]> err.no Git - linux-2.6/blob - net/tipc/link.c
tipc: Fix minor bugs in link session number handling
[linux-2.6] / 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, TIPC_OK, 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                                          TIPC_OK, 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 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 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 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 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                  TIPC_OK, 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 && !skb_cloned(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 TIPC_OK;
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 TIPC_OK;
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 TIPC_OK;
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         }
1632         while (res == TIPC_OK);
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 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 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 (!skb_cloned(buf)) {
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                         return;
1731                 } else {
1732                         /* Don't retransmit if driver already has the buffer */
1733                 }
1734         } else {
1735                 /* Detect repeated retransmit failures on uncongested bearer */
1736
1737                 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1738                         if (++l_ptr->stale_count > 100) {
1739                                 link_retransmit_failure(l_ptr, buf);
1740                                 return;
1741                         }
1742                 } else {
1743                         l_ptr->last_retransmitted = msg_seqno(msg);
1744                         l_ptr->stale_count = 1;
1745                 }
1746         }
1747
1748         while (retransmits && (buf != l_ptr->next_out) && buf && !skb_cloned(buf)) {
1749                 msg = buf_msg(buf);
1750                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1751                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1752                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1753                         msg_dbg(buf_msg(buf), ">RETR>");
1754                         buf = buf->next;
1755                         retransmits--;
1756                         l_ptr->stats.retransmitted++;
1757                 } else {
1758                         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1759                         l_ptr->stats.bearer_congs++;
1760                         l_ptr->retransm_queue_head = msg_seqno(buf_msg(buf));
1761                         l_ptr->retransm_queue_size = retransmits;
1762                         return;
1763                 }
1764         }
1765
1766         l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1767 }
1768
1769 /*
1770  * link_recv_non_seq: Receive packets which are outside
1771  *                    the link sequence flow
1772  */
1773
1774 static void link_recv_non_seq(struct sk_buff *buf)
1775 {
1776         struct tipc_msg *msg = buf_msg(buf);
1777
1778         if (msg_user(msg) ==  LINK_CONFIG)
1779                 tipc_disc_recv_msg(buf);
1780         else
1781                 tipc_bclink_recv_pkt(buf);
1782 }
1783
1784 /**
1785  * link_insert_deferred_queue - insert deferred messages back into receive chain
1786  */
1787
1788 static struct sk_buff *link_insert_deferred_queue(struct link *l_ptr,
1789                                                   struct sk_buff *buf)
1790 {
1791         u32 seq_no;
1792
1793         if (l_ptr->oldest_deferred_in == NULL)
1794                 return buf;
1795
1796         seq_no = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1797         if (seq_no == mod(l_ptr->next_in_no)) {
1798                 l_ptr->newest_deferred_in->next = buf;
1799                 buf = l_ptr->oldest_deferred_in;
1800                 l_ptr->oldest_deferred_in = NULL;
1801                 l_ptr->deferred_inqueue_sz = 0;
1802         }
1803         return buf;
1804 }
1805
1806 /**
1807  * link_recv_buf_validate - validate basic format of received message
1808  *
1809  * This routine ensures a TIPC message has an acceptable header, and at least
1810  * as much data as the header indicates it should.  The routine also ensures
1811  * that the entire message header is stored in the main fragment of the message
1812  * buffer, to simplify future access to message header fields.
1813  *
1814  * Note: Having extra info present in the message header or data areas is OK.
1815  * TIPC will ignore the excess, under the assumption that it is optional info
1816  * introduced by a later release of the protocol.
1817  */
1818
1819 static int link_recv_buf_validate(struct sk_buff *buf)
1820 {
1821         static u32 min_data_hdr_size[8] = {
1822                 SHORT_H_SIZE, MCAST_H_SIZE, LONG_H_SIZE, DIR_MSG_H_SIZE,
1823                 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1824                 };
1825
1826         struct tipc_msg *msg;
1827         u32 tipc_hdr[2];
1828         u32 size;
1829         u32 hdr_size;
1830         u32 min_hdr_size;
1831
1832         if (unlikely(buf->len < MIN_H_SIZE))
1833                 return 0;
1834
1835         msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1836         if (msg == NULL)
1837                 return 0;
1838
1839         if (unlikely(msg_version(msg) != TIPC_VERSION))
1840                 return 0;
1841
1842         size = msg_size(msg);
1843         hdr_size = msg_hdr_sz(msg);
1844         min_hdr_size = msg_isdata(msg) ?
1845                 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1846
1847         if (unlikely((hdr_size < min_hdr_size) ||
1848                      (size < hdr_size) ||
1849                      (buf->len < size) ||
1850                      (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1851                 return 0;
1852
1853         return pskb_may_pull(buf, hdr_size);
1854 }
1855
1856 void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *tb_ptr)
1857 {
1858         read_lock_bh(&tipc_net_lock);
1859         while (head) {
1860                 struct bearer *b_ptr;
1861                 struct node *n_ptr;
1862                 struct link *l_ptr;
1863                 struct sk_buff *crs;
1864                 struct sk_buff *buf = head;
1865                 struct tipc_msg *msg;
1866                 u32 seq_no;
1867                 u32 ackd;
1868                 u32 released = 0;
1869                 int type;
1870
1871                 b_ptr = (struct bearer *)tb_ptr;
1872                 TIPC_SKB_CB(buf)->handle = b_ptr;
1873
1874                 head = head->next;
1875
1876                 /* Ensure message is well-formed */
1877
1878                 if (unlikely(!link_recv_buf_validate(buf)))
1879                         goto cont;
1880
1881                 /* Ensure message data is a single contiguous unit */
1882
1883                 if (unlikely(buf_linearize(buf))) {
1884                         goto cont;
1885                 }
1886
1887                 /* Handle arrival of a non-unicast link message */
1888
1889                 msg = buf_msg(buf);
1890
1891                 if (unlikely(msg_non_seq(msg))) {
1892                         link_recv_non_seq(buf);
1893                         continue;
1894                 }
1895
1896                 if (unlikely(!msg_short(msg) &&
1897                              (msg_destnode(msg) != tipc_own_addr)))
1898                         goto cont;
1899
1900                 /* Locate unicast link endpoint that should handle message */
1901
1902                 n_ptr = tipc_node_find(msg_prevnode(msg));
1903                 if (unlikely(!n_ptr))
1904                         goto cont;
1905                 tipc_node_lock(n_ptr);
1906
1907                 l_ptr = n_ptr->links[b_ptr->identity];
1908                 if (unlikely(!l_ptr)) {
1909                         tipc_node_unlock(n_ptr);
1910                         goto cont;
1911                 }
1912
1913                 /* Validate message sequence number info */
1914
1915                 seq_no = msg_seqno(msg);
1916                 ackd = msg_ack(msg);
1917
1918                 /* Release acked messages */
1919
1920                 if (less(n_ptr->bclink.acked, msg_bcast_ack(msg))) {
1921                         if (tipc_node_is_up(n_ptr) && n_ptr->bclink.supported)
1922                                 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1923                 }
1924
1925                 crs = l_ptr->first_out;
1926                 while ((crs != l_ptr->next_out) &&
1927                        less_eq(msg_seqno(buf_msg(crs)), ackd)) {
1928                         struct sk_buff *next = crs->next;
1929
1930                         buf_discard(crs);
1931                         crs = next;
1932                         released++;
1933                 }
1934                 if (released) {
1935                         l_ptr->first_out = crs;
1936                         l_ptr->out_queue_size -= released;
1937                 }
1938
1939                 /* Try sending any messages link endpoint has pending */
1940
1941                 if (unlikely(l_ptr->next_out))
1942                         tipc_link_push_queue(l_ptr);
1943                 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1944                         tipc_link_wakeup_ports(l_ptr, 0);
1945                 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1946                         l_ptr->stats.sent_acks++;
1947                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1948                 }
1949
1950                 /* Now (finally!) process the incoming message */
1951
1952 protocol_check:
1953                 if (likely(link_working_working(l_ptr))) {
1954                         if (likely(seq_no == mod(l_ptr->next_in_no))) {
1955                                 l_ptr->next_in_no++;
1956                                 if (unlikely(l_ptr->oldest_deferred_in))
1957                                         head = link_insert_deferred_queue(l_ptr,
1958                                                                           head);
1959                                 if (likely(msg_is_dest(msg, tipc_own_addr))) {
1960 deliver:
1961                                         if (likely(msg_isdata(msg))) {
1962                                                 tipc_node_unlock(n_ptr);
1963                                                 tipc_port_recv_msg(buf);
1964                                                 continue;
1965                                         }
1966                                         switch (msg_user(msg)) {
1967                                         case MSG_BUNDLER:
1968                                                 l_ptr->stats.recv_bundles++;
1969                                                 l_ptr->stats.recv_bundled +=
1970                                                         msg_msgcnt(msg);
1971                                                 tipc_node_unlock(n_ptr);
1972                                                 tipc_link_recv_bundle(buf);
1973                                                 continue;
1974                                         case ROUTE_DISTRIBUTOR:
1975                                                 tipc_node_unlock(n_ptr);
1976                                                 tipc_cltr_recv_routing_table(buf);
1977                                                 continue;
1978                                         case NAME_DISTRIBUTOR:
1979                                                 tipc_node_unlock(n_ptr);
1980                                                 tipc_named_recv(buf);
1981                                                 continue;
1982                                         case CONN_MANAGER:
1983                                                 tipc_node_unlock(n_ptr);
1984                                                 tipc_port_recv_proto_msg(buf);
1985                                                 continue;
1986                                         case MSG_FRAGMENTER:
1987                                                 l_ptr->stats.recv_fragments++;
1988                                                 if (tipc_link_recv_fragment(&l_ptr->defragm_buf,
1989                                                                             &buf, &msg)) {
1990                                                         l_ptr->stats.recv_fragmented++;
1991                                                         goto deliver;
1992                                                 }
1993                                                 break;
1994                                         case CHANGEOVER_PROTOCOL:
1995                                                 type = msg_type(msg);
1996                                                 if (link_recv_changeover_msg(&l_ptr, &buf)) {
1997                                                         msg = buf_msg(buf);
1998                                                         seq_no = msg_seqno(msg);
1999                                                         TIPC_SKB_CB(buf)->handle
2000                                                                 = b_ptr;
2001                                                         if (type == ORIGINAL_MSG)
2002                                                                 goto deliver;
2003                                                         goto protocol_check;
2004                                                 }
2005                                                 break;
2006                                         }
2007                                 }
2008                                 tipc_node_unlock(n_ptr);
2009                                 tipc_net_route_msg(buf);
2010                                 continue;
2011                         }
2012                         link_handle_out_of_seq_msg(l_ptr, buf);
2013                         head = link_insert_deferred_queue(l_ptr, head);
2014                         tipc_node_unlock(n_ptr);
2015                         continue;
2016                 }
2017
2018                 if (msg_user(msg) == LINK_PROTOCOL) {
2019                         link_recv_proto_msg(l_ptr, buf);
2020                         head = link_insert_deferred_queue(l_ptr, head);
2021                         tipc_node_unlock(n_ptr);
2022                         continue;
2023                 }
2024                 msg_dbg(msg,"NSEQ<REC<");
2025                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2026
2027                 if (link_working_working(l_ptr)) {
2028                         /* Re-insert in front of queue */
2029                         msg_dbg(msg,"RECV-REINS:");
2030                         buf->next = head;
2031                         head = buf;
2032                         tipc_node_unlock(n_ptr);
2033                         continue;
2034                 }
2035                 tipc_node_unlock(n_ptr);
2036 cont:
2037                 buf_discard(buf);
2038         }
2039         read_unlock_bh(&tipc_net_lock);
2040 }
2041
2042 /*
2043  * link_defer_buf(): Sort a received out-of-sequence packet
2044  *                   into the deferred reception queue.
2045  * Returns the increase of the queue length,i.e. 0 or 1
2046  */
2047
2048 u32 tipc_link_defer_pkt(struct sk_buff **head,
2049                         struct sk_buff **tail,
2050                         struct sk_buff *buf)
2051 {
2052         struct sk_buff *prev = NULL;
2053         struct sk_buff *crs = *head;
2054         u32 seq_no = msg_seqno(buf_msg(buf));
2055
2056         buf->next = NULL;
2057
2058         /* Empty queue ? */
2059         if (*head == NULL) {
2060                 *head = *tail = buf;
2061                 return 1;
2062         }
2063
2064         /* Last ? */
2065         if (less(msg_seqno(buf_msg(*tail)), seq_no)) {
2066                 (*tail)->next = buf;
2067                 *tail = buf;
2068                 return 1;
2069         }
2070
2071         /* Scan through queue and sort it in */
2072         do {
2073                 struct tipc_msg *msg = buf_msg(crs);
2074
2075                 if (less(seq_no, msg_seqno(msg))) {
2076                         buf->next = crs;
2077                         if (prev)
2078                                 prev->next = buf;
2079                         else
2080                                 *head = buf;
2081                         return 1;
2082                 }
2083                 if (seq_no == msg_seqno(msg)) {
2084                         break;
2085                 }
2086                 prev = crs;
2087                 crs = crs->next;
2088         }
2089         while (crs);
2090
2091         /* Message is a duplicate of an existing message */
2092
2093         buf_discard(buf);
2094         return 0;
2095 }
2096
2097 /**
2098  * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
2099  */
2100
2101 static void link_handle_out_of_seq_msg(struct link *l_ptr,
2102                                        struct sk_buff *buf)
2103 {
2104         u32 seq_no = msg_seqno(buf_msg(buf));
2105
2106         if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
2107                 link_recv_proto_msg(l_ptr, buf);
2108                 return;
2109         }
2110
2111         dbg("rx OOS msg: seq_no %u, expecting %u (%u)\n",
2112             seq_no, mod(l_ptr->next_in_no), l_ptr->next_in_no);
2113
2114         /* Record OOS packet arrival (force mismatch on next timeout) */
2115
2116         l_ptr->checkpoint--;
2117
2118         /*
2119          * Discard packet if a duplicate; otherwise add it to deferred queue
2120          * and notify peer of gap as per protocol specification
2121          */
2122
2123         if (less(seq_no, mod(l_ptr->next_in_no))) {
2124                 l_ptr->stats.duplicates++;
2125                 buf_discard(buf);
2126                 return;
2127         }
2128
2129         if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
2130                                 &l_ptr->newest_deferred_in, buf)) {
2131                 l_ptr->deferred_inqueue_sz++;
2132                 l_ptr->stats.deferred_recv++;
2133                 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
2134                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
2135         } else
2136                 l_ptr->stats.duplicates++;
2137 }
2138
2139 /*
2140  * Send protocol message to the other endpoint.
2141  */
2142 void tipc_link_send_proto_msg(struct link *l_ptr, u32 msg_typ, int probe_msg,
2143                               u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
2144 {
2145         struct sk_buff *buf = NULL;
2146         struct tipc_msg *msg = l_ptr->pmsg;
2147         u32 msg_size = sizeof(l_ptr->proto_msg);
2148
2149         if (link_blocked(l_ptr))
2150                 return;
2151         msg_set_type(msg, msg_typ);
2152         msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
2153         msg_set_bcast_ack(msg, mod(l_ptr->owner->bclink.last_in));
2154         msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
2155
2156         if (msg_typ == STATE_MSG) {
2157                 u32 next_sent = mod(l_ptr->next_out_no);
2158
2159                 if (!tipc_link_is_up(l_ptr))
2160                         return;
2161                 if (l_ptr->next_out)
2162                         next_sent = msg_seqno(buf_msg(l_ptr->next_out));
2163                 msg_set_next_sent(msg, next_sent);
2164                 if (l_ptr->oldest_deferred_in) {
2165                         u32 rec = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
2166                         gap = mod(rec - mod(l_ptr->next_in_no));
2167                 }
2168                 msg_set_seq_gap(msg, gap);
2169                 if (gap)
2170                         l_ptr->stats.sent_nacks++;
2171                 msg_set_link_tolerance(msg, tolerance);
2172                 msg_set_linkprio(msg, priority);
2173                 msg_set_max_pkt(msg, ack_mtu);
2174                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
2175                 msg_set_probe(msg, probe_msg != 0);
2176                 if (probe_msg) {
2177                         u32 mtu = l_ptr->max_pkt;
2178
2179                         if ((mtu < l_ptr->max_pkt_target) &&
2180                             link_working_working(l_ptr) &&
2181                             l_ptr->fsm_msg_cnt) {
2182                                 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2183                                 if (l_ptr->max_pkt_probes == 10) {
2184                                         l_ptr->max_pkt_target = (msg_size - 4);
2185                                         l_ptr->max_pkt_probes = 0;
2186                                         msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2187                                 }
2188                                 l_ptr->max_pkt_probes++;
2189                         }
2190
2191                         l_ptr->stats.sent_probes++;
2192                 }
2193                 l_ptr->stats.sent_states++;
2194         } else {                /* RESET_MSG or ACTIVATE_MSG */
2195                 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
2196                 msg_set_seq_gap(msg, 0);
2197                 msg_set_next_sent(msg, 1);
2198                 msg_set_link_tolerance(msg, l_ptr->tolerance);
2199                 msg_set_linkprio(msg, l_ptr->priority);
2200                 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
2201         }
2202
2203         if (tipc_node_has_redundant_links(l_ptr->owner)) {
2204                 msg_set_redundant_link(msg);
2205         } else {
2206                 msg_clear_redundant_link(msg);
2207         }
2208         msg_set_linkprio(msg, l_ptr->priority);
2209
2210         /* Ensure sequence number will not fit : */
2211
2212         msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
2213
2214         /* Congestion? */
2215
2216         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
2217                 if (!l_ptr->proto_msg_queue) {
2218                         l_ptr->proto_msg_queue =
2219                                 buf_acquire(sizeof(l_ptr->proto_msg));
2220                 }
2221                 buf = l_ptr->proto_msg_queue;
2222                 if (!buf)
2223                         return;
2224                 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2225                 return;
2226         }
2227         msg_set_timestamp(msg, jiffies_to_msecs(jiffies));
2228
2229         /* Message can be sent */
2230
2231         msg_dbg(msg, ">>");
2232
2233         buf = buf_acquire(msg_size);
2234         if (!buf)
2235                 return;
2236
2237         skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2238         msg_set_size(buf_msg(buf), msg_size);
2239
2240         if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
2241                 l_ptr->unacked_window = 0;
2242                 buf_discard(buf);
2243                 return;
2244         }
2245
2246         /* New congestion */
2247         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
2248         l_ptr->proto_msg_queue = buf;
2249         l_ptr->stats.bearer_congs++;
2250 }
2251
2252 /*
2253  * Receive protocol message :
2254  * Note that network plane id propagates through the network, and may
2255  * change at any time. The node with lowest address rules
2256  */
2257
2258 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf)
2259 {
2260         u32 rec_gap = 0;
2261         u32 max_pkt_info;
2262         u32 max_pkt_ack;
2263         u32 msg_tol;
2264         struct tipc_msg *msg = buf_msg(buf);
2265
2266         dbg("AT(%u):", jiffies_to_msecs(jiffies));
2267         msg_dbg(msg, "<<");
2268         if (link_blocked(l_ptr))
2269                 goto exit;
2270
2271         /* record unnumbered packet arrival (force mismatch on next timeout) */
2272
2273         l_ptr->checkpoint--;
2274
2275         if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
2276                 if (tipc_own_addr > msg_prevnode(msg))
2277                         l_ptr->b_ptr->net_plane = msg_net_plane(msg);
2278
2279         l_ptr->owner->permit_changeover = msg_redundant_link(msg);
2280
2281         switch (msg_type(msg)) {
2282
2283         case RESET_MSG:
2284                 if (!link_working_unknown(l_ptr) &&
2285                     (l_ptr->peer_session != INVALID_SESSION)) {
2286                         if (msg_session(msg) == l_ptr->peer_session) {
2287                                 dbg("Duplicate RESET: %u<->%u\n",
2288                                     msg_session(msg), l_ptr->peer_session);
2289                                 break; /* duplicate: ignore */
2290                         }
2291                 }
2292                 /* fall thru' */
2293         case ACTIVATE_MSG:
2294                 /* Update link settings according other endpoint's values */
2295
2296                 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2297
2298                 if ((msg_tol = msg_link_tolerance(msg)) &&
2299                     (msg_tol > l_ptr->tolerance))
2300                         link_set_supervision_props(l_ptr, msg_tol);
2301
2302                 if (msg_linkprio(msg) > l_ptr->priority)
2303                         l_ptr->priority = msg_linkprio(msg);
2304
2305                 max_pkt_info = msg_max_pkt(msg);
2306                 if (max_pkt_info) {
2307                         if (max_pkt_info < l_ptr->max_pkt_target)
2308                                 l_ptr->max_pkt_target = max_pkt_info;
2309                         if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2310                                 l_ptr->max_pkt = l_ptr->max_pkt_target;
2311                 } else {
2312                         l_ptr->max_pkt = l_ptr->max_pkt_target;
2313                 }
2314                 l_ptr->owner->bclink.supported = (max_pkt_info != 0);
2315
2316                 link_state_event(l_ptr, msg_type(msg));
2317
2318                 l_ptr->peer_session = msg_session(msg);
2319                 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2320
2321                 /* Synchronize broadcast sequence numbers */
2322                 if (!tipc_node_has_redundant_links(l_ptr->owner)) {
2323                         l_ptr->owner->bclink.last_in = mod(msg_last_bcast(msg));
2324                 }
2325                 break;
2326         case STATE_MSG:
2327
2328                 if ((msg_tol = msg_link_tolerance(msg)))
2329                         link_set_supervision_props(l_ptr, msg_tol);
2330
2331                 if (msg_linkprio(msg) &&
2332                     (msg_linkprio(msg) != l_ptr->priority)) {
2333                         warn("Resetting link <%s>, priority change %u->%u\n",
2334                              l_ptr->name, l_ptr->priority, msg_linkprio(msg));
2335                         l_ptr->priority = msg_linkprio(msg);
2336                         tipc_link_reset(l_ptr); /* Enforce change to take effect */
2337                         break;
2338                 }
2339                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2340                 l_ptr->stats.recv_states++;
2341                 if (link_reset_unknown(l_ptr))
2342                         break;
2343
2344                 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2345                         rec_gap = mod(msg_next_sent(msg) -
2346                                       mod(l_ptr->next_in_no));
2347                 }
2348
2349                 max_pkt_ack = msg_max_pkt(msg);
2350                 if (max_pkt_ack > l_ptr->max_pkt) {
2351                         dbg("Link <%s> updated MTU %u -> %u\n",
2352                             l_ptr->name, l_ptr->max_pkt, max_pkt_ack);
2353                         l_ptr->max_pkt = max_pkt_ack;
2354                         l_ptr->max_pkt_probes = 0;
2355                 }
2356
2357                 max_pkt_ack = 0;
2358                 if (msg_probe(msg)) {
2359                         l_ptr->stats.recv_probes++;
2360                         if (msg_size(msg) > sizeof(l_ptr->proto_msg)) {
2361                                 max_pkt_ack = msg_size(msg);
2362                         }
2363                 }
2364
2365                 /* Protocol message before retransmits, reduce loss risk */
2366
2367                 tipc_bclink_check_gap(l_ptr->owner, msg_last_bcast(msg));
2368
2369                 if (rec_gap || (msg_probe(msg))) {
2370                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2371                                                  0, rec_gap, 0, 0, max_pkt_ack);
2372                 }
2373                 if (msg_seq_gap(msg)) {
2374                         msg_dbg(msg, "With Gap:");
2375                         l_ptr->stats.recv_nacks++;
2376                         tipc_link_retransmit(l_ptr, l_ptr->first_out,
2377                                              msg_seq_gap(msg));
2378                 }
2379                 break;
2380         default:
2381                 msg_dbg(buf_msg(buf), "<DISCARDING UNKNOWN<");
2382         }
2383 exit:
2384         buf_discard(buf);
2385 }
2386
2387
2388 /*
2389  * tipc_link_tunnel(): Send one message via a link belonging to
2390  * another bearer. Owner node is locked.
2391  */
2392 void tipc_link_tunnel(struct link *l_ptr,
2393                       struct tipc_msg *tunnel_hdr,
2394                       struct tipc_msg  *msg,
2395                       u32 selector)
2396 {
2397         struct link *tunnel;
2398         struct sk_buff *buf;
2399         u32 length = msg_size(msg);
2400
2401         tunnel = l_ptr->owner->active_links[selector & 1];
2402         if (!tipc_link_is_up(tunnel)) {
2403                 warn("Link changeover error, "
2404                      "tunnel link no longer available\n");
2405                 return;
2406         }
2407         msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2408         buf = buf_acquire(length + INT_H_SIZE);
2409         if (!buf) {
2410                 warn("Link changeover error, "
2411                      "unable to send tunnel msg\n");
2412                 return;
2413         }
2414         skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
2415         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
2416         dbg("%c->%c:", l_ptr->b_ptr->net_plane, tunnel->b_ptr->net_plane);
2417         msg_dbg(buf_msg(buf), ">SEND>");
2418         tipc_link_send_buf(tunnel, buf);
2419 }
2420
2421
2422
2423 /*
2424  * changeover(): Send whole message queue via the remaining link
2425  *               Owner node is locked.
2426  */
2427
2428 void tipc_link_changeover(struct link *l_ptr)
2429 {
2430         u32 msgcount = l_ptr->out_queue_size;
2431         struct sk_buff *crs = l_ptr->first_out;
2432         struct link *tunnel = l_ptr->owner->active_links[0];
2433         struct tipc_msg tunnel_hdr;
2434         int split_bundles;
2435
2436         if (!tunnel)
2437                 return;
2438
2439         if (!l_ptr->owner->permit_changeover) {
2440                 warn("Link changeover error, "
2441                      "peer did not permit changeover\n");
2442                 return;
2443         }
2444
2445         msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2446                  ORIGINAL_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
2447         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2448         msg_set_msgcnt(&tunnel_hdr, msgcount);
2449         dbg("Link changeover requires %u tunnel messages\n", msgcount);
2450
2451         if (!l_ptr->first_out) {
2452                 struct sk_buff *buf;
2453
2454                 buf = buf_acquire(INT_H_SIZE);
2455                 if (buf) {
2456                         skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2457                         msg_set_size(&tunnel_hdr, INT_H_SIZE);
2458                         dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2459                             tunnel->b_ptr->net_plane);
2460                         msg_dbg(&tunnel_hdr, "EMPTY>SEND>");
2461                         tipc_link_send_buf(tunnel, buf);
2462                 } else {
2463                         warn("Link changeover error, "
2464                              "unable to send changeover msg\n");
2465                 }
2466                 return;
2467         }
2468
2469         split_bundles = (l_ptr->owner->active_links[0] !=
2470                          l_ptr->owner->active_links[1]);
2471
2472         while (crs) {
2473                 struct tipc_msg *msg = buf_msg(crs);
2474
2475                 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2476                         struct tipc_msg *m = msg_get_wrapped(msg);
2477                         unchar* pos = (unchar*)m;
2478
2479                         msgcount = msg_msgcnt(msg);
2480                         while (msgcount--) {
2481                                 msg_set_seqno(m,msg_seqno(msg));
2482                                 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2483                                                  msg_link_selector(m));
2484                                 pos += align(msg_size(m));
2485                                 m = (struct tipc_msg *)pos;
2486                         }
2487                 } else {
2488                         tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2489                                          msg_link_selector(msg));
2490                 }
2491                 crs = crs->next;
2492         }
2493 }
2494
2495 void tipc_link_send_duplicate(struct link *l_ptr, struct link *tunnel)
2496 {
2497         struct sk_buff *iter;
2498         struct tipc_msg tunnel_hdr;
2499
2500         msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2501                  DUPLICATE_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
2502         msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2503         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2504         iter = l_ptr->first_out;
2505         while (iter) {
2506                 struct sk_buff *outbuf;
2507                 struct tipc_msg *msg = buf_msg(iter);
2508                 u32 length = msg_size(msg);
2509
2510                 if (msg_user(msg) == MSG_BUNDLER)
2511                         msg_set_type(msg, CLOSED_MSG);
2512                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));   /* Update */
2513                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2514                 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2515                 outbuf = buf_acquire(length + INT_H_SIZE);
2516                 if (outbuf == NULL) {
2517                         warn("Link changeover error, "
2518                              "unable to send duplicate msg\n");
2519                         return;
2520                 }
2521                 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2522                 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2523                                                length);
2524                 dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2525                     tunnel->b_ptr->net_plane);
2526                 msg_dbg(buf_msg(outbuf), ">SEND>");
2527                 tipc_link_send_buf(tunnel, outbuf);
2528                 if (!tipc_link_is_up(l_ptr))
2529                         return;
2530                 iter = iter->next;
2531         }
2532 }
2533
2534
2535
2536 /**
2537  * buf_extract - extracts embedded TIPC message from another message
2538  * @skb: encapsulating message buffer
2539  * @from_pos: offset to extract from
2540  *
2541  * Returns a new message buffer containing an embedded message.  The
2542  * encapsulating message itself is left unchanged.
2543  */
2544
2545 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2546 {
2547         struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2548         u32 size = msg_size(msg);
2549         struct sk_buff *eb;
2550
2551         eb = buf_acquire(size);
2552         if (eb)
2553                 skb_copy_to_linear_data(eb, msg, size);
2554         return eb;
2555 }
2556
2557 /*
2558  *  link_recv_changeover_msg(): Receive tunneled packet sent
2559  *  via other link. Node is locked. Return extracted buffer.
2560  */
2561
2562 static int link_recv_changeover_msg(struct link **l_ptr,
2563                                     struct sk_buff **buf)
2564 {
2565         struct sk_buff *tunnel_buf = *buf;
2566         struct link *dest_link;
2567         struct tipc_msg *msg;
2568         struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2569         u32 msg_typ = msg_type(tunnel_msg);
2570         u32 msg_count = msg_msgcnt(tunnel_msg);
2571
2572         dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2573         if (!dest_link) {
2574                 msg_dbg(tunnel_msg, "NOLINK/<REC<");
2575                 goto exit;
2576         }
2577         if (dest_link == *l_ptr) {
2578                 err("Unexpected changeover message on link <%s>\n",
2579                     (*l_ptr)->name);
2580                 goto exit;
2581         }
2582         dbg("%c<-%c:", dest_link->b_ptr->net_plane,
2583             (*l_ptr)->b_ptr->net_plane);
2584         *l_ptr = dest_link;
2585         msg = msg_get_wrapped(tunnel_msg);
2586
2587         if (msg_typ == DUPLICATE_MSG) {
2588                 if (less(msg_seqno(msg), mod(dest_link->next_in_no))) {
2589                         msg_dbg(tunnel_msg, "DROP/<REC<");
2590                         goto exit;
2591                 }
2592                 *buf = buf_extract(tunnel_buf,INT_H_SIZE);
2593                 if (*buf == NULL) {
2594                         warn("Link changeover error, duplicate msg dropped\n");
2595                         goto exit;
2596                 }
2597                 msg_dbg(tunnel_msg, "TNL<REC<");
2598                 buf_discard(tunnel_buf);
2599                 return 1;
2600         }
2601
2602         /* First original message ?: */
2603
2604         if (tipc_link_is_up(dest_link)) {
2605                 msg_dbg(tunnel_msg, "UP/FIRST/<REC<");
2606                 info("Resetting link <%s>, changeover initiated by peer\n",
2607                      dest_link->name);
2608                 tipc_link_reset(dest_link);
2609                 dest_link->exp_msg_count = msg_count;
2610                 dbg("Expecting %u tunnelled messages\n", msg_count);
2611                 if (!msg_count)
2612                         goto exit;
2613         } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2614                 msg_dbg(tunnel_msg, "BLK/FIRST/<REC<");
2615                 dest_link->exp_msg_count = msg_count;
2616                 dbg("Expecting %u tunnelled messages\n", msg_count);
2617                 if (!msg_count)
2618                         goto exit;
2619         }
2620
2621         /* Receive original message */
2622
2623         if (dest_link->exp_msg_count == 0) {
2624                 warn("Link switchover error, "
2625                      "got too many tunnelled messages\n");
2626                 msg_dbg(tunnel_msg, "OVERDUE/DROP/<REC<");
2627                 dbg_print_link(dest_link, "LINK:");
2628                 goto exit;
2629         }
2630         dest_link->exp_msg_count--;
2631         if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2632                 msg_dbg(tunnel_msg, "DROP/DUPL/<REC<");
2633                 goto exit;
2634         } else {
2635                 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2636                 if (*buf != NULL) {
2637                         msg_dbg(tunnel_msg, "TNL<REC<");
2638                         buf_discard(tunnel_buf);
2639                         return 1;
2640                 } else {
2641                         warn("Link changeover error, original msg dropped\n");
2642                 }
2643         }
2644 exit:
2645         *buf = NULL;
2646         buf_discard(tunnel_buf);
2647         return 0;
2648 }
2649
2650 /*
2651  *  Bundler functionality:
2652  */
2653 void tipc_link_recv_bundle(struct sk_buff *buf)
2654 {
2655         u32 msgcount = msg_msgcnt(buf_msg(buf));
2656         u32 pos = INT_H_SIZE;
2657         struct sk_buff *obuf;
2658
2659         msg_dbg(buf_msg(buf), "<BNDL<: ");
2660         while (msgcount--) {
2661                 obuf = buf_extract(buf, pos);
2662                 if (obuf == NULL) {
2663                         warn("Link unable to unbundle message(s)\n");
2664                         break;
2665                 }
2666                 pos += align(msg_size(buf_msg(obuf)));
2667                 msg_dbg(buf_msg(obuf), "     /");
2668                 tipc_net_route_msg(obuf);
2669         }
2670         buf_discard(buf);
2671 }
2672
2673 /*
2674  *  Fragmentation/defragmentation:
2675  */
2676
2677
2678 /*
2679  * tipc_link_send_long_buf: Entry for buffers needing fragmentation.
2680  * The buffer is complete, inclusive total message length.
2681  * Returns user data length.
2682  */
2683 int tipc_link_send_long_buf(struct link *l_ptr, struct sk_buff *buf)
2684 {
2685         struct tipc_msg *inmsg = buf_msg(buf);
2686         struct tipc_msg fragm_hdr;
2687         u32 insize = msg_size(inmsg);
2688         u32 dsz = msg_data_sz(inmsg);
2689         unchar *crs = buf->data;
2690         u32 rest = insize;
2691         u32 pack_sz = link_max_pkt(l_ptr);
2692         u32 fragm_sz = pack_sz - INT_H_SIZE;
2693         u32 fragm_no = 1;
2694         u32 destaddr = msg_destnode(inmsg);
2695
2696         if (msg_short(inmsg))
2697                 destaddr = l_ptr->addr;
2698
2699         if (msg_routed(inmsg))
2700                 msg_set_prevnode(inmsg, tipc_own_addr);
2701
2702         /* Prepare reusable fragment header: */
2703
2704         msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2705                  TIPC_OK, INT_H_SIZE, destaddr);
2706         msg_set_link_selector(&fragm_hdr, msg_link_selector(inmsg));
2707         msg_set_long_msgno(&fragm_hdr, mod(l_ptr->long_msg_seq_no++));
2708         msg_set_fragm_no(&fragm_hdr, fragm_no);
2709         l_ptr->stats.sent_fragmented++;
2710
2711         /* Chop up message: */
2712
2713         while (rest > 0) {
2714                 struct sk_buff *fragm;
2715
2716                 if (rest <= fragm_sz) {
2717                         fragm_sz = rest;
2718                         msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2719                 }
2720                 fragm = buf_acquire(fragm_sz + INT_H_SIZE);
2721                 if (fragm == NULL) {
2722                         warn("Link unable to fragment message\n");
2723                         dsz = -ENOMEM;
2724                         goto exit;
2725                 }
2726                 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2727                 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2728                 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2729                                                fragm_sz);
2730                 /*  Send queued messages first, if any: */
2731
2732                 l_ptr->stats.sent_fragments++;
2733                 tipc_link_send_buf(l_ptr, fragm);
2734                 if (!tipc_link_is_up(l_ptr))
2735                         return dsz;
2736                 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
2737                 rest -= fragm_sz;
2738                 crs += fragm_sz;
2739                 msg_set_type(&fragm_hdr, FRAGMENT);
2740         }
2741 exit:
2742         buf_discard(buf);
2743         return dsz;
2744 }
2745
2746 /*
2747  * A pending message being re-assembled must store certain values
2748  * to handle subsequent fragments correctly. The following functions
2749  * help storing these values in unused, available fields in the
2750  * pending message. This makes dynamic memory allocation unecessary.
2751  */
2752
2753 static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2754 {
2755         msg_set_seqno(buf_msg(buf), seqno);
2756 }
2757
2758 static u32 get_fragm_size(struct sk_buff *buf)
2759 {
2760         return msg_ack(buf_msg(buf));
2761 }
2762
2763 static void set_fragm_size(struct sk_buff *buf, u32 sz)
2764 {
2765         msg_set_ack(buf_msg(buf), sz);
2766 }
2767
2768 static u32 get_expected_frags(struct sk_buff *buf)
2769 {
2770         return msg_bcast_ack(buf_msg(buf));
2771 }
2772
2773 static void set_expected_frags(struct sk_buff *buf, u32 exp)
2774 {
2775         msg_set_bcast_ack(buf_msg(buf), exp);
2776 }
2777
2778 static u32 get_timer_cnt(struct sk_buff *buf)
2779 {
2780         return msg_reroute_cnt(buf_msg(buf));
2781 }
2782
2783 static void incr_timer_cnt(struct sk_buff *buf)
2784 {
2785         msg_incr_reroute_cnt(buf_msg(buf));
2786 }
2787
2788 /*
2789  * tipc_link_recv_fragment(): Called with node lock on. Returns
2790  * the reassembled buffer if message is complete.
2791  */
2792 int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2793                             struct tipc_msg **m)
2794 {
2795         struct sk_buff *prev = NULL;
2796         struct sk_buff *fbuf = *fb;
2797         struct tipc_msg *fragm = buf_msg(fbuf);
2798         struct sk_buff *pbuf = *pending;
2799         u32 long_msg_seq_no = msg_long_msgno(fragm);
2800
2801         *fb = NULL;
2802         msg_dbg(fragm,"FRG<REC<");
2803
2804         /* Is there an incomplete message waiting for this fragment? */
2805
2806         while (pbuf && ((msg_seqno(buf_msg(pbuf)) != long_msg_seq_no)
2807                         || (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2808                 prev = pbuf;
2809                 pbuf = pbuf->next;
2810         }
2811
2812         if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2813                 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2814                 u32 msg_sz = msg_size(imsg);
2815                 u32 fragm_sz = msg_data_sz(fragm);
2816                 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2817                 u32 max =  TIPC_MAX_USER_MSG_SIZE + LONG_H_SIZE;
2818                 if (msg_type(imsg) == TIPC_MCAST_MSG)
2819                         max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2820                 if (msg_size(imsg) > max) {
2821                         msg_dbg(fragm,"<REC<Oversized: ");
2822                         buf_discard(fbuf);
2823                         return 0;
2824                 }
2825                 pbuf = buf_acquire(msg_size(imsg));
2826                 if (pbuf != NULL) {
2827                         pbuf->next = *pending;
2828                         *pending = pbuf;
2829                         skb_copy_to_linear_data(pbuf, imsg,
2830                                                 msg_data_sz(fragm));
2831                         /*  Prepare buffer for subsequent fragments. */
2832
2833                         set_long_msg_seqno(pbuf, long_msg_seq_no);
2834                         set_fragm_size(pbuf,fragm_sz);
2835                         set_expected_frags(pbuf,exp_fragm_cnt - 1);
2836                 } else {
2837                         warn("Link unable to reassemble fragmented message\n");
2838                 }
2839                 buf_discard(fbuf);
2840                 return 0;
2841         } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2842                 u32 dsz = msg_data_sz(fragm);
2843                 u32 fsz = get_fragm_size(pbuf);
2844                 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2845                 u32 exp_frags = get_expected_frags(pbuf) - 1;
2846                 skb_copy_to_linear_data_offset(pbuf, crs,
2847                                                msg_data(fragm), dsz);
2848                 buf_discard(fbuf);
2849
2850                 /* Is message complete? */
2851
2852                 if (exp_frags == 0) {
2853                         if (prev)
2854                                 prev->next = pbuf->next;
2855                         else
2856                                 *pending = pbuf->next;
2857                         msg_reset_reroute_cnt(buf_msg(pbuf));
2858                         *fb = pbuf;
2859                         *m = buf_msg(pbuf);
2860                         return 1;
2861                 }
2862                 set_expected_frags(pbuf,exp_frags);
2863                 return 0;
2864         }
2865         dbg(" Discarding orphan fragment %x\n",fbuf);
2866         msg_dbg(fragm,"ORPHAN:");
2867         dbg("Pending long buffers:\n");
2868         dbg_print_buf_chain(*pending);
2869         buf_discard(fbuf);
2870         return 0;
2871 }
2872
2873 /**
2874  * link_check_defragm_bufs - flush stale incoming message fragments
2875  * @l_ptr: pointer to link
2876  */
2877
2878 static void link_check_defragm_bufs(struct link *l_ptr)
2879 {
2880         struct sk_buff *prev = NULL;
2881         struct sk_buff *next = NULL;
2882         struct sk_buff *buf = l_ptr->defragm_buf;
2883
2884         if (!buf)
2885                 return;
2886         if (!link_working_working(l_ptr))
2887                 return;
2888         while (buf) {
2889                 u32 cnt = get_timer_cnt(buf);
2890
2891                 next = buf->next;
2892                 if (cnt < 4) {
2893                         incr_timer_cnt(buf);
2894                         prev = buf;
2895                 } else {
2896                         dbg(" Discarding incomplete long buffer\n");
2897                         msg_dbg(buf_msg(buf), "LONG:");
2898                         dbg_print_link(l_ptr, "curr:");
2899                         dbg("Pending long buffers:\n");
2900                         dbg_print_buf_chain(l_ptr->defragm_buf);
2901                         if (prev)
2902                                 prev->next = buf->next;
2903                         else
2904                                 l_ptr->defragm_buf = buf->next;
2905                         buf_discard(buf);
2906                 }
2907                 buf = next;
2908         }
2909 }
2910
2911
2912
2913 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance)
2914 {
2915         l_ptr->tolerance = tolerance;
2916         l_ptr->continuity_interval =
2917                 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2918         l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2919 }
2920
2921
2922 void tipc_link_set_queue_limits(struct link *l_ptr, u32 window)
2923 {
2924         /* Data messages from this node, inclusive FIRST_FRAGM */
2925         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2926         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2927         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2928         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2929         /* Transiting data messages,inclusive FIRST_FRAGM */
2930         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2931         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2932         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2933         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2934         l_ptr->queue_limit[CONN_MANAGER] = 1200;
2935         l_ptr->queue_limit[ROUTE_DISTRIBUTOR] = 1200;
2936         l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2937         l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2938         /* FRAGMENT and LAST_FRAGMENT packets */
2939         l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2940 }
2941
2942 /**
2943  * link_find_link - locate link by name
2944  * @name - ptr to link name string
2945  * @node - ptr to area to be filled with ptr to associated node
2946  *
2947  * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2948  * this also prevents link deletion.
2949  *
2950  * Returns pointer to link (or 0 if invalid link name).
2951  */
2952
2953 static struct link *link_find_link(const char *name, struct node **node)
2954 {
2955         struct link_name link_name_parts;
2956         struct bearer *b_ptr;
2957         struct link *l_ptr;
2958
2959         if (!link_name_validate(name, &link_name_parts))
2960                 return NULL;
2961
2962         b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2963         if (!b_ptr)
2964                 return NULL;
2965
2966         *node = tipc_node_find(link_name_parts.addr_peer);
2967         if (!*node)
2968                 return NULL;
2969
2970         l_ptr = (*node)->links[b_ptr->identity];
2971         if (!l_ptr || strcmp(l_ptr->name, name))
2972                 return NULL;
2973
2974         return l_ptr;
2975 }
2976
2977 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2978                                      u16 cmd)
2979 {
2980         struct tipc_link_config *args;
2981         u32 new_value;
2982         struct link *l_ptr;
2983         struct node *node;
2984         int res;
2985
2986         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2987                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2988
2989         args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2990         new_value = ntohl(args->value);
2991
2992         if (!strcmp(args->name, tipc_bclink_name)) {
2993                 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2994                     (tipc_bclink_set_queue_limits(new_value) == 0))
2995                         return tipc_cfg_reply_none();
2996                 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2997                                                    " (cannot change setting on broadcast link)");
2998         }
2999
3000         read_lock_bh(&tipc_net_lock);
3001         l_ptr = link_find_link(args->name, &node);
3002         if (!l_ptr) {
3003                 read_unlock_bh(&tipc_net_lock);
3004                 return tipc_cfg_reply_error_string("link not found");
3005         }
3006
3007         tipc_node_lock(node);
3008         res = -EINVAL;
3009         switch (cmd) {
3010         case TIPC_CMD_SET_LINK_TOL:
3011                 if ((new_value >= TIPC_MIN_LINK_TOL) &&
3012                     (new_value <= TIPC_MAX_LINK_TOL)) {
3013                         link_set_supervision_props(l_ptr, new_value);
3014                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
3015                                                  0, 0, new_value, 0, 0);
3016                         res = TIPC_OK;
3017                 }
3018                 break;
3019         case TIPC_CMD_SET_LINK_PRI:
3020                 if ((new_value >= TIPC_MIN_LINK_PRI) &&
3021                     (new_value <= TIPC_MAX_LINK_PRI)) {
3022                         l_ptr->priority = new_value;
3023                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
3024                                                  0, 0, 0, new_value, 0);
3025                         res = TIPC_OK;
3026                 }
3027                 break;
3028         case TIPC_CMD_SET_LINK_WINDOW:
3029                 if ((new_value >= TIPC_MIN_LINK_WIN) &&
3030                     (new_value <= TIPC_MAX_LINK_WIN)) {
3031                         tipc_link_set_queue_limits(l_ptr, new_value);
3032                         res = TIPC_OK;
3033                 }
3034                 break;
3035         }
3036         tipc_node_unlock(node);
3037
3038         read_unlock_bh(&tipc_net_lock);
3039         if (res)
3040                 return tipc_cfg_reply_error_string("cannot change link setting");
3041
3042         return tipc_cfg_reply_none();
3043 }
3044
3045 /**
3046  * link_reset_statistics - reset link statistics
3047  * @l_ptr: pointer to link
3048  */
3049
3050 static void link_reset_statistics(struct link *l_ptr)
3051 {
3052         memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
3053         l_ptr->stats.sent_info = l_ptr->next_out_no;
3054         l_ptr->stats.recv_info = l_ptr->next_in_no;
3055 }
3056
3057 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
3058 {
3059         char *link_name;
3060         struct link *l_ptr;
3061         struct node *node;
3062
3063         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3064                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3065
3066         link_name = (char *)TLV_DATA(req_tlv_area);
3067         if (!strcmp(link_name, tipc_bclink_name)) {
3068                 if (tipc_bclink_reset_stats())
3069                         return tipc_cfg_reply_error_string("link not found");
3070                 return tipc_cfg_reply_none();
3071         }
3072
3073         read_lock_bh(&tipc_net_lock);
3074         l_ptr = link_find_link(link_name, &node);
3075         if (!l_ptr) {
3076                 read_unlock_bh(&tipc_net_lock);
3077                 return tipc_cfg_reply_error_string("link not found");
3078         }
3079
3080         tipc_node_lock(node);
3081         link_reset_statistics(l_ptr);
3082         tipc_node_unlock(node);
3083         read_unlock_bh(&tipc_net_lock);
3084         return tipc_cfg_reply_none();
3085 }
3086
3087 /**
3088  * percent - convert count to a percentage of total (rounding up or down)
3089  */
3090
3091 static u32 percent(u32 count, u32 total)
3092 {
3093         return (count * 100 + (total / 2)) / total;
3094 }
3095
3096 /**
3097  * tipc_link_stats - print link statistics
3098  * @name: link name
3099  * @buf: print buffer area
3100  * @buf_size: size of print buffer area
3101  *
3102  * Returns length of print buffer data string (or 0 if error)
3103  */
3104
3105 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
3106 {
3107         struct print_buf pb;
3108         struct link *l_ptr;
3109         struct node *node;
3110         char *status;
3111         u32 profile_total = 0;
3112
3113         if (!strcmp(name, tipc_bclink_name))
3114                 return tipc_bclink_stats(buf, buf_size);
3115
3116         tipc_printbuf_init(&pb, buf, buf_size);
3117
3118         read_lock_bh(&tipc_net_lock);
3119         l_ptr = link_find_link(name, &node);
3120         if (!l_ptr) {
3121                 read_unlock_bh(&tipc_net_lock);
3122                 return 0;
3123         }
3124         tipc_node_lock(node);
3125
3126         if (tipc_link_is_active(l_ptr))
3127                 status = "ACTIVE";
3128         else if (tipc_link_is_up(l_ptr))
3129                 status = "STANDBY";
3130         else
3131                 status = "DEFUNCT";
3132         tipc_printf(&pb, "Link <%s>\n"
3133                          "  %s  MTU:%u  Priority:%u  Tolerance:%u ms"
3134                          "  Window:%u packets\n",
3135                     l_ptr->name, status, link_max_pkt(l_ptr),
3136                     l_ptr->priority, l_ptr->tolerance, l_ptr->queue_limit[0]);
3137         tipc_printf(&pb, "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
3138                     l_ptr->next_in_no - l_ptr->stats.recv_info,
3139                     l_ptr->stats.recv_fragments,
3140                     l_ptr->stats.recv_fragmented,
3141                     l_ptr->stats.recv_bundles,
3142                     l_ptr->stats.recv_bundled);
3143         tipc_printf(&pb, "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
3144                     l_ptr->next_out_no - l_ptr->stats.sent_info,
3145                     l_ptr->stats.sent_fragments,
3146                     l_ptr->stats.sent_fragmented,
3147                     l_ptr->stats.sent_bundles,
3148                     l_ptr->stats.sent_bundled);
3149         profile_total = l_ptr->stats.msg_length_counts;
3150         if (!profile_total)
3151                 profile_total = 1;
3152         tipc_printf(&pb, "  TX profile sample:%u packets  average:%u octets\n"
3153                          "  0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
3154                          "-16354:%u%% -32768:%u%% -66000:%u%%\n",
3155                     l_ptr->stats.msg_length_counts,
3156                     l_ptr->stats.msg_lengths_total / profile_total,
3157                     percent(l_ptr->stats.msg_length_profile[0], profile_total),
3158                     percent(l_ptr->stats.msg_length_profile[1], profile_total),
3159                     percent(l_ptr->stats.msg_length_profile[2], profile_total),
3160                     percent(l_ptr->stats.msg_length_profile[3], profile_total),
3161                     percent(l_ptr->stats.msg_length_profile[4], profile_total),
3162                     percent(l_ptr->stats.msg_length_profile[5], profile_total),
3163                     percent(l_ptr->stats.msg_length_profile[6], profile_total));
3164         tipc_printf(&pb, "  RX states:%u probes:%u naks:%u defs:%u dups:%u\n",
3165                     l_ptr->stats.recv_states,
3166                     l_ptr->stats.recv_probes,
3167                     l_ptr->stats.recv_nacks,
3168                     l_ptr->stats.deferred_recv,
3169                     l_ptr->stats.duplicates);
3170         tipc_printf(&pb, "  TX states:%u probes:%u naks:%u acks:%u dups:%u\n",
3171                     l_ptr->stats.sent_states,
3172                     l_ptr->stats.sent_probes,
3173                     l_ptr->stats.sent_nacks,
3174                     l_ptr->stats.sent_acks,
3175                     l_ptr->stats.retransmitted);
3176         tipc_printf(&pb, "  Congestion bearer:%u link:%u  Send queue max:%u avg:%u\n",
3177                     l_ptr->stats.bearer_congs,
3178                     l_ptr->stats.link_congs,
3179                     l_ptr->stats.max_queue_sz,
3180                     l_ptr->stats.queue_sz_counts
3181                     ? (l_ptr->stats.accu_queue_sz / l_ptr->stats.queue_sz_counts)
3182                     : 0);
3183
3184         tipc_node_unlock(node);
3185         read_unlock_bh(&tipc_net_lock);
3186         return tipc_printbuf_validate(&pb);
3187 }
3188
3189 #define MAX_LINK_STATS_INFO 2000
3190
3191 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
3192 {
3193         struct sk_buff *buf;
3194         struct tlv_desc *rep_tlv;
3195         int str_len;
3196
3197         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3198                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3199
3200         buf = tipc_cfg_reply_alloc(TLV_SPACE(MAX_LINK_STATS_INFO));
3201         if (!buf)
3202                 return NULL;
3203
3204         rep_tlv = (struct tlv_desc *)buf->data;
3205
3206         str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
3207                                   (char *)TLV_DATA(rep_tlv), MAX_LINK_STATS_INFO);
3208         if (!str_len) {
3209                 buf_discard(buf);
3210                 return tipc_cfg_reply_error_string("link not found");
3211         }
3212
3213         skb_put(buf, TLV_SPACE(str_len));
3214         TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
3215
3216         return buf;
3217 }
3218
3219 #if 0
3220 int link_control(const char *name, u32 op, u32 val)
3221 {
3222         int res = -EINVAL;
3223         struct link *l_ptr;
3224         u32 bearer_id;
3225         struct node * node;
3226         u32 a;
3227
3228         a = link_name2addr(name, &bearer_id);
3229         read_lock_bh(&tipc_net_lock);
3230         node = tipc_node_find(a);
3231         if (node) {
3232                 tipc_node_lock(node);
3233                 l_ptr = node->links[bearer_id];
3234                 if (l_ptr) {
3235                         if (op == TIPC_REMOVE_LINK) {
3236                                 struct bearer *b_ptr = l_ptr->b_ptr;
3237                                 spin_lock_bh(&b_ptr->publ.lock);
3238                                 tipc_link_delete(l_ptr);
3239                                 spin_unlock_bh(&b_ptr->publ.lock);
3240                         }
3241                         if (op == TIPC_CMD_BLOCK_LINK) {
3242                                 tipc_link_reset(l_ptr);
3243                                 l_ptr->blocked = 1;
3244                         }
3245                         if (op == TIPC_CMD_UNBLOCK_LINK) {
3246                                 l_ptr->blocked = 0;
3247                         }
3248                         res = TIPC_OK;
3249                 }
3250                 tipc_node_unlock(node);
3251         }
3252         read_unlock_bh(&tipc_net_lock);
3253         return res;
3254 }
3255 #endif
3256
3257 /**
3258  * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
3259  * @dest: network address of destination node
3260  * @selector: used to select from set of active links
3261  *
3262  * If no active link can be found, uses default maximum packet size.
3263  */
3264
3265 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
3266 {
3267         struct node *n_ptr;
3268         struct link *l_ptr;
3269         u32 res = MAX_PKT_DEFAULT;
3270
3271         if (dest == tipc_own_addr)
3272                 return MAX_MSG_SIZE;
3273
3274         read_lock_bh(&tipc_net_lock);
3275         n_ptr = tipc_node_select(dest, selector);
3276         if (n_ptr) {
3277                 tipc_node_lock(n_ptr);
3278                 l_ptr = n_ptr->active_links[selector & 1];
3279                 if (l_ptr)
3280                         res = link_max_pkt(l_ptr);
3281                 tipc_node_unlock(n_ptr);
3282         }
3283         read_unlock_bh(&tipc_net_lock);
3284         return res;
3285 }
3286
3287 #if 0
3288 static void link_dump_rec_queue(struct link *l_ptr)
3289 {
3290         struct sk_buff *crs;
3291
3292         if (!l_ptr->oldest_deferred_in) {
3293                 info("Reception queue empty\n");
3294                 return;
3295         }
3296         info("Contents of Reception queue:\n");
3297         crs = l_ptr->oldest_deferred_in;
3298         while (crs) {
3299                 if (crs->data == (void *)0x0000a3a3) {
3300                         info("buffer %x invalid\n", crs);
3301                         return;
3302                 }
3303                 msg_dbg(buf_msg(crs), "In rec queue: \n");
3304                 crs = crs->next;
3305         }
3306 }
3307 #endif
3308
3309 static void link_dump_send_queue(struct link *l_ptr)
3310 {
3311         if (l_ptr->next_out) {
3312                 info("\nContents of unsent queue:\n");
3313                 dbg_print_buf_chain(l_ptr->next_out);
3314         }
3315         info("\nContents of send queue:\n");
3316         if (l_ptr->first_out) {
3317                 dbg_print_buf_chain(l_ptr->first_out);
3318         }
3319         info("Empty send queue\n");
3320 }
3321
3322 static void link_print(struct link *l_ptr, struct print_buf *buf,
3323                        const char *str)
3324 {
3325         tipc_printf(buf, str);
3326         if (link_reset_reset(l_ptr) || link_reset_unknown(l_ptr))
3327                 return;
3328         tipc_printf(buf, "Link %x<%s>:",
3329                     l_ptr->addr, l_ptr->b_ptr->publ.name);
3330         tipc_printf(buf, ": NXO(%u):", mod(l_ptr->next_out_no));
3331         tipc_printf(buf, "NXI(%u):", mod(l_ptr->next_in_no));
3332         tipc_printf(buf, "SQUE");
3333         if (l_ptr->first_out) {
3334                 tipc_printf(buf, "[%u..", msg_seqno(buf_msg(l_ptr->first_out)));
3335                 if (l_ptr->next_out)
3336                         tipc_printf(buf, "%u..",
3337                                     msg_seqno(buf_msg(l_ptr->next_out)));
3338                 tipc_printf(buf, "%u]",
3339                             msg_seqno(buf_msg
3340                                       (l_ptr->last_out)), l_ptr->out_queue_size);
3341                 if ((mod(msg_seqno(buf_msg(l_ptr->last_out)) -
3342                          msg_seqno(buf_msg(l_ptr->first_out)))
3343                      != (l_ptr->out_queue_size - 1))
3344                     || (l_ptr->last_out->next != NULL)) {
3345                         tipc_printf(buf, "\nSend queue inconsistency\n");
3346                         tipc_printf(buf, "first_out= %x ", l_ptr->first_out);
3347                         tipc_printf(buf, "next_out= %x ", l_ptr->next_out);
3348                         tipc_printf(buf, "last_out= %x ", l_ptr->last_out);
3349                         link_dump_send_queue(l_ptr);
3350                 }
3351         } else
3352                 tipc_printf(buf, "[]");
3353         tipc_printf(buf, "SQSIZ(%u)", l_ptr->out_queue_size);
3354         if (l_ptr->oldest_deferred_in) {
3355                 u32 o = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
3356                 u32 n = msg_seqno(buf_msg(l_ptr->newest_deferred_in));
3357                 tipc_printf(buf, ":RQUE[%u..%u]", o, n);
3358                 if (l_ptr->deferred_inqueue_sz != mod((n + 1) - o)) {
3359                         tipc_printf(buf, ":RQSIZ(%u)",
3360                                     l_ptr->deferred_inqueue_sz);
3361                 }
3362         }
3363         if (link_working_unknown(l_ptr))
3364                 tipc_printf(buf, ":WU");
3365         if (link_reset_reset(l_ptr))
3366                 tipc_printf(buf, ":RR");
3367         if (link_reset_unknown(l_ptr))
3368                 tipc_printf(buf, ":RU");
3369         if (link_working_working(l_ptr))
3370                 tipc_printf(buf, ":WW");
3371         tipc_printf(buf, "\n");
3372 }
3373