg(struct link *l_ptr,
120                                        struct sk_buff *buf);
121 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf);
122 static int  link_recv_changeover_msg(struct link **l_ptr, struct sk_buff **buf);
123 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance);
124 static int  link_send_sections_long(struct port *sender,
125                                     struct iovec const *msg_sect,
126                                     u32 num_sect, u32 destnode);
127 static void link_check_defragm_bufs(struct link *l_ptr);
128 static void link_state_event(struct link *l_ptr, u32 event);
129 static void link_reset_statistics(struct link *l_ptr);
130 static void link_print(struct link *l_ptr, struct print_buf *buf,
131                        const char *str);
132
133 /*
134  * Debugging code used by link routines only
135  *
136  * When debugging link problems on a system that has multiple links,
137  * the standard TIPC debugging routines may not be useful since they
138  * allow the output from multiple links to be intermixed.  For this reason
139  * routines of the form "dbg_link_XXX()" have been created that will capture
140  * debug info into a link's personal print buffer, which can then be dumped
141  * into the TIPC system log (TIPC_LOG) upon request.
142  *
143  * To enable per-link debugging, use LINK_LOG_BUF_SIZE to specify the size
144  * of the print buffer used by each link.  If LINK_LOG_BUF_SIZE is set to 0,
145  * the dbg_link_XXX() routines simply send their output to the standard
146  * debug print buffer (DBG_OUTPUT), if it has been defined; this can be useful
147  * when there is only a single link in the system being debugged.
148  *
149  * Notes:
150  * - When enabled, LINK_LOG_BUF_SIZE should be set to at least TIPC_PB_MIN_SIZE
151  * - "l_ptr" must be valid when using dbg_link_XXX() macros
152  */
153
154 #define LINK_LOG_BUF_SIZE 0
155
156 #define dbg_link(fmt, arg...) \
157         do { \
158                 if (LINK_LOG_BUF_SIZE) \
159                         tipc_printf(&l_ptr->print_buf, fmt, ## arg); \
160         } while (0)
161 #define dbg_link_msg(msg, txt) \
162         do { \
163                 if (LINK_LOG_BUF_SIZE) \
164                         tipc_msg_dbg(&l_ptr->print_buf, msg, txt); \
165         } while (0)
166 #define dbg_link_state(txt) \
167         do { \
168                 if (LINK_LOG_BUF_SIZE) \
169                         link_print(l_ptr, &l_ptr->print_buf, txt); \
170         } while (0)
171 #define dbg_link_dump() do { \
172         if (LINK_LOG_BUF_SIZE) { \
173                 tipc_printf(LOG, "\n\nDumping link <%s>:\n", l_ptr->name); \
174                 tipc_printbuf_move(LOG, &l_ptr->print_buf); \
175         } \
176 } while (0)
177
178 static void dbg_print_link(struct link *l_ptr, const char *str)
179 {
180         if (DBG_OUTPUT != TIPC_NULL)
181                 link_print(l_ptr, DBG_OUTPUT, str);
182 }
183
184 static void dbg_print_buf_chain(struct sk_buff *root_buf)
185 {
186         if (DBG_OUTPUT != TIPC_NULL) {
187                 struct sk_buff *buf = root_buf;
188
189                 while (buf) {
190                         msg_dbg(buf_msg(buf), "In chain: ");
191                         buf = buf->next;
192                 }
193         }
194 }
195
196 /*
197  *  Simple link routines
198  */
199
200 static unsigned int align(unsigned int i)
201 {
202         return (i + 3) & ~3u;
203 }
204
205 static int link_working_working(struct link *l_ptr)
206 {
207         return (l_ptr->state == WORKING_WORKING);
208 }
209
210 static int link_working_unknown(struct link *l_ptr)
211 {
212         return (l_ptr->state == WORKING_UNKNOWN);
213 }
214
215 static int link_reset_unknown(struct link *l_ptr)
216 {
217         return (l_ptr->state == RESET_UNKNOWN);
218 }
219
220 static int link_reset_reset(struct link *l_ptr)
221 {
222         return (l_ptr->state == RESET_RESET);
223 }
224
225 static int link_blocked(struct link *l_ptr)
226 {
227         return (l_ptr->exp_msg_count || l_ptr->blocked);
228 }
229
230 static int link_congested(struct link *l_ptr)
231 {
232         return (l_ptr->out_queue_size >= l_ptr->queue_limit[0]);
233 }
234
235 static u32 link_max_pkt(struct link *l_ptr)
236 {
237         return l_ptr->max_pkt;
238 }
239
240 static void link_init_max_pkt(struct link *l_ptr)
241 {
242         u32 max_pkt;
243
244         max_pkt = (l_ptr->b_ptr->publ.mtu & ~3);
245         if (max_pkt > MAX_MSG_SIZE)
246                 max_pkt = MAX_MSG_SIZE;
247
248         l_ptr->max_pkt_target = max_pkt;
249         if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
250                 l_ptr->max_pkt = l_ptr->max_pkt_target;
251         else
252                 l_ptr->max_pkt = MAX_PKT_DEFAULT;
253
254         l_ptr->max_pkt_probes = 0;
255 }
256
257 static u32 link_next_sent(struct link *l_ptr)
258 {
259         if (l_ptr->next_out)
260                 return msg_seqno(buf_msg(l_ptr->next_out));
261         return mod(l_ptr->next_out_no);
262 }
263
264 static u32 link_last_sent(struct link *l_ptr)
265 {
266         return mod(link_next_sent(l_ptr) - 1);
267 }
268
269 /*
270  *  Simple non-static link routines (i.e. referenced outside this file)
271  */
272
273 int tipc_link_is_up(struct link *l_ptr)
274 {
275         if (!l_ptr)
276                 return 0;
277         return (link_working_working(l_ptr) || link_working_unknown(l_ptr));
278 }
279
280 int tipc_link_is_active(struct link *l_ptr)
281 {
282         return ((l_ptr->owner->active_links[0] == l_ptr) ||
283                 (l_ptr->owner->active_links[1] == l_ptr));
284 }
285
286 /**
287  * link_name_validate - validate & (optionally) deconstruct link name
288  * @name - ptr to link name string
289  * @name_parts - ptr to area for link name components (or NULL if not needed)
290  *
291  * Returns 1 if link name is valid, otherwise 0.
292  */
293
294 static int link_name_validate(const char *name, struct link_name *name_parts)
295 {
296         char name_copy[TIPC_MAX_LINK_NAME];
297         char *addr_local;
298         char *if_local;
299         char *addr_peer;
300         char *if_peer;
301         char dummy;
302         u32 z_local, c_local, n_local;
303         u32 z_peer, c_peer, n_peer;
304         u32 if_local_len;
305         u32 if_peer_len;
306
307         /* copy link name & ensure length is OK */
308
309         name_copy[TIPC_MAX_LINK_NAME - 1] = 0;
310         /* need above in case non-Posix strncpy() doesn't pad with nulls */
311         strncpy(name_copy, name, TIPC_MAX_LINK_NAME);
312         if (name_copy[TIPC_MAX_LINK_NAME - 1] != 0)
313                 return 0;
314
315         /* ensure all component parts of link name are present */
316
317         addr_local = name_copy;
318         if ((if_local = strchr(addr_local, ':')) == NULL)
319                 return 0;
320         *(if_local++) = 0;
321         if ((addr_peer = strchr(if_local, '-')) == NULL)
322                 return 0;
323         *(addr_peer++) = 0;
324         if_local_len = addr_peer - if_local;
325         if ((if_peer = strchr(addr_peer, ':')) == NULL)
326                 return 0;
327         *(if_peer++) = 0;
328         if_peer_len = strlen(if_peer) + 1;
329
330         /* validate component parts of link name */
331
332         if ((sscanf(addr_local, "%u.%u.%u%c",
333                     &z_local, &c_local, &n_local, &dummy) != 3) ||
334             (sscanf(addr_peer, "%u.%u.%u%c",
335                     &z_peer, &c_peer, &n_peer, &dummy) != 3) ||
336             (z_local > 255) || (c_local > 4095) || (n_local > 4095) ||
337             (z_peer  > 255) || (c_peer  > 4095) || (n_peer  > 4095) ||
338             (if_local_len <= 1) || (if_local_len > TIPC_MAX_IF_NAME) ||
339             (if_peer_len  <= 1) || (if_peer_len  > TIPC_MAX_IF_NAME) ||
340             (strspn(if_local, tipc_alphabet) != (if_local_len - 1)) ||
341             (strspn(if_peer, tipc_alphabet) != (if_peer_len - 1)))
342                 return 0;
343
344         /* return link name components, if necessary */
345
346         if (name_parts) {
347                 name_parts->addr_local = tipc_addr(z_local, c_local, n_local);
348                 strcpy(name_parts->if_local, if_local);
349                 name_parts->addr_peer = tipc_addr(z_peer, c_peer, n_peer);
350                 strcpy(name_parts->if_peer, if_peer);
351         }
352         return 1;
353 }
354
355 /**
356  * link_timeout - handle expiration of link timer
357  * @l_ptr: pointer to link
358  *
359  * This routine must not grab "tipc_net_lock" to avoid a potential deadlock conflict
360  * with tipc_link_delete().  (There is no risk that the node will be deleted by
361  * another thread because tipc_link_delete() always cancels the link timer before
362  * tipc_node_delete() is called.)
363  */
364
365 static void link_timeout(struct link *l_ptr)
366 {
367         tipc_node_lock(l_ptr->owner);
368
369         /* update counters used in statistical profiling of send traffic */
370
371         l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
372         l_ptr->stats.queue_sz_counts++;
373
374         if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
375                 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
376
377         if (l_ptr->first_out) {
378                 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
379                 u32 length = msg_size(msg);
380
381                 if ((msg_user(msg) == MSG_FRAGMENTER) &&
382                     (msg_type(msg) == FIRST_FRAGMENT)) {
383                         length = msg_size(msg_get_wrapped(msg));
384                 }
385                 if (length) {
386                         l_ptr->stats.msg_lengths_total += length;
387                         l_ptr->stats.msg_length_counts++;
388                         if (length <= 64)
389                                 l_ptr->stats.msg_length_profile[0]++;
390                         else if (length <= 256)
391                                 l_ptr->stats.msg_length_profile[1]++;
392                         else if (length <= 1024)
393                                 l_ptr->stats.msg_length_profile[2]++;
394                         else if (length <= 4096)
395                                 l_ptr->stats.msg_length_profile[3]++;
396                         else if (length <= 16384)
397                                 l_ptr->stats.msg_length_profile[4]++;
398                         else if (length <= 32768)
399                                 l_ptr->stats.msg_length_profile[5]++;
400                         else
401                                 l_ptr->stats.msg_length_profile[6]++;
402                 }
403         }
404
405         /* do all other link processing performed on a periodic basis */
406
407         link_check_defragm_bufs(l_ptr);
408
409         link_state_event(l_ptr, TIMEOUT_EVT);
410
411         if (l_ptr->next_out)
412                 tipc_link_push_queue(l_ptr);
413
414         tipc_node_unlock(l_ptr->owner);
415 }
416
417 static void link_set_timer(struct link *l_ptr, u32 time)
418 {
419         k_start_timer(&l_ptr->timer, time);
420 }
421
422 /**
423  * tipc_link_create - create a new link
424  * @b_ptr: pointer to associated bearer
425  * @peer: network address of node at other end of link
426  * @media_addr: media address to use when sending messages over link
427  *
428  * Returns pointer to link.
429  */
430
431 struct link *tipc_link_create(struct bearer *b_ptr, const u32 peer,
432                               const struct tipc_media_addr *media_addr)
433 {
434         struct link *l_ptr;
435         struct tipc_msg *msg;
436         char *if_name;
437
438         l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
439         if (!l_ptr) {
440                 warn("Link creation failed, no memory\n");
441                 return NULL;
442         }
443
444         if (LINK_LOG_BUF_SIZE) {
445                 char *pb = kmalloc(LINK_LOG_BUF_SIZE, GFP_ATOMIC);
446
447                 if (!pb) {
448                         kfree(l_ptr);
449                         warn("Link creation failed, no memory for print buffer\n");
450                         return NULL;
451                 }
452                 tipc_printbuf_init(&l_ptr->print_buf, pb, LINK_LOG_BUF_SIZE);
453         }
454
455         l_ptr->addr = peer;
456         if_name = strchr(b_ptr->publ.name, ':') + 1;
457         sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:",
458                 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
459                 tipc_node(tipc_own_addr),
460                 if_name,
461                 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
462                 /* note: peer i/f is appended to link name by reset/activate */
463         memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
464         l_ptr->checkpoint = 1;
465         l_ptr->b_ptr = b_ptr;
466         link_set_supervision_props(l_ptr, b_ptr->media->tolerance);
467         l_ptr->state = RESET_UNKNOWN;
468
469         l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
470         msg = l_ptr->pmsg;
471         msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
472         msg_set_size(msg, sizeof(l_ptr->proto_msg));
473         msg_set_session(msg, (tipc_random & 0xffff));
474         msg_set_bearer_id(msg, b_ptr->identity);
475         strcpy((char *)msg_data(msg), if_name);
476
477         l_ptr->priority = b_ptr->priority;
478         tipc_link_set_queue_limits(l_ptr, b_ptr->media->window);
479
480         link_init_max_pkt(l_ptr);
481
482         l_ptr->next_out_no = 1;
483         INIT_LIST_HEAD(&l_ptr->waiting_ports);
484
485         link_reset_statistics(l_ptr);
486
487         l_ptr->owner = tipc_node_attach_link(l_ptr);
488         if (!l_ptr->owner) {
489                 if (LINK_LOG_BUF_SIZE)
490                         kfree(l_ptr->print_buf.buf);
491                 kfree(l_ptr);
492                 return NULL;
493         }
494
495         k_init_timer(&l_ptr->timer, (Handler)link_timeout, (unsigned long)l_ptr);
496         list_add_tail(&l_ptr->link_list, &b_ptr->links);
497         tipc_k_signal((Handler)tipc_link_start, (unsigned long)l_ptr);
498
499         dbg("tipc_link_create(): tolerance = %u,cont intv = %u, abort_limit = %u\n",
500             l_ptr->tolerance, l_ptr->continuity_interval, l_ptr->abort_limit);
501
502         return l_ptr;
503 }
504
505 /**
506  * tipc_link_delete - delete a link
507  * @l_ptr: pointer to link
508  *
509  * Note: 'tipc_net_lock' is write_locked, bearer is locked.
510  * This routine must not grab the node lock until after link timer cancellation
511  * to avoid a potential deadlock situation.
512  */
513
514 void tipc_link_delete(struct link *l_ptr)
515 {
516         if (!l_ptr) {
517                 err("Attempt to delete non-existent link\n");
518                 return;
519         }
520
521         dbg("tipc_link_delete()\n");
522
523         k_cancel_timer(&l_ptr->timer);
524
525         tipc_node_lock(l_ptr->owner);
526         tipc_link_reset(l_ptr);
527         tipc_node_detach_link(l_ptr->owner, l_ptr);
528         tipc_link_stop(l_ptr);
529         list_del_init(&l_ptr->link_list);
530         if (LINK_LOG_BUF_SIZE)
531                 kfree(l_ptr->print_buf.buf);
532         tipc_node_unlock(l_ptr->owner);
533         k_term_timer(&l_ptr->timer);
534         kfree(l_ptr);
535 }
536
537 void tipc_link_start(struct link *l_ptr)
538 {
539         dbg("tipc_link_start %x\n", l_ptr);
540         link_state_event(l_ptr, STARTING_EVT);
541 }
542
543 /**
544  * link_schedule_port - schedule port for deferred sending
545  * @l_ptr: pointer to link
546  * @origport: reference to sending port
547  * @sz: amount of data to be sent
548  *
549  * Schedules port for renewed sending of messages after link congestion
550  * has abated.
551  */
552
553 static int link_schedule_port(struct link *l_ptr, u32 origport, u32 sz)
554 {
555         struct port *p_ptr;
556
557         spin_lock_bh(&tipc_port_list_lock);
558         p_ptr = tipc_port_lock(origport);
559         if (p_ptr) {
560                 if (!p_ptr->wakeup)
561                         goto exit;
562                 if (!list_empty(&p_ptr->wait_list))
563                         goto exit;
564                 p_ptr->congested_link = l_ptr;
565                 p_ptr->publ.congested = 1;
566                 p_ptr->waiting_pkts = 1 + ((sz - 1) / link_max_pkt(l_ptr));
567                 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
568                 l_ptr->stats.link_congs++;
569 exit:
570                 tipc_port_unlock(p_ptr);
571         }
572         spin_unlock_bh(&tipc_port_list_lock);
573         return -ELINKCONG;
574 }
575
576 void tipc_link_wakeup_ports(struct link *l_ptr, int all)
577 {
578         struct port *p_ptr;
579         struct port *temp_p_ptr;
580         int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
581
582         if (all)
583                 win = 100000;
584         if (win <= 0)
585                 return;
586         if (!spin_trylock_bh(&tipc_port_list_lock))
587                 return;
588         if (link_congested(l_ptr))
589                 goto exit;
590         list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
591                                  wait_list) {
592                 if (win <= 0)
593                         break;
594                 list_del_init(&p_ptr->wait_list);
595                 p_ptr->congested_link = NULL;
596                 spin_lock_bh(p_ptr->publ.lock);
597                 p_ptr->publ.congested = 0;
598                 p_ptr->wakeup(&p_ptr->publ);
599                 win -= p_ptr->waiting_pkts;
600                 spin_unlock_bh(p_ptr->publ.lock);
601         }
602
603 exit:
604         spin_unlock_bh(&tipc_port_list_lock);
605 }
606
607 /**
608  * link_release_outqueue - purge link's outbound message queue
609  * @l_ptr: pointer to link
610  */
611
612 static void link_release_outqueue(struct link *l_ptr)
613 {
614         struct sk_buff *buf = l_ptr->first_out;
615         struct sk_buff *next;
616
617         while (buf) {
618                 next = buf->next;
619                 buf_discard(buf);
620                 buf = next;
621         }
622         l_ptr->first_out = NULL;
623         l_ptr->out_queue_size = 0;
624 }
625
626 /**
627  * tipc_link_reset_fragments - purge link's inbound message fragments queue
628  * @l_ptr: pointer to link
629  */
630
631 void tipc_link_reset_fragments(struct link *l_ptr)
632 {
633         struct sk_buff *buf = l_ptr->defragm_buf;
634         struct sk_buff *next;
635
636         while (buf) {
637                 next = buf->next;
638                 buf_discard(buf);
639                 buf = next;
640         }
641         l_ptr->defragm_buf = NULL;
642 }
643
644 /**
645  * tipc_link_stop - purge all inbound and outbound messages associated with link
646  * @l_ptr: pointer to link
647  */
648
649 void tipc_link_stop(struct link *l_ptr)
650 {
651         struct sk_buff *buf;
652         struct sk_buff *next;
653
654         buf = l_ptr->oldest_deferred_in;
655         while (buf) {
656                 next = buf->next;
657                 buf_discard(buf);
658                 buf = next;
659         }
660
661         buf = l_ptr->first_out;
662         while (buf) {
663                 next = buf->next;
664                 buf_discard(buf);
665                 buf = next;
666         }
667
668         tipc_link_reset_fragments(l_ptr);
669
670         buf_discard(l_ptr->proto_msg_queue);
671         l_ptr->proto_msg_queue = NULL;
672 }
673
674 #if 0
675
676 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
677
678 static void link_recv_event(struct link_event *ev)
679 {
680         ev->fcn(ev->addr, ev->name, ev->up);
681         kfree(ev);
682 }
683
684 static void link_send_event(void (*fcn)(u32 a, char *n, int up),
685                             struct link *l_ptr, int up)
686 {
687         struct link_event *ev;
688
689         ev = kmalloc(sizeof(*ev), GFP_ATOMIC);
690         if (!ev) {
691                 warn("Link event allocation failure\n");
692                 return;
693         }
694         ev->addr = l_ptr->addr;
695         ev->up = up;
696         ev->fcn = fcn;
697         memcpy(ev->name, l_ptr->name, TIPC_MAX_LINK_NAME);
698         tipc_k_signal((Handler)link_recv_event, (unsigned long)ev);
699 }
700
701 #else
702
703 #define link_send_event(fcn, l_ptr, up) do { } while (0)
704
705 #endif
706
707 void tipc_link_reset(struct link *l_ptr)
708 {
709         struct sk_buff *buf;
710         u32 prev_state = l_ptr->state;
711         u32 checkpoint = l_ptr->next_in_no;
712         int was_active_link = tipc_link_is_active(l_ptr);
713
714         msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
715
716         /* Link is down, accept any session */
717         l_ptr->peer_session = INVALID_SESSION;
718
719         /* Prepare for max packet size negotiation */
720         link_init_max_pkt(l_ptr);
721
722         l_ptr->state = RESET_UNKNOWN;
723         dbg_link_state("Resetting Link\n");
724
725         if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
726                 return;
727
728         tipc_node_link_down(l_ptr->owner, l_ptr);
729         tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
730 #if 0
731         tipc_printf(TIPC_CONS, "\nReset link <%s>\n", l_ptr->name);
732         dbg_link_dump();
733 #endif
734         if (was_active_link && tipc_node_has_active_links(l_ptr->owner) &&
735             l_ptr->owner->permit_changeover) {
736                 l_ptr->reset_checkpoint = checkpoint;
737                 l_ptr->exp_msg_count = START_CHANGEOVER;
738         }
739
740         /* Clean up all queues: */
741
742         link_release_outqueue(l_ptr);
743         buf_discard(l_ptr->proto_msg_queue);
744         l_ptr->proto_msg_queue = NULL;
745         buf = l_ptr->oldest_deferred_in;
746         while (buf) {
747                 struct sk_buff *next = buf->next;
748                 buf_discard(buf);
749                 buf = next;
750         }
751         if (!list_empty(&l_ptr->waiting_ports))
752                 tipc_link_wakeup_ports(l_ptr, 1);
753
754         l_ptr->retransm_queue_head = 0;
755         l_ptr->retransm_queue_size = 0;
756         l_ptr->last_out = NULL;
757         l_ptr->first_out = NULL;
758         l_ptr->next_out = NULL;
759         l_ptr->unacked_window = 0;
760         l_ptr->checkpoint = 1;
761         l_ptr->next_out_no = 1;
762         l_ptr->deferred_inqueue_sz = 0;
763         l_ptr->oldest_deferred_in = NULL;
764         l_ptr->newest_deferred_in = NULL;
765         l_ptr->fsm_msg_cnt = 0;
766         l_ptr->stale_count = 0;
767         link_reset_statistics(l_ptr);
768
769         link_send_event(tipc_cfg_link_event, l_ptr, 0);
770         if (!in_own_cluster(l_ptr->addr))
771                 link_send_event(tipc_disc_link_event, l_ptr, 0);
772 }
773
774
775 static void link_activate(struct link *l_ptr)
776 {
777         l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
778         tipc_node_link_up(l_ptr->owner, l_ptr);
779         tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
780         link_send_event(tipc_cfg_link_event, l_ptr, 1);
781         if (!in_own_cluster(l_ptr->addr))
782                 link_send_event(tipc_disc_link_event, l_ptr, 1);
783 }
784
785 /**
786  * link_state_event - link finite state machine
787  * @l_ptr: pointer to link
788  * @event: state machine event to process
789  */
790
791 static void link_state_event(struct link *l_ptr, unsigned event)
792 {
793         struct link *other;
794         u32 cont_intv = l_ptr->continuity_interval;
795
796         if (!l_ptr->started && (event != STARTING_EVT))
797                 return;         /* Not yet. */
798
799         if (link_blocked(l_ptr)) {
800                 if (event == TIMEOUT_EVT) {
801                         link_set_timer(l_ptr, cont_intv);
802                 }
803                 return;   /* Changeover going on */
804         }
805         dbg_link("STATE_EV: <%s> ", l_ptr->name);
806
807         switch (l_ptr->state) {
808         case WORKING_WORKING:
809                 dbg_link("WW/");
810                 switch (event) {
811                 case TRAFFIC_MSG_EVT:
812                         dbg_link("TRF-");
813                         /* fall through */
814                 case ACTIVATE_MSG:
815                         dbg_link("ACT\n");
816                         break;
817                 case TIMEOUT_EVT:
818                         dbg_link("TIM ");
819                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
820                                 l_ptr->checkpoint = l_ptr->next_in_no;
821                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
822                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
823                                                                  0, 0, 0, 0, 0);
824                                         l_ptr->fsm_msg_cnt++;
825                                 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
826                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
827                                                                  1, 0, 0, 0, 0);
828                                         l_ptr->fsm_msg_cnt++;
829                                 }
830                                 link_set_timer(l_ptr, cont_intv);
831                                 break;
832                         }
833                         dbg_link(" -> WU\n");
834                         l_ptr->state = WORKING_UNKNOWN;
835                         l_ptr->fsm_msg_cnt = 0;
836                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
837                         l_ptr->fsm_msg_cnt++;
838                         link_set_timer(l_ptr, cont_intv / 4);
839                         break;
840                 case RESET_MSG:
841                         dbg_link("RES -> RR\n");
842                         info("Resetting link <%s>, requested by peer\n",
843                              l_ptr->name);
844                         tipc_link_reset(l_ptr);
845                         l_ptr->state = RESET_RESET;
846                         l_ptr->fsm_msg_cnt = 0;
847                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
848                         l_ptr->fsm_msg_cnt++;
849                         link_set_timer(l_ptr, cont_intv);
850                         break;
851                 default:
852                         err("Unknown link event %u in WW state\n", event);
853                 }
854                 break;
855         case WORKING_UNKNOWN:
856                 dbg_link("WU/");
857                 switch (event) {
858                 case TRAFFIC_MSG_EVT:
859                         dbg_link("TRF-");
860                 case ACTIVATE_MSG:
861                         dbg_link("ACT -> WW\n");
862                         l_ptr->state = WORKING_WORKING;
863                         l_ptr->fsm_msg_cnt = 0;
864                         link_set_timer(l_ptr, cont_intv);
865                         break;
866                 case RESET_MSG:
867                         dbg_link("RES -> RR\n");
868                         info("Resetting link <%s>, requested by peer "
869                              "while probing\n", l_ptr->name);
870                         tipc_link_reset(l_ptr);
871                         l_ptr->state = RESET_RESET;
872                         l_ptr->fsm_msg_cnt = 0;
873                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
874                         l_ptr->fsm_msg_cnt++;
875                         link_set_timer(l_ptr, cont_intv);
876                         break;
877                 case TIMEOUT_EVT:
878                         dbg_link("TIM ");
879                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
880                                 dbg_link("-> WW\n");
881                                 l_ptr->state = WORKING_WORKING;
882                                 l_ptr->fsm_msg_cnt = 0;
883                                 l_ptr->checkpoint = l_ptr->next_in_no;
884                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
885                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
886                                                                  0, 0, 0, 0, 0);
887                                         l_ptr->fsm_msg_cnt++;
888                                 }
889                                 link_set_timer(l_ptr, cont_intv);
890                         } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
891                                 dbg_link("Probing %u/%u,timer = %u ms)\n",
892                                          l_ptr->fsm_msg_cnt, l_ptr->abort_limit,
893                                          cont_intv / 4);
894                                 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
895                                                          1, 0, 0, 0, 0);
896                                 l_ptr->fsm_msg_cnt++;
897                                 link_set_timer(l_ptr, cont_intv / 4);
898                         } else {        /* Link has failed */
899                                 dbg_link("-> RU (%u probes unanswered)\n",
900                                          l_ptr->fsm_msg_cnt);
901                                 warn("Resetting link <%s>, peer not responding\n",
902                                      l_ptr->name);
903                                 tipc_link_reset(l_ptr);
904                                 l_ptr->state = RESET_UNKNOWN;
905                                 l_ptr->fsm_msg_cnt = 0;
906                                 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
907                                                          0, 0, 0, 0, 0);
908                                 l_ptr->fsm_msg_cnt++;
909                                 link_set_timer(l_ptr, cont_intv);
910                         }
911                         break;
912                 default:
913                         err("Unknown link event %u in WU state\n", event);
914                 }
915                 break;
916         case RESET_UNKNOWN:
917                 dbg_link("RU/");
918                 switch (event) {
919                 case TRAFFIC_MSG_EVT:
920                         dbg_link("TRF-\n");
921                         break;
922                 case ACTIVATE_MSG:
923                         other = l_ptr->owner->active_links[0];
924                         if (other && link_working_unknown(other)) {
925                                 dbg_link("ACT\n");
926                                 break;
927                         }
928                         dbg_link("ACT -> WW\n");
929                         l_ptr->state = WORKING_WORKING;
930                         l_ptr->fsm_msg_cnt = 0;
931                         link_activate(l_ptr);
932                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
933                         l_ptr->fsm_msg_cnt++;
934                         link_set_timer(l_ptr, cont_intv);
935                         break;
936                 case RESET_MSG:
937                         dbg_link("RES\n");
938                         dbg_link(" -> RR\n");
939                         l_ptr->state = RESET_RESET;
940                         l_ptr->fsm_msg_cnt = 0;
941                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
942                         l_ptr->fsm_msg_cnt++;
943                         link_set_timer(l_ptr, cont_intv);
944                         break;
945                 case STARTING_EVT:
946                         dbg_link("START-");
947                         l_ptr->started = 1;
948                         /* fall through */
949                 case TIMEOUT_EVT:
950                         dbg_link("TIM\n");
951                         tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
952                         l_ptr->fsm_msg_cnt++;
953                         link_set_timer(l_ptr, cont_intv);
954                         break;
955                 default:
956                         err("Unknown link event %u in RU state\n", event);
957                 }
958                 break;
959         case RESET_RESET:
960                 dbg_link("RR/ ");
961                 switch (event) {
962                 case TRAFFIC_MSG_EVT:
963                         dbg_link("TRF-");
964                         /* fall through */
965                 case ACTIVATE_MSG:
966                         other = l_ptr->owner->active_links[0];
967                         if (other && link_working_unknown(other)) {
968                                 dbg_link("ACT\n");
969                                 break;
970                         }
971                         dbg_link("ACT -> WW\n");
972                         l_ptr->state = WORKING_WORKING;
973                         l_ptr->fsm_msg_cnt = 0;
974                         link_activate(l_ptr);
975                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
976                         l_ptr->fsm_msg_cnt++;
977                         link_set_timer(l_ptr, cont_intv);
978                         break;
979                 case RESET_MSG:
980                         dbg_link("RES\n");
981                         break;
982                 case TIMEOUT_EVT:
983                         dbg_link("TIM\n");
984                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
985                         l_ptr->fsm_msg_cnt++;
986                         link_set_timer(l_ptr, cont_intv);
987                         dbg_link("fsm_msg_cnt %u\n", l_ptr->fsm_msg_cnt);
988                         break;
989                 default:
990                         err("Unknown link event %u in RR state\n", event);
991                 }
992                 break;
993         default:
994                 err("Unknown link state %u/%u\n", l_ptr->state, event);
995         }
996 }
997
998 /*
999  * link_bundle_buf(): Append contents of a buffer to
1000  * the tail of an existing one.
1001  */
1002
1003 static int link_bundle_buf(struct link *l_ptr,
1004                            struct sk_buff *bundler,
1005                            struct sk_buff *buf)
1006 {
1007         struct tipc_msg *bundler_msg = buf_msg(bundler);
1008         struct tipc_msg *msg = buf_msg(buf);
1009         u32 size = msg_size(msg);
1010         u32 bundle_size = msg_size(bundler_msg);
1011         u32 to_pos = align(bundle_size);
1012         u32 pad = to_pos - bundle_size;
1013
1014         if (msg_user(bundler_msg) != MSG_BUNDLER)
1015                 return 0;
1016         if (msg_type(bundler_msg) != OPEN_MSG)
1017                 return 0;
1018         if (skb_tailroom(bundler) < (pad + size))
1019                 return 0;
1020         if (link_max_pkt(l_ptr) < (to_pos + size))
1021                 return 0;
1022
1023         skb_put(bundler, pad + size);
1024         skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
1025         msg_set_size(bundler_msg, to_pos + size);
1026         msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
1027         dbg("Packed msg # %u(%u octets) into pos %u in buf(#%u)\n",
1028             msg_msgcnt(bundler_msg), size, to_pos, msg_seqno(bundler_msg));
1029         msg_dbg(msg, "PACKD:");
1030         buf_discard(buf);
1031         l_ptr->stats.sent_bundled++;
1032         return 1;
1033 }
1034
1035 static void link_add_to_outqueue(struct link *l_ptr,
1036                                  struct sk_buff *buf,
1037                                  struct tipc_msg *msg)
1038 {
1039         u32 ack = mod(l_ptr->next_in_no - 1);
1040         u32 seqno = mod(l_ptr->next_out_no++);
1041
1042         msg_set_word(msg, 2, ((ack << 16) | seqno));
1043         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1044         buf->next = NULL;
1045         if (l_ptr->first_out) {
1046                 l_ptr->last_out->next = buf;
1047                 l_ptr->last_out = buf;
1048         } else
1049                 l_ptr->first_out = l_ptr->last_out = buf;
1050         l_ptr->out_queue_size++;
1051 }
1052
1053 /*
1054  * tipc_link_send_buf() is the 'full path' for messages, called from
1055  * inside TIPC when the 'fast path' in tipc_send_buf
1056  * has failed, and from link_send()
1057  */
1058
1059 int tipc_link_send_buf(struct link *l_ptr, struct sk_buff *buf)
1060 {
1061         struct tipc_msg *msg = buf_msg(buf);
1062         u32 size = msg_size(msg);
1063         u32 dsz = msg_data_sz(msg);
1064         u32 queue_size = l_ptr->out_queue_size;
1065         u32 imp = msg_tot_importance(msg);
1066         u32 queue_limit = l_ptr->queue_limit[imp];
1067         u32 max_packet = link_max_pkt(l_ptr);
1068
1069         msg_set_prevnode(msg, tipc_own_addr);   /* If routed message */
1070
1071         /* Match msg importance against queue limits: */
1072
1073         if (unlikely(queue_size >= queue_limit)) {
1074                 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
1075                         return link_schedule_port(l_ptr, msg_origport(msg),
1076                                                   size);
1077                 }
1078                 msg_dbg(msg, "TIPC: Congestion, throwing away\n");
1079                 buf_discard(buf);
1080                 if (imp > CONN_MANAGER) {
1081                         warn("Resetting link <%s>, send queue full", l_ptr->name);
1082                         tipc_link_reset(l_ptr);
1083                 }
1084                 return dsz;
1085         }
1086
1087         /* Fragmentation needed ? */
1088
1089         if (size > max_packet)
1090                 return tipc_link_send_long_buf(l_ptr, buf);
1091
1092         /* Packet can be queued or sent: */
1093
1094         if (queue_size > l_ptr->stats.max_queue_sz)
1095                 l_ptr->stats.max_queue_sz = queue_size;
1096
1097         if (likely(!tipc_bearer_congested(l_ptr->b_ptr, l_ptr) &&
1098                    !link_congested(l_ptr))) {
1099                 link_add_to_outqueue(l_ptr, buf, msg);
1100
1101                 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr))) {
1102                         l_ptr->unacked_window = 0;
1103                 } else {
1104                         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1105                         l_ptr->stats.bearer_congs++;
1106                         l_ptr->next_out = buf;
1107                 }
1108                 return dsz;
1109         }
1110         /* Congestion: can message be bundled ?: */
1111
1112         if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
1113             (msg_user(msg) != MSG_FRAGMENTER)) {
1114
1115                 /* Try adding message to an existing bundle */
1116
1117                 if (l_ptr->next_out &&
1118                     link_bundle_buf(l_ptr, l_ptr->last_out, buf)) {
1119                         tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1120                         return dsz;
1121                 }
1122
1123                 /* Try creating a new bundle */
1124
1125                 if (size <= max_packet * 2 / 3) {
1126                         struct sk_buff *bundler = buf_acquire(max_packet);
1127                         struct tipc_msg bundler_hdr;
1128
1129                         if (bundler) {
1130                                 msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
1131                                          INT_H_SIZE, l_ptr->addr);
1132                                 skb_copy_to_linear_data(bundler, &bundler_hdr,
1133                                                         INT_H_SIZE);
1134                                 skb_trim(bundler, INT_H_SIZE);
1135                                 link_bundle_buf(l_ptr, bundler, buf);
1136                                 buf = bundler;
1137                                 msg = buf_msg(buf);
1138                                 l_ptr->stats.sent_bundles++;
1139                         }
1140                 }
1141         }
1142         if (!l_ptr->next_out)
1143                 l_ptr->next_out = buf;
1144         link_add_to_outqueue(l_ptr, buf, msg);
1145         tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1146         return dsz;
1147 }
1148
1149 /*
1150  * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
1151  * not been selected yet, and the the owner node is not locked
1152  * Called by TIPC internal users, e.g. the name distributor
1153  */
1154
1155 int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
1156 {
1157         struct link *l_ptr;
1158         struct tipc_node *n_ptr;
1159         int res = -ELINKCONG;
1160
1161         read_lock_bh(&tipc_net_lock);
1162         n_ptr = tipc_node_select(dest, selector);
1163         if (n_ptr) {
1164                 tipc_node_lock(n_ptr);
1165                 l_ptr = n_ptr->active_links[selector & 1];
1166                 if (l_ptr) {
1167                         dbg("tipc_link_send: found link %x for dest %x\n", l_ptr, dest);
1168                         res = tipc_link_send_buf(l_ptr, buf);
1169                 } else {
1170                         dbg("Attempt to send msg to unreachable node:\n");
1171                         msg_dbg(buf_msg(buf),">>>");
1172                         buf_discard(buf);
1173                 }
1174                 tipc_node_unlock(n_ptr);
1175         } else {
1176                 dbg("Attempt to send msg to unknown node:\n");
1177                 msg_dbg(buf_msg(buf),">>>");
1178                 buf_discard(buf);
1179         }
1180         read_unlock_bh(&tipc_net_lock);
1181         return res;
1182 }
1183
1184 /*
1185  * link_send_buf_fast: Entry for data messages where the
1186  * destination link is known and the header is complete,
1187  * inclusive total message length. Very time critical.
1188  * Link is locked. Returns user data length.
1189  */
1190
1191 static int link_send_buf_fast(struct link *l_ptr, struct sk_buff *buf,
1192                               u32 *used_max_pkt)
1193 {
1194         struct tipc_msg *msg = buf_msg(buf);
1195         int res = msg_data_sz(msg);
1196
1197         if (likely(!link_congested(l_ptr))) {
1198                 if (likely(msg_size(msg) <= link_max_pkt(l_ptr))) {
1199                         if (likely(list_empty(&l_ptr->b_ptr->cong_links))) {
1200                                 link_add_to_outqueue(l_ptr, buf, msg);
1201                                 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf,
1202                                                             &l_ptr->media_addr))) {
1203                                         l_ptr->unacked_window = 0;
1204                                         msg_dbg(msg,"SENT_FAST:");
1205                                         return res;
1206                                 }
1207                                 dbg("failed sent fast...\n");
1208                                 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1209                                 l_ptr->stats.bearer_congs++;
1210                                 l_ptr->next_out = buf;
1211                                 return res;
1212                         }
1213                 }
1214                 else
1215                         *used_max_pkt = link_max_pkt(l_ptr);
1216         }
1217         return tipc_link_send_buf(l_ptr, buf);  /* All other cases */
1218 }
1219
1220 /*
1221  * tipc_send_buf_fast: Entry for data messages where the
1222  * destination node is known and the header is complete,
1223  * inclusive total message length.
1224  * Returns user data length.
1225  */
1226 int tipc_send_buf_fast(struct sk_buff *buf, u32 destnode)
1227 {
1228         struct link *l_ptr;
1229         struct tipc_node *n_ptr;
1230         int res;
1231         u32 selector = msg_origport(buf_msg(buf)) & 1;
1232         u32 dummy;
1233
1234         if (destnode == tipc_own_addr)
1235                 return tipc_port_recv_msg(buf);
1236
1237         read_lock_bh(&tipc_net_lock);
1238         n_ptr = tipc_node_select(destnode, selector);
1239         if (likely(n_ptr)) {
1240                 tipc_node_lock(n_ptr);
1241                 l_ptr = n_ptr->active_links[selector];
1242                 dbg("send_fast: buf %x selected %x, destnode = %x\n",
1243                     buf, l_ptr, destnode);
1244                 if (likely(l_ptr)) {
1245                         res = link_send_buf_fast(l_ptr, buf, &dummy);
1246                         tipc_node_unlock(n_ptr);
1247                         read_unlock_bh(&tipc_net_lock);
1248                         return res;
1249                 }
1250                 tipc_node_unlock(n_ptr);
1251         }
1252         read_unlock_bh(&tipc_net_lock);
1253         res = msg_data_sz(buf_msg(buf));
1254         tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1255         return res;
1256 }
1257
1258
1259 /*
1260  * tipc_link_send_sections_fast: Entry for messages where the
1261  * destination processor is known and the header is complete,
1262  * except for total message length.
1263  * Returns user data length or errno.
1264  */
1265 int tipc_link_send_sections_fast(struct port *sender,
1266                                  struct iovec const *msg_sect,
1267                                  const u32 num_sect,
1268                                  u32 destaddr)
1269 {
1270         struct tipc_msg *hdr = &sender->publ.phdr;
1271         struct link *l_ptr;
1272         struct sk_buff *buf;
1273         struct tipc_node *node;
1274         int res;
1275         u32 selector = msg_origport(hdr) & 1;
1276
1277 again:
1278         /*
1279          * Try building message using port's max_pkt hint.
1280          * (Must not hold any locks while building message.)
1281          */
1282
1283         res = msg_build(hdr, msg_sect, num_sect, sender->publ.max_pkt,
1284                         !sender->user_port, &buf);
1285
1286         read_lock_bh(&tipc_net_lock);
1287         node = tipc_node_select(destaddr, selector);
1288         if (likely(node)) {
1289                 tipc_node_lock(node);
1290                 l_ptr = node->active_links[selector];
1291                 if (likely(l_ptr)) {
1292                         if (likely(buf)) {
1293                                 res = link_send_buf_fast(l_ptr, buf,
1294                                                          &sender->publ.max_pkt);
1295                                 if (unlikely(res < 0))
1296                                         buf_discard(buf);
1297 exit:
1298                                 tipc_node_unlock(node);
1299                                 read_unlock_bh(&tipc_net_lock);
1300                                 return res;
1301                         }
1302
1303                         /* Exit if build request was invalid */
1304
1305                         if (unlikely(res < 0))
1306                                 goto exit;
1307
1308                         /* Exit if link (or bearer) is congested */
1309
1310                         if (link_congested(l_ptr) ||
1311                             !list_empty(&l_ptr->b_ptr->cong_links)) {
1312                                 res = link_schedule_port(l_ptr,
1313                                                          sender->publ.ref, res);
1314                                 goto exit;
1315                         }
1316
1317                         /*
1318                          * Message size exceeds max_pkt hint; update hint,
1319                          * then re-try fast path or fragment the message
1320                          */
1321
1322                         sender->publ.max_pkt = link_max_pkt(l_ptr);
1323                         tipc_node_unlock(node);
1324                         read_unlock_bh(&tipc_net_lock);
1325
1326
1327                         if ((msg_hdr_sz(hdr) + res) <= sender->publ.max_pkt)
1328                                 goto again;
1329
1330                         return link_send_sections_long(sender, msg_sect,
1331                                                        num_sect, destaddr);
1332                 }
1333                 tipc_node_unlock(node);
1334         }
1335         read_unlock_bh(&tipc_net_lock);
1336
1337         /* Couldn't find a link to the destination node */
1338
1339         if (buf)
1340                 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1341         if (res >= 0)
1342                 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1343                                                  TIPC_ERR_NO_NODE);
1344         return res;
1345 }
1346
1347 /*
1348  * link_send_sections_long(): Entry for long messages where the
1349  * destination node is known and the header is complete,
1350  * inclusive total message length.
1351  * Link and bearer congestion status have been checked to be ok,
1352  * and are ignored if they change.
1353  *
1354  * Note that fragments do not use the full link MTU so that they won't have
1355  * to undergo refragmentation if link changeover causes them to be sent
1356  * over another link with an additional tunnel header added as prefix.
1357  * (Refragmentation will still occur if the other link has a smaller MTU.)
1358  *
1359  * Returns user data length or errno.
1360  */
1361 static int link_send_sections_long(struct port *sender,
1362                                    struct iovec const *msg_sect,
1363                                    u32 num_sect,
1364                                    u32 destaddr)
1365 {
1366         struct link *l_ptr;
1367         struct tipc_node *node;
1368         struct tipc_msg *hdr = &sender->publ.phdr;
1369         u32 dsz = msg_data_sz(hdr);
1370         u32 max_pkt,fragm_sz,rest;
1371         struct tipc_msg fragm_hdr;
1372         struct sk_buff *buf,*buf_chain,*prev;
1373         u32 fragm_crs,fragm_rest,hsz,sect_rest;
1374         const unchar *sect_crs;
1375         int curr_sect;
1376         u32 fragm_no;
1377
1378 again:
1379         fragm_no = 1;
1380         max_pkt = sender->publ.max_pkt - INT_H_SIZE;
1381                 /* leave room for tunnel header in case of link changeover */
1382         fragm_sz = max_pkt - INT_H_SIZE;
1383                 /* leave room for fragmentation header in each fragment */
1384         rest = dsz;
1385         fragm_crs = 0;
1386         fragm_rest = 0;
1387         sect_rest = 0;
1388         sect_crs = NULL;
1389         curr_sect = -1;
1390
1391         /* Prepare reusable fragment header: */
1392
1393         msg_dbg(hdr, ">FRAGMENTING>");
1394         msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1395                  INT_H_SIZE, msg_destnode(hdr));
1396         msg_set_link_selector(&fragm_hdr, sender->publ.ref);
1397         msg_set_size(&fragm_hdr, max_pkt);
1398         msg_set_fragm_no(&fragm_hdr, 1);
1399
1400         /* Prepare header of first fragment: */
1401
1402         buf_chain = buf = buf_acquire(max_pkt);
1403         if (!buf)
1404                 return -ENOMEM;
1405         buf->next = NULL;
1406         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1407         hsz = msg_hdr_sz(hdr);
1408         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1409         msg_dbg(buf_msg(buf), ">BUILD>");
1410
1411         /* Chop up message: */
1412
1413         fragm_crs = INT_H_SIZE + hsz;
1414         fragm_rest = fragm_sz - hsz;
1415
1416         do {            /* For all sections */
1417                 u32 sz;
1418
1419                 if (!sect_rest) {
1420                         sect_rest = msg_sect[++curr_sect].iov_len;
1421                         sect_crs = (const unchar *)msg_sect[curr_sect].iov_base;
1422                 }
1423
1424                 if (sect_rest < fragm_rest)
1425                         sz = sect_rest;
1426                 else
1427                         sz = fragm_rest;
1428
1429                 if (likely(!sender->user_port)) {
1430                         if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1431 error:
1432                                 for (; buf_chain; buf_chain = buf) {
1433                                         buf = buf_chain->next;
1434                                         buf_discard(buf_chain);
1435                                 }
1436                                 return -EFAULT;
1437                         }
1438                 } else
1439                         skb_copy_to_linear_data_offset(buf, fragm_crs,
1440                                                        sect_crs, sz);
1441                 sect_crs += sz;
1442                 sect_rest -= sz;
1443                 fragm_crs += sz;
1444                 fragm_rest -= sz;
1445                 rest -= sz;
1446
1447                 if (!fragm_rest && rest) {
1448
1449                         /* Initiate new fragment: */
1450                         if (rest <= fragm_sz) {
1451                                 fragm_sz = rest;
1452                                 msg_set_type(&fragm_hdr,LAST_FRAGMENT);
1453                         } else {
1454                                 msg_set_type(&fragm_hdr, FRAGMENT);
1455                         }
1456                         msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1457                         msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1458                         prev = buf;
1459                         buf = buf_acquire(fragm_sz + INT_H_SIZE);
1460                         if (!buf)
1461                                 goto error;
1462
1463                         buf->next = NULL;
1464                         prev->next = buf;
1465                         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1466                         fragm_crs = INT_H_SIZE;
1467                         fragm_rest = fragm_sz;
1468                         msg_dbg(buf_msg(buf),"  >BUILD>");
1469                 }
1470         }
1471         while (rest > 0);
1472
1473         /*
1474          * Now we have a buffer chain. Select a link and check
1475          * that packet size is still OK
1476          */
1477         node = tipc_node_select(destaddr, sender->publ.ref & 1);
1478         if (likely(node)) {
1479                 tipc_node_lock(node);
1480                 l_ptr = node->active_links[sender->publ.ref & 1];
1481                 if (!l_ptr) {
1482                         tipc_node_unlock(node);
1483                         goto reject;
1484                 }
1485                 if (link_max_pkt(l_ptr) < max_pkt) {
1486                         sender->publ.max_pkt = link_max_pkt(l_ptr);
1487                         tipc_node_unlock(node);
1488                         for (; buf_chain; buf_chain = buf) {
1489                                 buf = buf_chain->next;
1490                                 buf_discard(buf_chain);
1491                         }
1492                         goto again;
1493                 }
1494         } else {
1495 reject:
1496                 for (; buf_chain; buf_chain = buf) {
1497                         buf = buf_chain->next;
1498                         buf_discard(buf_chain);
1499                 }
1500                 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1501                                                  TIPC_ERR_NO_NODE);
1502         }
1503
1504         /* Append whole chain to send queue: */
1505
1506         buf = buf_chain;
1507         l_ptr->long_msg_seq_no = mod(l_ptr->long_msg_seq_no + 1);
1508         if (!l_ptr->next_out)
1509                 l_ptr->next_out = buf_chain;
1510         l_ptr->stats.sent_fragmented++;
1511         while (buf) {
1512                 struct sk_buff *next = buf->next;
1513                 struct tipc_msg *msg = buf_msg(buf);
1514
1515                 l_ptr->stats.sent_fragments++;
1516                 msg_set_long_msgno(msg, l_ptr->long_msg_seq_no);
1517                 link_add_to_outqueue(l_ptr, buf, msg);
1518                 msg_dbg(msg, ">ADD>");
1519                 buf = next;
1520         }
1521
1522         /* Send it, if possible: */
1523
1524         tipc_link_push_queue(l_ptr);
1525         tipc_node_unlock(node);
1526         return dsz;
1527 }
1528
1529 /*
1530  * tipc_link_push_packet: Push one unsent packet to the media
1531  */
1532 u32 tipc_link_push_packet(struct link *l_ptr)
1533 {
1534         struct sk_buff *buf = l_ptr->first_out;
1535         u32 r_q_size = l_ptr->retransm_queue_size;
1536         u32 r_q_head = l_ptr->retransm_queue_head;
1537
1538         /* Step to position where retransmission failed, if any,    */
1539         /* consider that buffers may have been released in meantime */
1540
1541         if (r_q_size && buf) {
1542                 u32 last = lesser(mod(r_q_head + r_q_size),
1543                                   link_last_sent(l_ptr));
1544                 u32 first = msg_seqno(buf_msg(buf));
1545
1546                 while (buf && less(first, r_q_head)) {
1547                         first = mod(first + 1);
1548                         buf = buf->next;
1549                 }
1550                 l_ptr->retransm_queue_head = r_q_head = first;
1551                 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1552         }
1553
1554         /* Continue retransmission now, if there is anything: */
1555
1556         if (r_q_size && buf) {
1557                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1558                 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1559                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1560                         msg_dbg(buf_msg(buf), ">DEF-RETR>");
1561                         l_ptr->retransm_queue_head = mod(++r_q_head);
1562                         l_ptr->retransm_queue_size = --r_q_size;
1563                         l_ptr->stats.retransmitted++;
1564                         return 0;
1565                 } else {
1566                         l_ptr->stats.bearer_congs++;
1567                         msg_dbg(buf_msg(buf), "|>DEF-RETR>");
1568                         return PUSH_FAILED;
1569                 }
1570         }
1571
1572         /* Send deferred protocol message, if any: */
1573
1574         buf = l_ptr->proto_msg_queue;
1575         if (buf) {
1576                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1577                 msg_set_bcast_ack(buf_msg(buf),l_ptr->owner->bclink.last_in);
1578                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1579                         msg_dbg(buf_msg(buf), ">DEF-PROT>");
1580                         l_ptr->unacked_window = 0;
1581                         buf_discard(buf);
1582                         l_ptr->proto_msg_queue = NULL;
1583                         return 0;
1584                 } else {
1585                         msg_dbg(buf_msg(buf), "|>DEF-PROT>");
1586                         l_ptr->stats.bearer_congs++;
1587                         return PUSH_FAILED;
1588                 }
1589         }
1590
1591         /* Send one deferred data message, if send window not full: */
1592
1593         buf = l_ptr->next_out;
1594         if (buf) {
1595                 struct tipc_msg *msg = buf_msg(buf);
1596                 u32 next = msg_seqno(msg);
1597                 u32 first = msg_seqno(buf_msg(l_ptr->first_out));
1598
1599                 if (mod(next - first) < l_ptr->queue_limit[0]) {
1600                         msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1601                         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1602                         if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1603                                 if (msg_user(msg) == MSG_BUNDLER)
1604                                         msg_set_type(msg, CLOSED_MSG);
1605                                 msg_dbg(msg, ">PUSH-DATA>");
1606                                 l_ptr->next_out = buf->next;
1607                                 return 0;
1608                         } else {
1609                                 msg_dbg(msg, "|PUSH-DATA|");
1610                                 l_ptr->stats.bearer_congs++;
1611                                 return PUSH_FAILED;
1612                         }
1613                 }
1614         }
1615         return PUSH_FINISHED;
1616 }
1617
1618 /*
1619  * push_queue(): push out the unsent messages of a link where
1620  *               congestion has abated. Node is locked
1621  */
1622 void tipc_link_push_queue(struct link *l_ptr)
1623 {
1624         u32 res;
1625
1626         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr))
1627                 return;
1628
1629         do {
1630                 res = tipc_link_push_packet(l_ptr);
1631         } while (!res);
1632
1633         if (res == PUSH_FAILED)
1634                 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1635 }
1636
1637 static void link_reset_all(unsigned long addr)
1638 {
1639         struct tipc_node *n_ptr;
1640         char addr_string[16];
1641         u32 i;
1642
1643         read_lock_bh(&tipc_net_lock);
1644         n_ptr = tipc_node_find((u32)addr);
1645         if (!n_ptr) {
1646                 read_unlock_bh(&tipc_net_lock);
1647                 return; /* node no longer exists */
1648         }
1649
1650         tipc_node_lock(n_ptr);
1651
1652         warn("Resetting all links to %s\n",
1653              addr_string_fill(addr_string, n_ptr->addr));
1654
1655         for (i = 0; i < MAX_BEARERS; i++) {
1656                 if (n_ptr->links[i]) {
1657                         link_print(n_ptr->links[i], TIPC_OUTPUT,
1658                                    "Resetting link\n");
1659                         tipc_link_reset(n_ptr->links[i]);
1660                 }
1661         }
1662
1663         tipc_node_unlock(n_ptr);
1664         read_unlock_bh(&tipc_net_lock);
1665 }
1666
1667 static void link_retransmit_failure(struct link *l_ptr, struct sk_buff *buf)
1668 {
1669         struct tipc_msg *msg = buf_msg(buf);
1670
1671         warn("Retransmission failure on link <%s>\n", l_ptr->name);
1672         tipc_msg_dbg(TIPC_OUTPUT, msg, ">RETR-FAIL>");
1673
1674         if (l_ptr->addr) {
1675
1676                 /* Handle failure on standard link */
1677
1678                 link_print(l_ptr, TIPC_OUTPUT, "Resetting link\n");
1679                 tipc_link_reset(l_ptr);
1680
1681         } else {
1682
1683                 /* Handle failure on broadcast link */
1684
1685                 struct tipc_node *n_ptr;
1686                 char addr_string[16];
1687
1688                 tipc_printf(TIPC_OUTPUT, "Msg seq number: %u,  ", msg_seqno(msg));
1689                 tipc_printf(TIPC_OUTPUT, "Outstanding acks: %lu\n",
1690                                      (unsigned long) TIPC_SKB_CB(buf)->handle);
1691
1692                 n_ptr = l_ptr->owner->next;
1693                 tipc_node_lock(n_ptr);
1694
1695                 addr_string_fill(addr_string, n_ptr->addr);
1696                 tipc_printf(TIPC_OUTPUT, "Multicast link info for %s\n", addr_string);
1697                 tipc_printf(TIPC_OUTPUT, "Supported: %d,  ", n_ptr->bclink.supported);
1698                 tipc_printf(TIPC_OUTPUT, "Acked: %u\n", n_ptr->bclink.acked);
1699                 tipc_printf(TIPC_OUTPUT, "Last in: %u,  ", n_ptr->bclink.last_in);
1700                 tipc_printf(TIPC_OUTPUT, "Gap after: %u,  ", n_ptr->bclink.gap_after);
1701                 tipc_printf(TIPC_OUTPUT, "Gap to: %u\n", n_ptr->bclink.gap_to);
1702                 tipc_printf(TIPC_OUTPUT, "Nack sync: %u\n\n", n_ptr->bclink.nack_sync);
1703
1704                 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1705
1706                 tipc_node_unlock(n_ptr);
1707
1708                 l_ptr->stale_count = 0;
1709         }
1710 }
1711
1712 void tipc_link_retransmit(struct link *l_ptr, struct sk_buff *buf,
1713                           u32 retransmits)
1714 {
1715         struct tipc_msg *msg;
1716
1717         if (!buf)
1718                 return;
1719
1720         msg = buf_msg(buf);
1721
1722         dbg("Retransmitting %u in link %x\n", retransmits, l_ptr);
1723
1724         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1725                 if (l_ptr->retransm_queue_size == 0) {
1726                         msg_dbg(msg, ">NO_RETR->BCONG>");
1727                         dbg_print_link(l_ptr, "   ");
1728                         l_ptr->retransm_queue_head = msg_seqno(msg);
1729                         l_ptr->retransm_queue_size = retransmits;
1730                 } else {
1731                         err("Unexpected retransmit on link %s (qsize=%d)\n",
1732                             l_ptr->name, l_ptr->retransm_queue_size);
1733                 }
1734                 return;
1735         } else {
1736                 /* Detect repeated retransmit failures on uncongested bearer */
1737
1738                 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1739                         if (++l_ptr->stale_count > 100) {
1740                                 link_retransmit_failure(l_ptr, buf);
1741                                 return;
1742                         }
1743                 } else {
1744                         l_ptr->last_retransmitted = msg_seqno(msg);
1745                         l_ptr->stale_count = 1;
1746                 }
1747         }
1748
1749         while (retransmits && (buf != l_ptr->next_out) && buf) {
1750                 msg = buf_msg(buf);
1751                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1752                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1753                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1754                         msg_dbg(buf_msg(buf), ">RETR>");
1755                         buf = buf->next;
1756                         retransmits--;
1757                         l_ptr->stats.retransmitted++;
1758                 } else {
1759                         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1760                         l_ptr->stats.bearer_congs++;
1761                         l_ptr->retransm_queue_head = msg_seqno(buf_msg(buf));
1762                         l_ptr->retransm_queue_size = retransmits;
1763                         return;
1764                 }
1765         }
1766
1767         l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1768 }
1769
1770 /**
1771  * link_insert_deferred_queue - insert deferred messages back into receive chain
1772  */
1773
1774 static struct sk_buff *link_insert_deferred_queue(struct link *l_ptr,
1775                                                   struct sk_buff *buf)
1776 {
1777         u32 seq_no;
1778
1779         if (l_ptr->oldest_deferred_in == NULL)
1780                 return buf;
1781
1782         seq_no = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1783         if (seq_no == mod(l_ptr->next_in_no)) {
1784                 l_ptr->newest_deferred_in->next = buf;
1785                 buf = l_ptr->oldest_deferred_in;
1786                 l_ptr->oldest_deferred_in = NULL;
1787                 l_ptr->deferred_inqueue_sz = 0;
1788         }
1789         return buf;
1790 }
1791
1792 /**
1793  * link_recv_buf_validate - validate basic format of received message
1794  *
1795  * This routine ensures a TIPC message has an acceptable header, and at least
1796  * as much data as the header indicates it should.  The routine also ensures
1797  * that the entire message header is stored in the main fragment of the message
1798  * buffer, to simplify future access to message header fields.
1799  *
1800  * Note: Having extra info present in the message header or data areas is OK.
1801  * TIPC will ignore the excess, under the assumption that it is optional info
1802  * introduced by a later release of the protocol.
1803  */
1804
1805 static int link_recv_buf_validate(struct sk_buff *buf)
1806 {
1807         static u32 min_data_hdr_size[8] = {
1808                 SHORT_H_SIZE, MCAST_H_SIZE, LONG_H_SIZE, DIR_MSG_H_SIZE,
1809                 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1810                 };
1811
1812         struct tipc_msg *msg;
1813         u32 tipc_hdr[2];
1814         u32 size;
1815         u32 hdr_size;
1816         u32 min_hdr_size;
1817
1818         if (unlikely(buf->len < MIN_H_SIZE))
1819                 return 0;
1820
1821         msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1822         if (msg == NULL)
1823                 return 0;
1824
1825         if (unlikely(msg_version(msg) != TIPC_VERSION))
1826                 return 0;
1827
1828         size = msg_size(msg);
1829         hdr_size = msg_hdr_sz(msg);
1830         min_hdr_size = msg_isdata(msg) ?
1831                 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1832
1833         if (unlikely((hdr_size < min_hdr_size) ||
1834                      (size < hdr_size) ||
1835                      (buf->len < size) ||
1836                      (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1837                 return 0;
1838
1839         return pskb_may_pull(buf, hdr_size);
1840 }
1841
1842 void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *tb_ptr)
1843 {
1844         read_lock_bh(&tipc_net_lock);
1845         while (head) {
1846                 struct bearer *b_ptr = (struct bearer *)tb_ptr;
1847                 struct tipc_node *n_ptr;
1848                 struct link *l_ptr;
1849                 struct sk_buff *crs;
1850                 struct sk_buff *buf = head;
1851                 struct tipc_msg *msg;
1852                 u32 seq_no;
1853                 u32 ackd;
1854                 u32 released = 0;
1855                 int type;
1856
1857                 head = head->next;
1858
1859                 /* Ensure message is well-formed */
1860
1861                 if (unlikely(!link_recv_buf_validate(buf)))
1862                         goto cont;
1863
1864                 /* Ensure message data is a single contiguous unit */
1865
1866                 if (unlikely(buf_linearize(buf))) {
1867                         goto cont;
1868                 }
1869
1870                 /* Handle arrival of a non-unicast link message */
1871
1872                 msg = buf_msg(buf);
1873
1874                 if (unlikely(msg_non_seq(msg))) {
1875                         if (msg_user(msg) ==  LINK_CONFIG)
1876                                 tipc_disc_recv_msg(buf, b_ptr);
1877                         else
1878                                 tipc_bclink_recv_pkt(buf);
1879                         continue;
1880                 }
1881
1882                 if (unlikely(!msg_short(msg) &&
1883                              (msg_destnode(msg) != tipc_own_addr)))
1884                         goto cont;
1885
1886                 /* Discard non-routeable messages destined for another node */
1887
1888                 if (unlikely(!msg_isdata(msg) &&
1889                              (msg_destnode(msg) != tipc_own_addr))) {
1890                         if ((msg_user(msg) != CONN_MANAGER) &&
1891                             (msg_user(msg) != MSG_FRAGMENTER))
1892                                 goto cont;
1893                 }
1894
1895                 /* Locate unicast link endpoint that should handle message */
1896
1897                 n_ptr = tipc_node_find(msg_prevnode(msg));
1898                 if (unlikely(!n_ptr))
1899                         goto cont;
1900                 tipc_node_lock(n_ptr);
1901
1902                 l_ptr = n_ptr->links[b_ptr->identity];
1903                 if (unlikely(!l_ptr)) {
1904                         tipc_node_unlock(n_ptr);
1905                         goto cont;
1906                 }
1907
1908                 /* Validate message sequence number info */
1909
1910                 seq_no = msg_seqno(msg);
1911                 ackd = msg_ack(msg);
1912
1913                 /* Release acked messages */
1914
1915                 if (less(n_ptr->bclink.acked, msg_bcast_ack(msg))) {
1916                         if (tipc_node_is_up(n_ptr) && n_ptr->bclink.supported)
1917                                 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1918                 }
1919
1920                 crs = l_ptr->first_out;
1921                 while ((crs != l_ptr->next_out) &&
1922                        less_eq(msg_seqno(buf_msg(crs)), ackd)) {
1923                         struct sk_buff *next = crs->next;
1924
1925                         buf_discard(crs);
1926                         crs = next;
1927                         released++;
1928                 }
1929                 if (released) {
1930                         l_ptr->first_out = crs;
1931                         l_ptr->out_queue_size -= released;
1932                 }
1933
1934                 /* Try sending any messages link endpoint has pending */
1935
1936                 if (unlikely(l_ptr->next_out))
1937                         tipc_link_push_queue(l_ptr);
1938                 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1939                         tipc_link_wakeup_ports(l_ptr, 0);
1940                 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1941                         l_ptr->stats.sent_acks++;
1942                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1943                 }
1944
1945                 /* Now (finally!) process the incoming message */
1946
1947 protocol_check:
1948                 if (likely(link_working_working(l_ptr))) {
1949                         if (likely(seq_no == mod(l_ptr->next_in_no))) {
1950                                 l_ptr->next_in_no++;
1951                                 if (unlikely(l_ptr->oldest_deferred_in))
1952                                         head = link_insert_deferred_queue(l_ptr,
1953                                                                           head);
1954                                 if (likely(msg_is_dest(msg, tipc_own_addr))) {
1955 deliver:
1956                                         if (likely(msg_isdata(msg))) {
1957                                                 tipc_node_unlock(n_ptr);
1958                                                 tipc_port_recv_msg(buf);
1959                                                 continue;
1960                                         }
1961                                         switch (msg_user(msg)) {
1962                                         case MSG_BUNDLER:
1963                                                 l_ptr->stats.recv_bundles++;
1964                                                 l_ptr->stats.recv_bundled +=
1965                                                         msg_msgcnt(msg);
1966                                                 tipc_node_unlock(n_ptr);
1967                                                 tipc_link_recv_bundle(buf);
1968                                                 continue;
1969                                         case ROUTE_DISTRIBUTOR:
1970                                                 tipc_node_unlock(n_ptr);
1971                                                 tipc_cltr_recv_routing_table(buf);
1972                                                 continue;
1973                                         case NAME_DISTRIBUTOR:
1974                                                 tipc_node_unlock(n_ptr);
1975                                                 tipc_named_recv(buf);
1976                                                 continue;
1977                                         case CONN_MANAGER:
1978                                                 tipc_node_unlock(n_ptr);
1979                                                 tipc_port_recv_proto_msg(buf);
1980                                                 continue;
1981                                         case MSG_FRAGMENTER:
1982                                                 l_ptr->stats.recv_fragments++;
1983                                                 if (tipc_link_recv_fragment(&l_ptr->defragm_buf,
1984                                                                             &buf, &msg)) {
1985                                                         l_ptr->stats.recv_fragmented++;
1986                                                         goto deliver;
1987                                                 }
1988                                                 break;
1989                                         case CHANGEOVER_PROTOCOL:
1990                                                 type = msg_type(msg);
1991                                                 if (link_recv_changeover_msg(&l_ptr, &buf)) {
1992                                                         msg = buf_msg(buf);
1993                                                         seq_no = msg_seqno(msg);
1994                                                         if (type == ORIGINAL_MSG)
1995                                                                 goto deliver;
1996                                                         goto protocol_check;
1997                                                 }
1998                                                 break;
1999                                         }
2000                                 }
2001                                 tipc_node_unlock(n_ptr);
2002                                 tipc_net_route_msg(buf);
2003                                 continue;
2004                         }
2005                         link_handle_out_of_seq_msg(l_ptr, buf);
2006                         head = link_insert_deferred_queue(l_ptr, head);
2007                         tipc_node_unlock(n_ptr);
2008                         continue;
2009                 }
2010
2011                 if (msg_user(msg) == LINK_PROTOCOL) {
2012                         link_recv_proto_msg(l_ptr, buf);
2013                         head = link_insert_deferred_queue(l_ptr, head);
2014                         tipc_node_unlock(n_ptr);
2015                         continue;
2016                 }
2017                 msg_dbg(msg,"NSEQ<REC<");
2018                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2019
2020                 if (link_working_working(l_ptr)) {
2021                         /* Re-insert in front of queue */
2022                         msg_dbg(msg,"RECV-REINS:");
2023                         buf->next = head;
2024                         head = buf;
2025                         tipc_node_unlock(n_ptr);
2026                         continue;
2027                 }
2028                 tipc_node_unlock(n_ptr);
2029 cont:
2030                 buf_discard(buf);
2031         }
2032         read_unlock_bh(&tipc_net_lock);
2033 }
2034
2035 /*
2036  * link_defer_buf(): Sort a received out-of-sequence packet
2037  *                   into the deferred reception queue.
2038  * Returns the increase of the queue length,i.e. 0 or 1
2039  */
2040
2041 u32 tipc_link_defer_pkt(struct sk_buff **head,
2042                         struct sk_buff **tail,
2043                         struct sk_buff *buf)
2044 {
2045         struct sk_buff *prev = NULL;
2046         struct sk_buff *crs = *head;
2047         u32 seq_no = msg_seqno(buf_msg(buf));
2048
2049         buf->next = NULL;
2050
2051         /* Empty queue ? */
2052         if (*head == NULL) {
2053                 *head = *tail = buf;
2054                 return 1;
2055         }
2056
2057         /* Last ? */
2058         if (less(msg_seqno(buf_msg(*tail)), seq_no)) {
2059                 (*tail)->next = buf;
2060                 *tail = buf;
2061                 return 1;
2062         }
2063
2064         /* Scan through queue and sort it in */
2065         do {
2066                 struct tipc_msg *msg = buf_msg(crs);
2067
2068                 if (less(seq_no, msg_seqno(msg))) {
2069                         buf->next = crs;
2070                         if (prev)
2071                                 prev->next = buf;
2072                         else
2073                                 *head = buf;
2074                         return 1;
2075                 }
2076                 if (seq_no == msg_seqno(msg)) {
2077                         break;
2078                 }
2079                 prev = crs;
2080                 crs = crs->next;
2081         }
2082         while (crs);
2083
2084         /* Message is a duplicate of an existing message */
2085
2086         buf_discard(buf);
2087         return 0;
2088 }
2089
2090 /**
2091  * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
2092  */
2093
2094 static void link_handle_out_of_seq_msg(struct link *l_ptr,
2095                                        struct sk_buff *buf)
2096 {
2097         u32 seq_no = msg_seqno(buf_msg(buf));
2098
2099         if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
2100                 link_recv_proto_msg(l_ptr, buf);
2101                 return;
2102         }
2103
2104         dbg("rx OOS msg: seq_no %u, expecting %u (%u)\n",
2105             seq_no, mod(l_ptr->next_in_no), l_ptr->next_in_no);
2106
2107         /* Record OOS packet arrival (force mismatch on next timeout) */
2108
2109         l_ptr->checkpoint--;
2110
2111         /*
2112          * Discard packet if a duplicate; otherwise add it to deferred queue
2113          * and notify peer of gap as per protocol specification
2114          */
2115
2116         if (less(seq_no, mod(l_ptr->next_in_no))) {
2117                 l_ptr->stats.duplicates++;
2118                 buf_discard(buf);
2119                 return;
2120         }
2121
2122         if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
2123                                 &l_ptr->newest_deferred_in, buf)) {
2124                 l_ptr->deferred_inqueue_sz++;
2125                 l_ptr->stats.deferred_recv++;
2126                 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
2127                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
2128         } else
2129                 l_ptr->stats.duplicates++;
2130 }
2131
2132 /*
2133  * Send protocol message to the other endpoint.
2134  */
2135 void tipc_link_send_proto_msg(struct link *l_ptr, u32 msg_typ, int probe_msg,
2136                               u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
2137 {
2138         struct sk_buff *buf = NULL;
2139         struct tipc_msg *msg = l_ptr->pmsg;
2140         u32 msg_size = sizeof(l_ptr->proto_msg);
2141
2142         if (link_blocked(l_ptr))
2143                 return;
2144         msg_set_type(msg, msg_typ);
2145         msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
2146         msg_set_bcast_ack(msg, mod(l_ptr->owner->bclink.last_in));
2147         msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
2148
2149         if (msg_typ == STATE_MSG) {
2150                 u32 next_sent = mod(l_ptr->next_out_no);
2151
2152                 if (!tipc_link_is_up(l_ptr))
2153                         return;
2154                 if (l_ptr->next_out)
2155                         next_sent = msg_seqno(buf_msg(l_ptr->next_out));
2156                 msg_set_next_sent(msg, next_sent);
2157                 if (l_ptr->oldest_deferred_in) {
2158                         u32 rec = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
2159                         gap = mod(rec - mod(l_ptr->next_in_no));
2160                 }
2161                 msg_set_seq_gap(msg, gap);
2162                 if (gap)
2163                         l_ptr->stats.sent_nacks++;
2164                 msg_set_link_tolerance(msg, tolerance);
2165                 msg_set_linkprio(msg, priority);
2166                 msg_set_max_pkt(msg, ack_mtu);
2167                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
2168                 msg_set_probe(msg, probe_msg != 0);
2169                 if (probe_msg) {
2170                         u32 mtu = l_ptr->max_pkt;
2171
2172                         if ((mtu < l_ptr->max_pkt_target) &&
2173                             link_working_working(l_ptr) &&
2174                             l_ptr->fsm_msg_cnt) {
2175                                 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2176                                 if (l_ptr->max_pkt_probes == 10) {
2177                                         l_ptr->max_pkt_target = (msg_size - 4);
2178                                         l_ptr->max_pkt_probes = 0;
2179                                         msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2180                                 }
2181                                 l_ptr->max_pkt_probes++;
2182                         }
2183
2184                         l_ptr->stats.sent_probes++;
2185                 }
2186                 l_ptr->stats.sent_states++;
2187         } else {                /* RESET_MSG or ACTIVATE_MSG */
2188                 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
2189                 msg_set_seq_gap(msg, 0);
2190                 msg_set_next_sent(msg, 1);
2191                 msg_set_link_tolerance(msg, l_ptr->tolerance);
2192                 msg_set_linkprio(msg, l_ptr->priority);
2193                 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
2194         }
2195
2196         if (tipc_node_has_redundant_links(l_ptr->owner)) {
2197                 msg_set_redundant_link(msg);
2198         } else {
2199                 msg_clear_redundant_link(msg);
2200         }
2201         msg_set_linkprio(msg, l_ptr->priority);
2202
2203         /* Ensure sequence number will not fit : */
2204
2205         msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
2206
2207         /* Congestion? */
2208
2209         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
2210                 if (!l_ptr->proto_msg_queue) {
2211                         l_ptr->proto_msg_queue =
2212                                 buf_acquire(sizeof(l_ptr->proto_msg));
2213                 }
2214                 buf = l_ptr->proto_msg_queue;
2215                 if (!buf)
2216                         return;
2217                 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2218                 return;
2219         }
2220         msg_set_timestamp(msg, jiffies_to_msecs(jiffies));
2221
2222         /* Message can be sent */
2223
2224         msg_dbg(msg, ">>");
2225
2226         buf = buf_acquire(msg_size);
2227         if (!buf)
2228                 return;
2229
2230         skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2231         msg_set_size(buf_msg(buf), msg_size);
2232
2233         if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
2234                 l_ptr->unacked_window = 0;
2235                 buf_discard(buf);
2236                 return;
2237         }
2238
2239         /* New congestion */
2240         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
2241         l_ptr->proto_msg_queue = buf;
2242         l_ptr->stats.bearer_congs++;
2243 }
2244
2245 /*
2246  * Receive protocol message :
2247  * Note that network plane id propagates through the network, and may
2248  * change at any time. The node with lowest address rules
2249  */
2250
2251 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf)
2252 {
2253         u32 rec_gap = 0;
2254         u32 max_pkt_info;
2255         u32 max_pkt_ack;
2256         u32 msg_tol;
2257         struct tipc_msg *msg = buf_msg(buf);
2258
2259         dbg("AT(%u):", jiffies_to_msecs(jiffies));
2260         msg_dbg(msg, "<<");
2261         if (link_blocked(l_ptr))
2262                 goto exit;
2263
2264         /* record unnumbered packet arrival (force mismatch on next timeout) */
2265
2266         l_ptr->checkpoint--;
2267
2268         if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
2269                 if (tipc_own_addr > msg_prevnode(msg))
2270                         l_ptr->b_ptr->net_plane = msg_net_plane(msg);
2271
2272         l_ptr->owner->permit_changeover = msg_redundant_link(msg);
2273
2274         switch (msg_type(msg)) {
2275
2276         case RESET_MSG:
2277                 if (!link_working_unknown(l_ptr) &&
2278                     (l_ptr->peer_session != INVALID_SESSION)) {
2279                         if (msg_session(msg) == l_ptr->peer_session) {
2280                                 dbg("Duplicate RESET: %u<->%u\n",
2281                                     msg_session(msg), l_ptr->peer_session);
2282                                 break; /* duplicate: ignore */
2283                         }
2284                 }
2285                 /* fall thru' */
2286         case ACTIVATE_MSG:
2287                 /* Update link settings according other endpoint's values */
2288
2289                 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2290
2291                 if ((msg_tol = msg_link_tolerance(msg)) &&
2292                     (msg_tol > l_ptr->tolerance))
2293                         link_set_supervision_props(l_ptr, msg_tol);
2294
2295                 if (msg_linkprio(msg) > l_ptr->priority)
2296                         l_ptr->priority = msg_linkprio(msg);
2297
2298                 max_pkt_info = msg_max_pkt(msg);
2299                 if (max_pkt_info) {
2300                         if (max_pkt_info < l_ptr->max_pkt_target)
2301                                 l_ptr->max_pkt_target = max_pkt_info;
2302                         if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2303                                 l_ptr->max_pkt = l_ptr->max_pkt_target;
2304                 } else {
2305                         l_ptr->max_pkt = l_ptr->max_pkt_target;
2306                 }
2307                 l_ptr->owner->bclink.supported = (max_pkt_info != 0);
2308
2309                 link_state_event(l_ptr, msg_type(msg));
2310
2311                 l_ptr->peer_session = msg_session(msg);
2312                 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2313
2314                 /* Synchronize broadcast sequence numbers */
2315                 if (!tipc_node_has_redundant_links(l_ptr->owner)) {
2316                         l_ptr->owner->bclink.last_in = mod(msg_last_bcast(msg));
2317                 }
2318                 break;
2319         case STATE_MSG:
2320
2321                 if ((msg_tol = msg_link_tolerance(msg)))
2322                         link_set_supervision_props(l_ptr, msg_tol);
2323
2324                 if (msg_linkprio(msg) &&
2325                     (msg_linkprio(msg) != l_ptr->priority)) {
2326                         warn("Resetting link <%s>, priority change %u->%u\n",
2327                              l_ptr->name, l_ptr->priority, msg_linkprio(msg));
2328                         l_ptr->priority = msg_linkprio(msg);
2329                         tipc_link_reset(l_ptr); /* Enforce change to take effect */
2330                         break;
2331                 }
2332                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2333                 l_ptr->stats.recv_states++;
2334                 if (link_reset_unknown(l_ptr))
2335                         break;
2336
2337                 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2338                         rec_gap = mod(msg_next_sent(msg) -
2339                                       mod(l_ptr->next_in_no));
2340                 }
2341
2342                 max_pkt_ack = msg_max_pkt(msg);
2343                 if (max_pkt_ack > l_ptr->max_pkt) {
2344                         dbg("Link <%s> updated MTU %u -> %u\n",
2345                             l_ptr->name, l_ptr->max_pkt, max_pkt_ack);
2346                         l_ptr->max_pkt = max_pkt_ack;
2347                         l_ptr->max_pkt_probes = 0;
2348                 }
2349
2350                 max_pkt_ack = 0;
2351                 if (msg_probe(msg)) {
2352                         l_ptr->stats.recv_probes++;
2353                         if (msg_size(msg) > sizeof(l_ptr->proto_msg)) {
2354                                 max_pkt_ack = msg_size(msg);
2355                         }
2356                 }
2357
2358                 /* Protocol message before retransmits, reduce loss risk */
2359
2360                 tipc_bclink_check_gap(l_ptr->owner, msg_last_bcast(msg));
2361
2362                 if (rec_gap || (msg_probe(msg))) {
2363                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2364                                                  0, rec_gap, 0, 0, max_pkt_ack);
2365                 }
2366                 if (msg_seq_gap(msg)) {
2367                         msg_dbg(msg, "With Gap:");
2368                         l_ptr->stats.recv_nacks++;
2369                         tipc_link_retransmit(l_ptr, l_ptr->first_out,
2370                                              msg_seq_gap(msg));
2371                 }
2372                 break;
2373         default:
2374                 msg_dbg(buf_msg(buf), "<DISCARDING UNKNOWN<");
2375         }
2376 exit:
2377         buf_discard(buf);
2378 }
2379
2380
2381 /*
2382  * tipc_link_tunnel(): Send one message via a link belonging to
2383  * another bearer. Owner node is locked.
2384  */
2385 void tipc_link_tunnel(struct link *l_ptr,
2386                       struct tipc_msg *tunnel_hdr,
2387                       struct tipc_msg  *msg,
2388                       u32 selector)
2389 {
2390         struct link *tunnel;
2391         struct sk_buff *buf;
2392         u32 length = msg_size(msg);
2393
2394         tunnel = l_ptr->owner->active_links[selector & 1];
2395         if (!tipc_link_is_up(tunnel)) {
2396                 warn("Link changeover error, "
2397                      "tunnel link no longer available\n");
2398                 return;
2399         }
2400         msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2401         buf = buf_acquire(length + INT_H_SIZE);
2402         if (!buf) {
2403                 warn("Link changeover error, "
2404                      "unable to send tunnel msg\n");
2405                 return;
2406         }
2407         skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
2408         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
2409         dbg("%c->%c:", l_ptr->b_ptr->net_plane, tunnel->b_ptr->net_plane);
2410         msg_dbg(buf_msg(buf), ">SEND>");
2411         tipc_link_send_buf(tunnel, buf);
2412 }
2413
2414
2415
2416 /*
2417  * changeover(): Send whole message queue via the remaining link
2418  *               Owner node is locked.
2419  */
2420
2421 void tipc_link_changeover(struct link *l_ptr)
2422 {
2423         u32 msgcount = l_ptr->out_queue_size;
2424         struct sk_buff *crs = l_ptr->first_out;
2425         struct link *tunnel = l_ptr->owner->active_links[0];
2426         struct tipc_msg tunnel_hdr;
2427         int split_bundles;
2428
2429         if (!tunnel)
2430                 return;
2431
2432         if (!l_ptr->owner->permit_changeover) {
2433                 warn("Link changeover error, "
2434                      "peer did not permit changeover\n");
2435                 return;
2436         }
2437
2438         msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2439                  ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
2440         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2441         msg_set_msgcnt(&tunnel_hdr, msgcount);
2442         dbg("Link changeover requires %u tunnel messages\n", msgcount);
2443
2444         if (!l_ptr->first_out) {
2445                 struct sk_buff *buf;
2446
2447                 buf = buf_acquire(INT_H_SIZE);
2448                 if (buf) {
2449                         skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2450                         msg_set_size(&tunnel_hdr, INT_H_SIZE);
2451                         dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2452                             tunnel->b_ptr->net_plane);
2453                         msg_dbg(&tunnel_hdr, "EMPTY>SEND>");
2454                         tipc_link_send_buf(tunnel, buf);
2455                 } else {
2456                         warn("Link changeover error, "
2457                              "unable to send changeover msg\n");
2458                 }
2459                 return;
2460         }
2461
2462         split_bundles = (l_ptr->owner->active_links[0] !=
2463                          l_ptr->owner->active_links[1]);
2464
2465         while (crs) {
2466                 struct tipc_msg *msg = buf_msg(crs);
2467
2468                 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2469                         struct tipc_msg *m = msg_get_wrapped(msg);
2470                         unchar* pos = (unchar*)m;
2471
2472                         msgcount = msg_msgcnt(msg);
2473                         while (msgcount--) {
2474                                 msg_set_seqno(m,msg_seqno(msg));
2475                                 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2476                                                  msg_link_selector(m));
2477                                 pos += align(msg_size(m));
2478                                 m = (struct tipc_msg *)pos;
2479                         }
2480                 } else {
2481                         tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2482                                          msg_link_selector(msg));
2483                 }
2484                 crs = crs->next;
2485         }
2486 }
2487
2488 void tipc_link_send_duplicate(struct link *l_ptr, struct link *tunnel)
2489 {
2490         struct sk_buff *iter;
2491         struct tipc_msg tunnel_hdr;
2492
2493         msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2494                  DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
2495         msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2496         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2497         iter = l_ptr->first_out;
2498         while (iter) {
2499                 struct sk_buff *outbuf;
2500                 struct tipc_msg *msg = buf_msg(iter);
2501                 u32 length = msg_size(msg);
2502
2503                 if (msg_user(msg) == MSG_BUNDLER)
2504                         msg_set_type(msg, CLOSED_MSG);
2505                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));   /* Update */
2506                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2507                 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2508                 outbuf = buf_acquire(length + INT_H_SIZE);
2509                 if (outbuf == NULL) {
2510                         warn("Link changeover error, "
2511                              "unable to send duplicate msg\n");
2512                         return;
2513                 }
2514                 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2515                 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2516                                                length);
2517                 dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2518                     tunnel->b_ptr->net_plane);
2519                 msg_dbg(buf_msg(outbuf), ">SEND>");
2520                 tipc_link_send_buf(tunnel, outbuf);
2521                 if (!tipc_link_is_up(l_ptr))
2522                         return;
2523                 iter = iter->next;
2524         }
2525 }
2526
2527
2528
2529 /**
2530  * buf_extract - extracts embedded TIPC message from another message
2531  * @skb: encapsulating message buffer
2532  * @from_pos: offset to extract from
2533  *
2534  * Returns a new message buffer containing an embedded message.  The
2535  * encapsulating message itself is left unchanged.
2536  */
2537
2538 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2539 {
2540         struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2541         u32 size = msg_size(msg);
2542         struct sk_buff *eb;
2543
2544         eb = buf_acquire(size);
2545         if (eb)
2546                 skb_copy_to_linear_data(eb, msg, size);
2547         return eb;
2548 }
2549
2550 /*
2551  *  link_recv_changeover_msg(): Receive tunneled packet sent
2552  *  via other link. Node is locked. Return extracted buffer.
2553  */
2554
2555 static int link_recv_changeover_msg(struct link **l_ptr,
2556                                     struct sk_buff **buf)
2557 {
2558         struct sk_buff *tunnel_buf = *buf;
2559         struct link *dest_link;
2560         struct tipc_msg *msg;
2561         struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2562         u32 msg_typ = msg_type(tunnel_msg);
2563         u32 msg_count = msg_msgcnt(tunnel_msg);
2564
2565         dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2566         if (!dest_link) {
2567                 msg_dbg(tunnel_msg, "NOLINK/<REC<");
2568                 goto exit;
2569         }
2570         if (dest_link == *l_ptr) {
2571                 err("Unexpected changeover message on link <%s>\n",
2572                     (*l_ptr)->name);
2573                 goto exit;
2574         }
2575         dbg("%c<-%c:", dest_link->b_ptr->net_plane,
2576             (*l_ptr)->b_ptr->net_plane);
2577         *l_ptr = dest_link;
2578         msg = msg_get_wrapped(tunnel_msg);
2579
2580         if (msg_typ == DUPLICATE_MSG) {
2581                 if (less(msg_seqno(msg), mod(dest_link->next_in_no))) {
2582                         msg_dbg(tunnel_msg, "DROP/<REC<");
2583                         goto exit;
2584                 }
2585                 *buf = buf_extract(tunnel_buf,INT_H_SIZE);
2586                 if (*buf == NULL) {
2587                         warn("Link changeover error, duplicate msg dropped\n");
2588                         goto exit;
2589                 }
2590                 msg_dbg(tunnel_msg, "TNL<REC<");
2591                 buf_discard(tunnel_buf);
2592                 return 1;
2593         }
2594
2595         /* First original message ?: */
2596
2597         if (tipc_link_is_up(dest_link)) {
2598                 msg_dbg(tunnel_msg, "UP/FIRST/<REC<");
2599                 info("Resetting link <%s>, changeover initiated by peer\n",
2600                      dest_link->name);
2601                 tipc_link_reset(dest_link);
2602                 dest_link->exp_msg_count = msg_count;
2603                 dbg("Expecting %u tunnelled messages\n", msg_count);
2604                 if (!msg_count)
2605                         goto exit;
2606         } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2607                 msg_dbg(tunnel_msg, "BLK/FIRST/<REC<");
2608                 dest_link->exp_msg_count = msg_count;
2609                 dbg("Expecting %u tunnelled messages\n", msg_count);
2610                 if (!msg_count)
2611                         goto exit;
2612         }
2613
2614         /* Receive original message */
2615
2616         if (dest_link->exp_msg_count == 0) {
2617                 warn("Link switchover error, "
2618                      "got too many tunnelled messages\n");
2619                 msg_dbg(tunnel_msg, "OVERDUE/DROP/<REC<");
2620                 dbg_print_link(dest_link, "LINK:");
2621                 goto exit;
2622         }
2623         dest_link->exp_msg_count--;
2624         if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2625                 msg_dbg(tunnel_msg, "DROP/DUPL/<REC<");
2626                 goto exit;
2627         } else {
2628                 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2629                 if (*buf != NULL) {
2630                         msg_dbg(tunnel_msg, "TNL<REC<");
2631                         buf_discard(tunnel_buf);
2632                         return 1;
2633                 } else {
2634                         warn("Link changeover error, original msg dropped\n");
2635                 }
2636         }
2637 exit:
2638         *buf = NULL;
2639         buf_discard(tunnel_buf);
2640         return 0;
2641 }
2642
2643 /*
2644  *  Bundler functionality:
2645  */
2646 void tipc_link_recv_bundle(struct sk_buff *buf)
2647 {
2648         u32 msgcount = msg_msgcnt(buf_msg(buf));
2649         u32 pos = INT_H_SIZE;
2650         struct sk_buff *obuf;
2651
2652         msg_dbg(buf_msg(buf), "<BNDL<: ");
2653         while (msgcount--) {
2654                 obuf = buf_extract(buf, pos);
2655                 if (obuf == NULL) {
2656                         warn("Link unable to unbundle message(s)\n");
2657                         break;
2658                 }
2659                 pos += align(msg_size(buf_msg(obuf)));
2660                 msg_dbg(buf_msg(obuf), "     /");
2661                 tipc_net_route_msg(obuf);
2662         }
2663         buf_discard(buf);
2664 }
2665
2666 /*
2667  *  Fragmentation/defragmentation:
2668  */
2669
2670
2671 /*
2672  * tipc_link_send_long_buf: Entry for buffers needing fragmentation.
2673  * The buffer is complete, inclusive total message length.
2674  * Returns user data length.
2675  */
2676 int tipc_link_send_long_buf(struct link *l_ptr, struct sk_buff *buf)
2677 {
2678         struct tipc_msg *inmsg = buf_msg(buf);
2679         struct tipc_msg fragm_hdr;
2680         u32 insize = msg_size(inmsg);
2681         u32 dsz = msg_data_sz(inmsg);
2682         unchar *crs = buf->data;
2683         u32 rest = insize;
2684         u32 pack_sz = link_max_pkt(l_ptr);
2685         u32 fragm_sz = pack_sz - INT_H_SIZE;
2686         u32 fragm_no = 1;
2687         u32 destaddr;
2688
2689         if (msg_short(inmsg))
2690                 destaddr = l_ptr->addr;
2691         else
2692                 destaddr = msg_destnode(inmsg);
2693
2694         if (msg_routed(inmsg))
2695                 msg_set_prevnode(inmsg, tipc_own_addr);
2696
2697         /* Prepare reusable fragment header: */
2698
2699         msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2700                  INT_H_SIZE, destaddr);
2701         msg_set_link_selector(&fragm_hdr, msg_link_selector(inmsg));
2702         msg_set_long_msgno(&fragm_hdr, mod(l_ptr->long_msg_seq_no++));
2703         msg_set_fragm_no(&fragm_hdr, fragm_no);
2704         l_ptr->stats.sent_fragmented++;
2705
2706         /* Chop up message: */
2707
2708         while (rest > 0) {
2709                 struct sk_buff *fragm;
2710
2711                 if (rest <= fragm_sz) {
2712                         fragm_sz = rest;
2713                         msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2714                 }
2715                 fragm = buf_acquire(fragm_sz + INT_H_SIZE);
2716                 if (fragm == NULL) {
2717                         warn("Link unable to fragment message\n");
2718                         dsz = -ENOMEM;
2719                         goto exit;
2720                 }
2721                 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2722                 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2723                 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2724                                                fragm_sz);
2725                 /*  Send queued messages first, if any: */
2726
2727                 l_ptr->stats.sent_fragments++;
2728                 tipc_link_send_buf(l_ptr, fragm);
2729                 if (!tipc_link_is_up(l_ptr))
2730                         return dsz;
2731                 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
2732                 rest -= fragm_sz;
2733                 crs += fragm_sz;
2734                 msg_set_type(&fragm_hdr, FRAGMENT);
2735         }
2736 exit:
2737         buf_discard(buf);
2738         return dsz;
2739 }
2740
2741 /*
2742  * A pending message being re-assembled must store certain values
2743  * to handle subsequent fragments correctly. The following functions
2744  * help storing these values in unused, available fields in the
2745  * pending message. This makes dynamic memory allocation unecessary.
2746  */
2747
2748 static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2749 {
2750         msg_set_seqno(buf_msg(buf), seqno);
2751 }
2752
2753 static u32 get_fragm_size(struct sk_buff *buf)
2754 {
2755         return msg_ack(buf_msg(buf));
2756 }
2757
2758 static void set_fragm_size(struct sk_buff *buf, u32 sz)
2759 {
2760         msg_set_ack(buf_msg(buf), sz);
2761 }
2762
2763 static u32 get_expected_frags(struct sk_buff *buf)
2764 {
2765         return msg_bcast_ack(buf_msg(buf));
2766 }
2767
2768 static void set_expected_frags(struct sk_buff *buf, u32 exp)
2769 {
2770         msg_set_bcast_ack(buf_msg(buf), exp);
2771 }
2772
2773 static u32 get_timer_cnt(struct sk_buff *buf)
2774 {
2775         return msg_reroute_cnt(buf_msg(buf));
2776 }
2777
2778 static void incr_timer_cnt(struct sk_buff *buf)
2779 {
2780         msg_incr_reroute_cnt(buf_msg(buf));
2781 }
2782
2783 /*
2784  * tipc_link_recv_fragment(): Called with node lock on. Returns
2785  * the reassembled buffer if message is complete.
2786  */
2787 int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2788                             struct tipc_msg **m)
2789 {
2790         struct sk_buff *prev = NULL;
2791         struct sk_buff *fbuf = *fb;
2792         struct tipc_msg *fragm = buf_msg(fbuf);
2793         struct sk_buff *pbuf = *pending;
2794         u32 long_msg_seq_no = msg_long_msgno(fragm);
2795
2796         *fb = NULL;
2797         msg_dbg(fragm,"FRG<REC<");
2798
2799         /* Is there an incomplete message waiting for this fragment? */
2800
2801         while (pbuf && ((msg_seqno(buf_msg(pbuf)) != long_msg_seq_no) ||
2802                         (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2803                 prev = pbuf;
2804                 pbuf = pbuf->next;
2805         }
2806
2807         if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2808                 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2809                 u32 msg_sz = msg_size(imsg);
2810                 u32 fragm_sz = msg_data_sz(fragm);
2811                 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2812                 u32 max =  TIPC_MAX_USER_MSG_SIZE + LONG_H_SIZE;
2813                 if (msg_type(imsg) == TIPC_MCAST_MSG)
2814                         max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2815                 if (msg_size(imsg) > max) {
2816                         msg_dbg(fragm,"<REC<Oversized: ");
2817                         buf_discard(fbuf);
2818                         return 0;
2819                 }
2820                 pbuf = buf_acquire(msg_size(imsg));
2821                 if (pbuf != NULL) {
2822                         pbuf->next = *pending;
2823                         *pending = pbuf;
2824                         skb_copy_to_linear_data(pbuf, imsg,
2825                                                 msg_data_sz(fragm));
2826                         /*  Prepare buffer for subsequent fragments. */
2827
2828                         set_long_msg_seqno(pbuf, long_msg_seq_no);
2829                         set_fragm_size(pbuf,fragm_sz);
2830                         set_expected_frags(pbuf,exp_fragm_cnt - 1);
2831                 } else {
2832                         warn("Link unable to reassemble fragmented message\n");
2833                 }
2834                 buf_discard(fbuf);
2835                 return 0;
2836         } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2837                 u32 dsz = msg_data_sz(fragm);
2838                 u32 fsz = get_fragm_size(pbuf);
2839                 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2840                 u32 exp_frags = get_expected_frags(pbuf) - 1;
2841                 skb_copy_to_linear_data_offset(pbuf, crs,
2842                                                msg_data(fragm), dsz);
2843                 buf_discard(fbuf);
2844
2845                 /* Is message complete? */
2846
2847                 if (exp_frags == 0) {
2848                         if (prev)
2849                                 prev->next = pbuf->next;
2850                         else
2851                                 *pending = pbuf->next;
2852                         msg_reset_reroute_cnt(buf_msg(pbuf));
2853                         *fb = pbuf;
2854                         *m = buf_msg(pbuf);
2855                         return 1;
2856                 }
2857                 set_expected_frags(pbuf,exp_frags);
2858                 return 0;
2859         }
2860         dbg(" Discarding orphan fragment %x\n",fbuf);
2861         msg_dbg(fragm,"ORPHAN:");
2862         dbg("Pending long buffers:\n");
2863         dbg_print_buf_chain(*pending);
2864         buf_discard(fbuf);
2865         return 0;
2866 }
2867
2868 /**
2869  * link_check_defragm_bufs - flush stale incoming message fragments
2870  * @l_ptr: pointer to link
2871  */
2872
2873 static void link_check_defragm_bufs(struct link *l_ptr)
2874 {
2875         struct sk_buff *prev = NULL;
2876         struct sk_buff *next = NULL;
2877         struct sk_buff *buf = l_ptr->defragm_buf;
2878
2879         if (!buf)
2880                 return;
2881         if (!link_working_working(l_ptr))
2882                 return;
2883         while (buf) {
2884                 u32 cnt = get_timer_cnt(buf);
2885
2886                 next = buf->next;
2887                 if (cnt < 4) {
2888                         incr_timer_cnt(buf);
2889                         prev = buf;
2890                 } else {
2891                         dbg(" Discarding incomplete long buffer\n");
2892                         msg_dbg(buf_msg(buf), "LONG:");
2893                         dbg_print_link(l_ptr, "curr:");
2894                         dbg("Pending long buffers:\n");
2895                         dbg_print_buf_chain(l_ptr->defragm_buf);
2896                         if (prev)
2897                                 prev->next = buf->next;
2898                         else
2899                                 l_ptr->defragm_buf = buf->next;
2900                         buf_discard(buf);
2901                 }
2902                 buf = next;
2903         }
2904 }
2905
2906
2907
2908 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance)
2909 {
2910         l_ptr->tolerance = tolerance;
2911         l_ptr->continuity_interval =
2912                 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2913         l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2914 }
2915
2916
2917 void tipc_link_set_queue_limits(struct link *l_ptr, u32 window)
2918 {
2919         /* Data messages from this node, inclusive FIRST_FRAGM */
2920         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2921         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2922         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2923         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2924         /* Transiting data messages,inclusive FIRST_FRAGM */
2925         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2926         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2927         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2928         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2929         l_ptr->queue_limit[CONN_MANAGER] = 1200;
2930         l_ptr->queue_limit[ROUTE_DISTRIBUTOR] = 1200;
2931         l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2932         l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2933         /* FRAGMENT and LAST_FRAGMENT packets */
2934         l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2935 }
2936
2937 /**
2938  * link_find_link - locate link by name
2939  * @name - ptr to link name string
2940  * @node - ptr to area to be filled with ptr to associated node
2941  *
2942  * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2943  * this also prevents link deletion.
2944  *
2945  * Returns pointer to link (or 0 if invalid link name).
2946  */
2947
2948 static struct link *link_find_link(const char *name, struct tipc_node **node)
2949 {
2950         struct link_name link_name_parts;
2951         struct bearer *b_ptr;
2952         struct link *l_ptr;
2953
2954         if (!link_name_validate(name, &link_name_parts))
2955                 return NULL;
2956
2957         b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2958         if (!b_ptr)
2959                 return NULL;
2960
2961         *node = tipc_node_find(link_name_parts.addr_peer);
2962         if (!*node)
2963                 return NULL;
2964
2965         l_ptr = (*node)->links[b_ptr->identity];
2966         if (!l_ptr || strcmp(l_ptr->name, name))
2967                 return NULL;
2968
2969         return l_ptr;
2970 }
2971
2972 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2973                                      u16 cmd)
2974 {
2975         struct tipc_link_config *args;
2976         u32 new_value;
2977         struct link *l_ptr;
2978         struct tipc_node *node;
2979         int res;
2980
2981         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2982                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2983
2984         args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2985         new_value = ntohl(args->value);
2986
2987         if (!strcmp(args->name, tipc_bclink_name)) {
2988                 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2989                     (tipc_bclink_set_queue_limits(new_value) == 0))
2990                         return tipc_cfg_reply_none();
2991                 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2992                                                    " (cannot change setting on broadcast link)");
2993         }
2994
2995         read_lock_bh(&tipc_net_lock);
2996         l_ptr = link_find_link(args->name, &node);
2997         if (!l_ptr) {
2998                 read_unlock_bh(&tipc_net_lock);
2999                 return tipc_cfg_reply_error_string("link not found");
3000         }
3001
3002         tipc_node_lock(node);
3003         res = -EINVAL;
3004         switch (cmd) {
3005         case TIPC_CMD_SET_LINK_TOL:
3006                 if ((new_value >= TIPC_MIN_LINK_TOL) &&
3007                     (new_value <= TIPC_MAX_LINK_TOL)) {
3008                         link_set_supervision_props(l_ptr, new_value);
3009                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
3010                                                  0, 0, new_value, 0, 0);
3011                         res = 0;
3012                 }
3013                 break;
3014         case TIPC_CMD_SET_LINK_PRI:
3015                 if ((new_value >= TIPC_MIN_LINK_PRI) &&
3016                     (new_value <= TIPC_MAX_LINK_PRI)) {
3017                         l_ptr->priority = new_value;
3018                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
3019                                                  0, 0, 0, new_value, 0);
3020                         res = 0;
3021                 }
3022                 break;
3023         case TIPC_CMD_SET_LINK_WINDOW:
3024                 if ((new_value >= TIPC_MIN_LINK_WIN) &&
3025                     (new_value <= TIPC_MAX_LINK_WIN)) {
3026                         tipc_link_set_queue_limits(l_ptr, new_value);
3027                         res = 0;
3028                 }
3029                 break;
3030         }
3031         tipc_node_unlock(node);
3032
3033         read_unlock_bh(&tipc_net_lock);
3034         if (res)
3035                 return tipc_cfg_reply_error_string("cannot change link setting");
3036
3037         return tipc_cfg_reply_none();
3038 }
3039
3040 /**
3041  * link_reset_statistics - reset link statistics
3042  * @l_ptr: pointer to link
3043  */
3044
3045 static void link_reset_statistics(struct link *l_ptr)
3046 {
3047         memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
3048         l_ptr->stats.sent_info = l_ptr->next_out_no;
3049         l_ptr->stats.recv_info = l_ptr->next_in_no;
3050 }
3051
3052 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
3053 {
3054         char *link_name;
3055         struct link *l_ptr;
3056         struct tipc_node *node;
3057
3058         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3059                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3060
3061         link_name = (char *)TLV_DATA(req_tlv_area);
3062         if (!strcmp(link_name, tipc_bclink_name)) {
3063                 if (tipc_bclink_reset_stats())
3064                         return tipc_cfg_reply_error_string("link not found");
3065                 return tipc_cfg_reply_none();
3066         }
3067
3068         read_lock_bh(&tipc_net_lock);
3069         l_ptr = link_find_link(link_name, &node);
3070         if (!l_ptr) {
3071                 read_unlock_bh(&tipc_net_lock);
3072                 return tipc_cfg_reply_error_string("link not found");
3073         }
3074
3075         tipc_node_lock(node);
3076         link_reset_statistics(l_ptr);
3077         tipc_node_unlock(node);
3078         read_unlock_bh(&tipc_net_lock);
3079         return tipc_cfg_reply_none();
3080 }
3081
3082 /**
3083  * percent - convert count to a percentage of total (rounding up or down)
3084  */
3085
3086 static u32 percent(u32 count, u32 total)
3087 {
3088         return (count * 100 + (total / 2)) / total;
3089 }
3090
3091 /**
3092  * tipc_link_stats - print link statistics
3093  * @name: link name
3094  * @buf: print buffer area
3095  * @buf_size: size of print buffer area
3096  *
3097  * Returns length of print buffer data string (or 0 if error)
3098  */
3099
3100 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
3101 {
3102         struct print_buf pb;
3103         struct link *l_ptr;
3104         struct tipc_node *node;
3105         char *status;
3106         u32 profile_total = 0;
3107
3108         if (!strcmp(name, tipc_bclink_name))
3109                 return tipc_bclink_stats(buf, buf_size);
3110
3111         tipc_printbuf_init(&pb, buf, buf_size);
3112
3113         read_lock_bh(&tipc_net_lock);
3114         l_ptr = link_find_link(name, &node);
3115         if (!l_ptr) {
3116                 read_unlock_bh(&tipc_net_lock);
3117                 return 0;
3118         }
3119         tipc_node_lock(node);
3120
3121         if (tipc_link_is_active(l_ptr))
3122                 status = "ACTIVE";
3123         else if (tipc_link_is_up(l_ptr))
3124                 status = "STANDBY";
3125         else
3126                 status = "DEFUNCT";
3127         tipc_printf(&pb, "Link <%s>\n"
3128                          "  %s  MTU:%u  Priority:%u  Tolerance:%u ms"
3129                          "  Window:%u packets\n",
3130                     l_ptr->name, status, link_max_pkt(l_ptr),
3131                     l_ptr->priority, l_ptr->tolerance, l_ptr->queue_limit[0]);
3132         tipc_printf(&pb, "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
3133                     l_ptr->next_in_no - l_ptr->stats.recv_info,
3134                     l_ptr->stats.recv_fragments,
3135                     l_ptr->stats.recv_fragmented,
3136                     l_ptr->stats.recv_bundles,
3137                     l_ptr->stats.recv_bundled);
3138         tipc_printf(&pb, "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
3139                     l_ptr->next_out_no - l_ptr->stats.sent_info,
3140                     l_ptr->stats.sent_fragments,
3141                     l_ptr->stats.sent_fragmented,
3142                     l_ptr->stats.sent_bundles,
3143                     l_ptr->stats.sent_bundled);
3144         profile_total = l_ptr->stats.msg_length_counts;
3145         if (!profile_total)
3146                 profile_total = 1;
3147         tipc_printf(&pb, "  TX profile sample:%u packets  average:%u octets\n"
3148                          "  0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
3149                          "-16354:%u%% -32768:%u%% -66000:%u%%\n",
3150                     l_ptr->stats.msg_length_counts,
3151                     l_ptr->stats.msg_lengths_total / profile_total,
3152                     percent(l_ptr->stats.msg_length_profile[0], profile_total),
3153                     percent(l_ptr->stats.msg_length_profile[1], profile_total),
3154                     percent(l_ptr->stats.msg_length_profile[2], profile_total),
3155                     percent(l_ptr->stats.msg_length_profile[3], profile_total),
3156                     percent(l_ptr->stats.msg_length_profile[4], profile_total),
3157                     percent(l_ptr->stats.msg_length_profile[5], profile_total),
3158                     percent(l_ptr->stats.msg_length_profile[6], profile_total));
3159         tipc_printf(&pb, "  RX states:%u probes:%u naks:%u defs:%u dups:%u\n",
3160                     l_ptr->stats.recv_states,
3161                     l_ptr->stats.recv_probes,
3162                     l_ptr->stats.recv_nacks,
3163                     l_ptr->stats.deferred_recv,
3164                     l_ptr->stats.duplicates);
3165         tipc_printf(&pb, "  TX states:%u probes:%u naks:%u acks:%u dups:%u\n",
3166                     l_ptr->stats.sent_states,
3167                     l_ptr->stats.sent_probes,
3168                     l_ptr->stats.sent_nacks,
3169                     l_ptr->stats.sent_acks,
3170                     l_ptr->stats.retransmitted);
3171         tipc_printf(&pb, "  Congestion bearer:%u link:%u  Send queue max:%u avg:%u\n",
3172                     l_ptr->stats.bearer_congs,
3173                     l_ptr->stats.link_congs,
3174                     l_ptr->stats.max_queue_sz,
3175                     l_ptr->stats.queue_sz_counts
3176                     ? (l_ptr->stats.accu_queue_sz / l_ptr->stats.queue_sz_counts)
3177                     : 0);
3178
3179         tipc_node_unlock(node);
3180         read_unlock_bh(&tipc_net_lock);
3181         return tipc_printbuf_validate(&pb);
3182 }
3183
3184 #define MAX_LINK_STATS_INFO 2000
3185
3186 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
3187 {
3188         struct sk_buff *buf;
3189         struct tlv_desc *rep_tlv;
3190         int str_len;
3191
3192         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3193                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3194
3195         buf = tipc_cfg_reply_alloc(TLV_SPACE(MAX_LINK_STATS_INFO));
3196         if (!buf)
3197                 return NULL;
3198
3199         rep_tlv = (struct tlv_desc *)buf->data;
3200
3201         str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
3202                                   (char *)TLV_DATA(rep_tlv), MAX_LINK_STATS_INFO);
3203         if (!str_len) {
3204                 buf_discard(buf);
3205                 return tipc_cfg_reply_error_string("link not found");
3206         }
3207
3208         skb_put(buf, TLV_SPACE(str_len));
3209         TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
3210
3211         return buf;
3212 }
3213
3214 #if 0
3215 int link_control(const char *name, u32 op, u32 val)
3216 {
3217         int res = -EINVAL;
3218         struct link *l_ptr;
3219         u32 bearer_id;
3220         struct tipc_node * node;
3221         u32 a;
3222
3223         a = link_name2addr(name, &bearer_id);
3224         read_lock_bh(&tipc_net_lock);
3225         node = tipc_node_find(a);
3226         if (node) {
3227                 tipc_node_lock(node);
3228                 l_ptr = node->links[bearer_id];
3229                 if (l_ptr) {
3230                         if (op == TIPC_REMOVE_LINK) {
3231                                 struct bearer *b_ptr = l_ptr->b_ptr;
3232                                 spin_lock_bh(&b_ptr->publ.lock);
3233                                 tipc_link_delete(l_ptr);
3234                                 spin_unlock_bh(&b_ptr->publ.lock);
3235                         }
3236                         if (op == TIPC_CMD_BLOCK_LINK) {
3237                                 tipc_link_reset(l_ptr);
3238                                 l_ptr->blocked = 1;
3239                         }
3240                         if (op == TIPC_CMD_UNBLOCK_LINK) {
3241                                 l_ptr->blocked = 0;
3242                         }
3243                         res = 0;
3244                 }
3245                 tipc_node_unlock(node);
3246         }
3247         read_unlock_bh(&tipc_net_lock);
3248         return res;
3249 }
3250 #endif
3251
3252 /**
3253  * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
3254  * @dest: network address of destination node
3255  * @selector: used to select from set of active links
3256  *
3257  * If no active link can be found, uses default maximum packet size.
3258  */
3259
3260 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
3261 {
3262         struct tipc_node *n_ptr;
3263         struct link *l_ptr;
3264         u32 res = MAX_PKT_DEFAULT;
3265
3266         if (dest == tipc_own_addr)
3267                 return MAX_MSG_SIZE;
3268
3269         read_lock_bh(&tipc_net_lock);
3270         n_ptr = tipc_node_select(dest, selector);
3271         if (n_ptr) {
3272                 tipc_node_lock(n_ptr);
3273                 l_ptr = n_ptr->active_links[selector & 1];
3274                 if (l_ptr)
3275                         res = link_max_pkt(l_ptr);
3276                 tipc_node_unlock(n_ptr);
3277         }
3278         read_unlock_bh(&tipc_net_lock);
3279         return res;
3280 }
3281
3282 #if 0
3283 static void link_dump_rec_queue(struct link *l_ptr)
3284 {
3285         struct sk_buff *crs;
3286
3287         if (!l_ptr->oldest_deferred_in) {
3288                 info("Reception queue empty\n");
3289                 return;
3290         }
3291         info("Contents of Reception queue:\n");
3292         crs = l_ptr->oldest_deferred_in;
3293         while (crs) {
3294                 if (crs->data == (void *)0x0000a3a3) {
3295                         info("buffer %x invalid\n", crs);
3296                         return;
3297                 }
3298                 msg_dbg(buf_msg(crs), "In rec queue:\n");
3299                 crs = crs->next;
3300         }
3301 }
3302 #endif
3303
3304 static void link_dump_send_queue(struct link *l_ptr)
3305 {
3306         if (l_ptr->next_out) {
3307                 info("\nContents of unsent queue:\n");
3308                 dbg_print_buf_chain(l_ptr->next_out);
3309         }
3310         info("\nContents of send queue:\n");
3311         if (l_ptr->first_out) {
3312                 dbg_print_buf_chain(l_ptr->first_out);
3313         }
3314         info("Empty send queue\n");
3315 }
3316
3317 static void link_print(struct link *l_ptr, struct print_buf *buf,
3318                        const char *str)
3319 {
3320         tipc_printf(buf, str);
3321         if (link_reset_reset(l_ptr) || link_reset_unknown(l_ptr))
3322                 return;
3323         tipc_printf(buf, "Link %x<%s>:",
3324                     l_ptr->addr, l_ptr->b_ptr->publ.name);
3325         tipc_printf(buf, ": NXO(%u):", mod(l_ptr->next_out_no));
3326         tipc_printf(buf, "NXI(%u):", mod(l_ptr->next_in_no));
3327         tipc_printf(buf, "SQUE");
3328         if (l_ptr->first_out) {
3329                 tipc_printf(buf, "[%u..", msg_seqno(buf_msg(l_ptr->first_out)));
3330                 if (l_ptr->next_out)
3331                         tipc_printf(buf, "%u..",
3332                                     msg_seqno(buf_msg(l_ptr->next_out)));
3333                 tipc_printf(buf, "%u]",
3334                             msg_seqno(buf_msg
3335                                       (l_ptr->last_out)), l_ptr->out_queue_size);
3336                 if ((mod(msg_seqno(buf_msg(l_ptr->last_out)) -
3337                          msg_seqno(buf_msg(l_ptr->first_out)))
3338                      != (l_ptr->out_queue_size - 1)) ||
3339                     (l_ptr->last_out->next != NULL)) {
3340                         tipc_printf(buf, "\nSend queue inconsistency\n");
3341                         tipc_printf(buf, "first_out= %x ", l_ptr->first_out);
3342                         tipc_printf(buf, "next_out= %x ", l_ptr->next_out);
3343                         tipc_printf(buf, "last_out= %x ", l_ptr->last_out);
3344                         link_dump_send_queue(l_ptr);
3345                 }
3346         } else
3347                 tipc_printf(buf, "[]");
3348         tipc_printf(buf, "SQSIZ(%u)", l_ptr->out_queue_size);
3349         if (l_ptr->oldest_deferred_in) {
3350                 u32 o = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
3351                 u32 n = msg_seqno(buf_msg(l_ptr->newest_deferred_in));
3352                 tipc_printf(buf, ":RQUE[%u..%u]", o, n);
3353                 if (l_ptr->deferred_inqueue_sz != mod((n + 1) - o)) {
3354                         tipc_printf(buf, ":RQSIZ(%u)",
3355                                     l_ptr->deferred_inqueue_sz);
3356                 }
3357         }
3358         if (link_working_unknown(l_ptr))
3359                 tipc_printf(buf, ":WU");
3360         if (link_reset_reset(l_ptr))
3361                 tipc_printf(buf, ":RR");
3362         if (link_reset_unknown(l_ptr))
3363                 tipc_printf(buf, ":RU");
3364         if (link_working_working(l_ptr))
3365                 tipc_printf(buf, ":WW");
3366         tipc_printf(buf, "\n");
3367 }
3368