]> git.karo-electronics.de Git - mv-sheeva.git/blob - net/tipc/link.c
tipc: Optimize tipc_node_has_active_links()
[mv-sheeva.git] / net / tipc / link.c
1 /*
2  * net/tipc/link.c: TIPC link code
3  *
4  * Copyright (c) 1996-2007, Ericsson AB
5  * Copyright (c) 2004-2007, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36
37 #include "core.h"
38 #include "dbg.h"
39 #include "link.h"
40 #include "net.h"
41 #include "node.h"
42 #include "port.h"
43 #include "addr.h"
44 #include "node_subscr.h"
45 #include "name_distr.h"
46 #include "bearer.h"
47 #include "name_table.h"
48 #include "discover.h"
49 #include "config.h"
50 #include "bcast.h"
51
52
53 /*
54  * Out-of-range value for link session numbers
55  */
56
57 #define INVALID_SESSION 0x10000
58
59 /*
60  * Limit for deferred reception queue:
61  */
62
63 #define DEF_QUEUE_LIMIT 256u
64
65 /*
66  * Link state events:
67  */
68
69 #define  STARTING_EVT    856384768      /* link processing trigger */
70 #define  TRAFFIC_MSG_EVT 560815u        /* rx'd ??? */
71 #define  TIMEOUT_EVT     560817u        /* link timer expired */
72
73 /*
74  * The following two 'message types' is really just implementation
75  * data conveniently stored in the message header.
76  * They must not be considered part of the protocol
77  */
78 #define OPEN_MSG   0
79 #define CLOSED_MSG 1
80
81 /*
82  * State value stored in 'exp_msg_count'
83  */
84
85 #define START_CHANGEOVER 100000u
86
87 /**
88  * struct link_name - deconstructed link name
89  * @addr_local: network address of node at this end
90  * @if_local: name of interface at this end
91  * @addr_peer: network address of node at far end
92  * @if_peer: name of interface at far end
93  */
94
95 struct link_name {
96         u32 addr_local;
97         char if_local[TIPC_MAX_IF_NAME];
98         u32 addr_peer;
99         char if_peer[TIPC_MAX_IF_NAME];
100 };
101
102 #if 0
103
104 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
105
106 /**
107  * struct link_event - link up/down event notification
108  */
109
110 struct link_event {
111         u32 addr;
112         int up;
113         void (*fcn)(u32, char *, int);
114         char name[TIPC_MAX_LINK_NAME];
115 };
116
117 #endif
118
119 static void link_handle_out_of_seq_msg(struct link *l_ptr,
120                                        struct sk_buff *buf);
121 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf);
122 static int  link_recv_changeover_msg(struct link **l_ptr, struct sk_buff **buf);
123 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance);
124 static int  link_send_sections_long(struct port *sender,
125                                     struct iovec const *msg_sect,
126                                     u32 num_sect, u32 destnode);
127 static void link_check_defragm_bufs(struct link *l_ptr);
128 static void link_state_event(struct link *l_ptr, u32 event);
129 static void link_reset_statistics(struct link *l_ptr);
130 static void link_print(struct link *l_ptr, struct print_buf *buf,
131                        const char *str);
132
133 /*
134  * Debugging code used by link routines only
135  *
136  * When debugging link problems on a system that has multiple links,
137  * the standard TIPC debugging routines may not be useful since they
138  * allow the output from multiple links to be intermixed.  For this reason
139  * routines of the form "dbg_link_XXX()" have been created that will capture
140  * debug info into a link's personal print buffer, which can then be dumped
141  * into the TIPC system log (TIPC_LOG) upon request.
142  *
143  * To enable per-link debugging, use LINK_LOG_BUF_SIZE to specify the size
144  * of the print buffer used by each link.  If LINK_LOG_BUF_SIZE is set to 0,
145  * the dbg_link_XXX() routines simply send their output to the standard
146  * debug print buffer (DBG_OUTPUT), if it has been defined; this can be useful
147  * when there is only a single link in the system being debugged.
148  *
149  * Notes:
150  * - When enabled, LINK_LOG_BUF_SIZE should be set to at least TIPC_PB_MIN_SIZE
151  * - "l_ptr" must be valid when using dbg_link_XXX() macros
152  */
153
154 #define LINK_LOG_BUF_SIZE 0
155
156 #define dbg_link(fmt, arg...) \
157         do { \
158                 if (LINK_LOG_BUF_SIZE) \
159                         tipc_printf(&l_ptr->print_buf, fmt, ## arg); \
160         } while (0)
161 #define dbg_link_msg(msg, txt) \
162         do { \
163                 if (LINK_LOG_BUF_SIZE) \
164                         tipc_msg_dbg(&l_ptr->print_buf, msg, txt); \
165         } while (0)
166 #define dbg_link_state(txt) \
167         do { \
168                 if (LINK_LOG_BUF_SIZE) \
169                         link_print(l_ptr, &l_ptr->print_buf, txt); \
170         } while (0)
171 #define dbg_link_dump() do { \
172         if (LINK_LOG_BUF_SIZE) { \
173                 tipc_printf(LOG, "\n\nDumping link <%s>:\n", l_ptr->name); \
174                 tipc_printbuf_move(LOG, &l_ptr->print_buf); \
175         } \
176 } while (0)
177
178 static void dbg_print_link(struct link *l_ptr, const char *str)
179 {
180         if (DBG_OUTPUT != TIPC_NULL)
181                 link_print(l_ptr, DBG_OUTPUT, str);
182 }
183
184 static void dbg_print_buf_chain(struct sk_buff *root_buf)
185 {
186         if (DBG_OUTPUT != TIPC_NULL) {
187                 struct sk_buff *buf = root_buf;
188
189                 while (buf) {
190                         msg_dbg(buf_msg(buf), "In chain: ");
191                         buf = buf->next;
192                 }
193         }
194 }
195
196 /*
197  *  Simple link routines
198  */
199
200 static unsigned int align(unsigned int i)
201 {
202         return (i + 3) & ~3u;
203 }
204
205 static void link_init_max_pkt(struct link *l_ptr)
206 {
207         u32 max_pkt;
208
209         max_pkt = (l_ptr->b_ptr->publ.mtu & ~3);
210         if (max_pkt > MAX_MSG_SIZE)
211                 max_pkt = MAX_MSG_SIZE;
212
213         l_ptr->max_pkt_target = max_pkt;
214         if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
215                 l_ptr->max_pkt = l_ptr->max_pkt_target;
216         else
217                 l_ptr->max_pkt = MAX_PKT_DEFAULT;
218
219         l_ptr->max_pkt_probes = 0;
220 }
221
222 static u32 link_next_sent(struct link *l_ptr)
223 {
224         if (l_ptr->next_out)
225                 return msg_seqno(buf_msg(l_ptr->next_out));
226         return mod(l_ptr->next_out_no);
227 }
228
229 static u32 link_last_sent(struct link *l_ptr)
230 {
231         return mod(link_next_sent(l_ptr) - 1);
232 }
233
234 /*
235  *  Simple non-static link routines (i.e. referenced outside this file)
236  */
237
238 int tipc_link_is_up(struct link *l_ptr)
239 {
240         if (!l_ptr)
241                 return 0;
242         return (link_working_working(l_ptr) || link_working_unknown(l_ptr));
243 }
244
245 int tipc_link_is_active(struct link *l_ptr)
246 {
247         return ((l_ptr->owner->active_links[0] == l_ptr) ||
248                 (l_ptr->owner->active_links[1] == l_ptr));
249 }
250
251 /**
252  * link_name_validate - validate & (optionally) deconstruct link name
253  * @name - ptr to link name string
254  * @name_parts - ptr to area for link name components (or NULL if not needed)
255  *
256  * Returns 1 if link name is valid, otherwise 0.
257  */
258
259 static int link_name_validate(const char *name, struct link_name *name_parts)
260 {
261         char name_copy[TIPC_MAX_LINK_NAME];
262         char *addr_local;
263         char *if_local;
264         char *addr_peer;
265         char *if_peer;
266         char dummy;
267         u32 z_local, c_local, n_local;
268         u32 z_peer, c_peer, n_peer;
269         u32 if_local_len;
270         u32 if_peer_len;
271
272         /* copy link name & ensure length is OK */
273
274         name_copy[TIPC_MAX_LINK_NAME - 1] = 0;
275         /* need above in case non-Posix strncpy() doesn't pad with nulls */
276         strncpy(name_copy, name, TIPC_MAX_LINK_NAME);
277         if (name_copy[TIPC_MAX_LINK_NAME - 1] != 0)
278                 return 0;
279
280         /* ensure all component parts of link name are present */
281
282         addr_local = name_copy;
283         if ((if_local = strchr(addr_local, ':')) == NULL)
284                 return 0;
285         *(if_local++) = 0;
286         if ((addr_peer = strchr(if_local, '-')) == NULL)
287                 return 0;
288         *(addr_peer++) = 0;
289         if_local_len = addr_peer - if_local;
290         if ((if_peer = strchr(addr_peer, ':')) == NULL)
291                 return 0;
292         *(if_peer++) = 0;
293         if_peer_len = strlen(if_peer) + 1;
294
295         /* validate component parts of link name */
296
297         if ((sscanf(addr_local, "%u.%u.%u%c",
298                     &z_local, &c_local, &n_local, &dummy) != 3) ||
299             (sscanf(addr_peer, "%u.%u.%u%c",
300                     &z_peer, &c_peer, &n_peer, &dummy) != 3) ||
301             (z_local > 255) || (c_local > 4095) || (n_local > 4095) ||
302             (z_peer  > 255) || (c_peer  > 4095) || (n_peer  > 4095) ||
303             (if_local_len <= 1) || (if_local_len > TIPC_MAX_IF_NAME) ||
304             (if_peer_len  <= 1) || (if_peer_len  > TIPC_MAX_IF_NAME) ||
305             (strspn(if_local, tipc_alphabet) != (if_local_len - 1)) ||
306             (strspn(if_peer, tipc_alphabet) != (if_peer_len - 1)))
307                 return 0;
308
309         /* return link name components, if necessary */
310
311         if (name_parts) {
312                 name_parts->addr_local = tipc_addr(z_local, c_local, n_local);
313                 strcpy(name_parts->if_local, if_local);
314                 name_parts->addr_peer = tipc_addr(z_peer, c_peer, n_peer);
315                 strcpy(name_parts->if_peer, if_peer);
316         }
317         return 1;
318 }
319
320 /**
321  * link_timeout - handle expiration of link timer
322  * @l_ptr: pointer to link
323  *
324  * This routine must not grab "tipc_net_lock" to avoid a potential deadlock conflict
325  * with tipc_link_delete().  (There is no risk that the node will be deleted by
326  * another thread because tipc_link_delete() always cancels the link timer before
327  * tipc_node_delete() is called.)
328  */
329
330 static void link_timeout(struct link *l_ptr)
331 {
332         tipc_node_lock(l_ptr->owner);
333
334         /* update counters used in statistical profiling of send traffic */
335
336         l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
337         l_ptr->stats.queue_sz_counts++;
338
339         if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
340                 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
341
342         if (l_ptr->first_out) {
343                 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
344                 u32 length = msg_size(msg);
345
346                 if ((msg_user(msg) == MSG_FRAGMENTER) &&
347                     (msg_type(msg) == FIRST_FRAGMENT)) {
348                         length = msg_size(msg_get_wrapped(msg));
349                 }
350                 if (length) {
351                         l_ptr->stats.msg_lengths_total += length;
352                         l_ptr->stats.msg_length_counts++;
353                         if (length <= 64)
354                                 l_ptr->stats.msg_length_profile[0]++;
355                         else if (length <= 256)
356                                 l_ptr->stats.msg_length_profile[1]++;
357                         else if (length <= 1024)
358                                 l_ptr->stats.msg_length_profile[2]++;
359                         else if (length <= 4096)
360                                 l_ptr->stats.msg_length_profile[3]++;
361                         else if (length <= 16384)
362                                 l_ptr->stats.msg_length_profile[4]++;
363                         else if (length <= 32768)
364                                 l_ptr->stats.msg_length_profile[5]++;
365                         else
366                                 l_ptr->stats.msg_length_profile[6]++;
367                 }
368         }
369
370         /* do all other link processing performed on a periodic basis */
371
372         link_check_defragm_bufs(l_ptr);
373
374         link_state_event(l_ptr, TIMEOUT_EVT);
375
376         if (l_ptr->next_out)
377                 tipc_link_push_queue(l_ptr);
378
379         tipc_node_unlock(l_ptr->owner);
380 }
381
382 static void link_set_timer(struct link *l_ptr, u32 time)
383 {
384         k_start_timer(&l_ptr->timer, time);
385 }
386
387 /**
388  * tipc_link_create - create a new link
389  * @b_ptr: pointer to associated bearer
390  * @peer: network address of node at other end of link
391  * @media_addr: media address to use when sending messages over link
392  *
393  * Returns pointer to link.
394  */
395
396 struct link *tipc_link_create(struct bearer *b_ptr, const u32 peer,
397                               const struct tipc_media_addr *media_addr)
398 {
399         struct link *l_ptr;
400         struct tipc_msg *msg;
401         char *if_name;
402
403         l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
404         if (!l_ptr) {
405                 warn("Link creation failed, no memory\n");
406                 return NULL;
407         }
408
409         if (LINK_LOG_BUF_SIZE) {
410                 char *pb = kmalloc(LINK_LOG_BUF_SIZE, GFP_ATOMIC);
411
412                 if (!pb) {
413                         kfree(l_ptr);
414                         warn("Link creation failed, no memory for print buffer\n");
415                         return NULL;
416                 }
417                 tipc_printbuf_init(&l_ptr->print_buf, pb, LINK_LOG_BUF_SIZE);
418         }
419
420         l_ptr->addr = peer;
421         if_name = strchr(b_ptr->publ.name, ':') + 1;
422         sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:",
423                 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
424                 tipc_node(tipc_own_addr),
425                 if_name,
426                 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
427                 /* note: peer i/f is appended to link name by reset/activate */
428         memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
429         l_ptr->checkpoint = 1;
430         l_ptr->b_ptr = b_ptr;
431         link_set_supervision_props(l_ptr, b_ptr->media->tolerance);
432         l_ptr->state = RESET_UNKNOWN;
433
434         l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
435         msg = l_ptr->pmsg;
436         tipc_msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
437         msg_set_size(msg, sizeof(l_ptr->proto_msg));
438         msg_set_session(msg, (tipc_random & 0xffff));
439         msg_set_bearer_id(msg, b_ptr->identity);
440         strcpy((char *)msg_data(msg), if_name);
441
442         l_ptr->priority = b_ptr->priority;
443         tipc_link_set_queue_limits(l_ptr, b_ptr->media->window);
444
445         link_init_max_pkt(l_ptr);
446
447         l_ptr->next_out_no = 1;
448         INIT_LIST_HEAD(&l_ptr->waiting_ports);
449
450         link_reset_statistics(l_ptr);
451
452         l_ptr->owner = tipc_node_attach_link(l_ptr);
453         if (!l_ptr->owner) {
454                 if (LINK_LOG_BUF_SIZE)
455                         kfree(l_ptr->print_buf.buf);
456                 kfree(l_ptr);
457                 return NULL;
458         }
459
460         k_init_timer(&l_ptr->timer, (Handler)link_timeout, (unsigned long)l_ptr);
461         list_add_tail(&l_ptr->link_list, &b_ptr->links);
462         tipc_k_signal((Handler)tipc_link_start, (unsigned long)l_ptr);
463
464         dbg("tipc_link_create(): tolerance = %u,cont intv = %u, abort_limit = %u\n",
465             l_ptr->tolerance, l_ptr->continuity_interval, l_ptr->abort_limit);
466
467         return l_ptr;
468 }
469
470 /**
471  * tipc_link_delete - delete a link
472  * @l_ptr: pointer to link
473  *
474  * Note: 'tipc_net_lock' is write_locked, bearer is locked.
475  * This routine must not grab the node lock until after link timer cancellation
476  * to avoid a potential deadlock situation.
477  */
478
479 void tipc_link_delete(struct link *l_ptr)
480 {
481         if (!l_ptr) {
482                 err("Attempt to delete non-existent link\n");
483                 return;
484         }
485
486         dbg("tipc_link_delete()\n");
487
488         k_cancel_timer(&l_ptr->timer);
489
490         tipc_node_lock(l_ptr->owner);
491         tipc_link_reset(l_ptr);
492         tipc_node_detach_link(l_ptr->owner, l_ptr);
493         tipc_link_stop(l_ptr);
494         list_del_init(&l_ptr->link_list);
495         if (LINK_LOG_BUF_SIZE)
496                 kfree(l_ptr->print_buf.buf);
497         tipc_node_unlock(l_ptr->owner);
498         k_term_timer(&l_ptr->timer);
499         kfree(l_ptr);
500 }
501
502 void tipc_link_start(struct link *l_ptr)
503 {
504         dbg("tipc_link_start %x\n", l_ptr);
505         link_state_event(l_ptr, STARTING_EVT);
506 }
507
508 /**
509  * link_schedule_port - schedule port for deferred sending
510  * @l_ptr: pointer to link
511  * @origport: reference to sending port
512  * @sz: amount of data to be sent
513  *
514  * Schedules port for renewed sending of messages after link congestion
515  * has abated.
516  */
517
518 static int link_schedule_port(struct link *l_ptr, u32 origport, u32 sz)
519 {
520         struct port *p_ptr;
521
522         spin_lock_bh(&tipc_port_list_lock);
523         p_ptr = tipc_port_lock(origport);
524         if (p_ptr) {
525                 if (!p_ptr->wakeup)
526                         goto exit;
527                 if (!list_empty(&p_ptr->wait_list))
528                         goto exit;
529                 p_ptr->publ.congested = 1;
530                 p_ptr->waiting_pkts = 1 + ((sz - 1) / l_ptr->max_pkt);
531                 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
532                 l_ptr->stats.link_congs++;
533 exit:
534                 tipc_port_unlock(p_ptr);
535         }
536         spin_unlock_bh(&tipc_port_list_lock);
537         return -ELINKCONG;
538 }
539
540 void tipc_link_wakeup_ports(struct link *l_ptr, int all)
541 {
542         struct port *p_ptr;
543         struct port *temp_p_ptr;
544         int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
545
546         if (all)
547                 win = 100000;
548         if (win <= 0)
549                 return;
550         if (!spin_trylock_bh(&tipc_port_list_lock))
551                 return;
552         if (link_congested(l_ptr))
553                 goto exit;
554         list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
555                                  wait_list) {
556                 if (win <= 0)
557                         break;
558                 list_del_init(&p_ptr->wait_list);
559                 spin_lock_bh(p_ptr->publ.lock);
560                 p_ptr->publ.congested = 0;
561                 p_ptr->wakeup(&p_ptr->publ);
562                 win -= p_ptr->waiting_pkts;
563                 spin_unlock_bh(p_ptr->publ.lock);
564         }
565
566 exit:
567         spin_unlock_bh(&tipc_port_list_lock);
568 }
569
570 /**
571  * link_release_outqueue - purge link's outbound message queue
572  * @l_ptr: pointer to link
573  */
574
575 static void link_release_outqueue(struct link *l_ptr)
576 {
577         struct sk_buff *buf = l_ptr->first_out;
578         struct sk_buff *next;
579
580         while (buf) {
581                 next = buf->next;
582                 buf_discard(buf);
583                 buf = next;
584         }
585         l_ptr->first_out = NULL;
586         l_ptr->out_queue_size = 0;
587 }
588
589 /**
590  * tipc_link_reset_fragments - purge link's inbound message fragments queue
591  * @l_ptr: pointer to link
592  */
593
594 void tipc_link_reset_fragments(struct link *l_ptr)
595 {
596         struct sk_buff *buf = l_ptr->defragm_buf;
597         struct sk_buff *next;
598
599         while (buf) {
600                 next = buf->next;
601                 buf_discard(buf);
602                 buf = next;
603         }
604         l_ptr->defragm_buf = NULL;
605 }
606
607 /**
608  * tipc_link_stop - purge all inbound and outbound messages associated with link
609  * @l_ptr: pointer to link
610  */
611
612 void tipc_link_stop(struct link *l_ptr)
613 {
614         struct sk_buff *buf;
615         struct sk_buff *next;
616
617         buf = l_ptr->oldest_deferred_in;
618         while (buf) {
619                 next = buf->next;
620                 buf_discard(buf);
621                 buf = next;
622         }
623
624         buf = l_ptr->first_out;
625         while (buf) {
626                 next = buf->next;
627                 buf_discard(buf);
628                 buf = next;
629         }
630
631         tipc_link_reset_fragments(l_ptr);
632
633         buf_discard(l_ptr->proto_msg_queue);
634         l_ptr->proto_msg_queue = NULL;
635 }
636
637 #if 0
638
639 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
640
641 static void link_recv_event(struct link_event *ev)
642 {
643         ev->fcn(ev->addr, ev->name, ev->up);
644         kfree(ev);
645 }
646
647 static void link_send_event(void (*fcn)(u32 a, char *n, int up),
648                             struct link *l_ptr, int up)
649 {
650         struct link_event *ev;
651
652         ev = kmalloc(sizeof(*ev), GFP_ATOMIC);
653         if (!ev) {
654                 warn("Link event allocation failure\n");
655                 return;
656         }
657         ev->addr = l_ptr->addr;
658         ev->up = up;
659         ev->fcn = fcn;
660         memcpy(ev->name, l_ptr->name, TIPC_MAX_LINK_NAME);
661         tipc_k_signal((Handler)link_recv_event, (unsigned long)ev);
662 }
663
664 #else
665
666 #define link_send_event(fcn, l_ptr, up) do { } while (0)
667
668 #endif
669
670 void tipc_link_reset(struct link *l_ptr)
671 {
672         struct sk_buff *buf;
673         u32 prev_state = l_ptr->state;
674         u32 checkpoint = l_ptr->next_in_no;
675         int was_active_link = tipc_link_is_active(l_ptr);
676
677         msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
678
679         /* Link is down, accept any session */
680         l_ptr->peer_session = INVALID_SESSION;
681
682         /* Prepare for max packet size negotiation */
683         link_init_max_pkt(l_ptr);
684
685         l_ptr->state = RESET_UNKNOWN;
686         dbg_link_state("Resetting Link\n");
687
688         if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
689                 return;
690
691         tipc_node_link_down(l_ptr->owner, l_ptr);
692         tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
693 #if 0
694         tipc_printf(TIPC_CONS, "\nReset link <%s>\n", l_ptr->name);
695         dbg_link_dump();
696 #endif
697         if (was_active_link && tipc_node_has_active_links(l_ptr->owner) &&
698             l_ptr->owner->permit_changeover) {
699                 l_ptr->reset_checkpoint = checkpoint;
700                 l_ptr->exp_msg_count = START_CHANGEOVER;
701         }
702
703         /* Clean up all queues: */
704
705         link_release_outqueue(l_ptr);
706         buf_discard(l_ptr->proto_msg_queue);
707         l_ptr->proto_msg_queue = NULL;
708         buf = l_ptr->oldest_deferred_in;
709         while (buf) {
710                 struct sk_buff *next = buf->next;
711                 buf_discard(buf);
712                 buf = next;
713         }
714         if (!list_empty(&l_ptr->waiting_ports))
715                 tipc_link_wakeup_ports(l_ptr, 1);
716
717         l_ptr->retransm_queue_head = 0;
718         l_ptr->retransm_queue_size = 0;
719         l_ptr->last_out = NULL;
720         l_ptr->first_out = NULL;
721         l_ptr->next_out = NULL;
722         l_ptr->unacked_window = 0;
723         l_ptr->checkpoint = 1;
724         l_ptr->next_out_no = 1;
725         l_ptr->deferred_inqueue_sz = 0;
726         l_ptr->oldest_deferred_in = NULL;
727         l_ptr->newest_deferred_in = NULL;
728         l_ptr->fsm_msg_cnt = 0;
729         l_ptr->stale_count = 0;
730         link_reset_statistics(l_ptr);
731
732         link_send_event(tipc_cfg_link_event, l_ptr, 0);
733         if (!in_own_cluster(l_ptr->addr))
734                 link_send_event(tipc_disc_link_event, l_ptr, 0);
735 }
736
737
738 static void link_activate(struct link *l_ptr)
739 {
740         l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
741         tipc_node_link_up(l_ptr->owner, l_ptr);
742         tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
743         link_send_event(tipc_cfg_link_event, l_ptr, 1);
744         if (!in_own_cluster(l_ptr->addr))
745                 link_send_event(tipc_disc_link_event, l_ptr, 1);
746 }
747
748 /**
749  * link_state_event - link finite state machine
750  * @l_ptr: pointer to link
751  * @event: state machine event to process
752  */
753
754 static void link_state_event(struct link *l_ptr, unsigned event)
755 {
756         struct link *other;
757         u32 cont_intv = l_ptr->continuity_interval;
758
759         if (!l_ptr->started && (event != STARTING_EVT))
760                 return;         /* Not yet. */
761
762         if (link_blocked(l_ptr)) {
763                 if (event == TIMEOUT_EVT) {
764                         link_set_timer(l_ptr, cont_intv);
765                 }
766                 return;   /* Changeover going on */
767         }
768         dbg_link("STATE_EV: <%s> ", l_ptr->name);
769
770         switch (l_ptr->state) {
771         case WORKING_WORKING:
772                 dbg_link("WW/");
773                 switch (event) {
774                 case TRAFFIC_MSG_EVT:
775                         dbg_link("TRF-");
776                         /* fall through */
777                 case ACTIVATE_MSG:
778                         dbg_link("ACT\n");
779                         break;
780                 case TIMEOUT_EVT:
781                         dbg_link("TIM ");
782                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
783                                 l_ptr->checkpoint = l_ptr->next_in_no;
784                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
785                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
786                                                                  0, 0, 0, 0, 0);
787                                         l_ptr->fsm_msg_cnt++;
788                                 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
789                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
790                                                                  1, 0, 0, 0, 0);
791                                         l_ptr->fsm_msg_cnt++;
792                                 }
793                                 link_set_timer(l_ptr, cont_intv);
794                                 break;
795                         }
796                         dbg_link(" -> WU\n");
797                         l_ptr->state = WORKING_UNKNOWN;
798                         l_ptr->fsm_msg_cnt = 0;
799                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
800                         l_ptr->fsm_msg_cnt++;
801                         link_set_timer(l_ptr, cont_intv / 4);
802                         break;
803                 case RESET_MSG:
804                         dbg_link("RES -> RR\n");
805                         info("Resetting link <%s>, requested by peer\n",
806                              l_ptr->name);
807                         tipc_link_reset(l_ptr);
808                         l_ptr->state = RESET_RESET;
809                         l_ptr->fsm_msg_cnt = 0;
810                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
811                         l_ptr->fsm_msg_cnt++;
812                         link_set_timer(l_ptr, cont_intv);
813                         break;
814                 default:
815                         err("Unknown link event %u in WW state\n", event);
816                 }
817                 break;
818         case WORKING_UNKNOWN:
819                 dbg_link("WU/");
820                 switch (event) {
821                 case TRAFFIC_MSG_EVT:
822                         dbg_link("TRF-");
823                 case ACTIVATE_MSG:
824                         dbg_link("ACT -> WW\n");
825                         l_ptr->state = WORKING_WORKING;
826                         l_ptr->fsm_msg_cnt = 0;
827                         link_set_timer(l_ptr, cont_intv);
828                         break;
829                 case RESET_MSG:
830                         dbg_link("RES -> RR\n");
831                         info("Resetting link <%s>, requested by peer "
832                              "while probing\n", l_ptr->name);
833                         tipc_link_reset(l_ptr);
834                         l_ptr->state = RESET_RESET;
835                         l_ptr->fsm_msg_cnt = 0;
836                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
837                         l_ptr->fsm_msg_cnt++;
838                         link_set_timer(l_ptr, cont_intv);
839                         break;
840                 case TIMEOUT_EVT:
841                         dbg_link("TIM ");
842                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
843                                 dbg_link("-> WW\n");
844                                 l_ptr->state = WORKING_WORKING;
845                                 l_ptr->fsm_msg_cnt = 0;
846                                 l_ptr->checkpoint = l_ptr->next_in_no;
847                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
848                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
849                                                                  0, 0, 0, 0, 0);
850                                         l_ptr->fsm_msg_cnt++;
851                                 }
852                                 link_set_timer(l_ptr, cont_intv);
853                         } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
854                                 dbg_link("Probing %u/%u,timer = %u ms)\n",
855                                          l_ptr->fsm_msg_cnt, l_ptr->abort_limit,
856                                          cont_intv / 4);
857                                 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
858                                                          1, 0, 0, 0, 0);
859                                 l_ptr->fsm_msg_cnt++;
860                                 link_set_timer(l_ptr, cont_intv / 4);
861                         } else {        /* Link has failed */
862                                 dbg_link("-> RU (%u probes unanswered)\n",
863                                          l_ptr->fsm_msg_cnt);
864                                 warn("Resetting link <%s>, peer not responding\n",
865                                      l_ptr->name);
866                                 tipc_link_reset(l_ptr);
867                                 l_ptr->state = RESET_UNKNOWN;
868                                 l_ptr->fsm_msg_cnt = 0;
869                                 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
870                                                          0, 0, 0, 0, 0);
871                                 l_ptr->fsm_msg_cnt++;
872                                 link_set_timer(l_ptr, cont_intv);
873                         }
874                         break;
875                 default:
876                         err("Unknown link event %u in WU state\n", event);
877                 }
878                 break;
879         case RESET_UNKNOWN:
880                 dbg_link("RU/");
881                 switch (event) {
882                 case TRAFFIC_MSG_EVT:
883                         dbg_link("TRF-\n");
884                         break;
885                 case ACTIVATE_MSG:
886                         other = l_ptr->owner->active_links[0];
887                         if (other && link_working_unknown(other)) {
888                                 dbg_link("ACT\n");
889                                 break;
890                         }
891                         dbg_link("ACT -> WW\n");
892                         l_ptr->state = WORKING_WORKING;
893                         l_ptr->fsm_msg_cnt = 0;
894                         link_activate(l_ptr);
895                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
896                         l_ptr->fsm_msg_cnt++;
897                         link_set_timer(l_ptr, cont_intv);
898                         break;
899                 case RESET_MSG:
900                         dbg_link("RES\n");
901                         dbg_link(" -> RR\n");
902                         l_ptr->state = RESET_RESET;
903                         l_ptr->fsm_msg_cnt = 0;
904                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
905                         l_ptr->fsm_msg_cnt++;
906                         link_set_timer(l_ptr, cont_intv);
907                         break;
908                 case STARTING_EVT:
909                         dbg_link("START-");
910                         l_ptr->started = 1;
911                         /* fall through */
912                 case TIMEOUT_EVT:
913                         dbg_link("TIM\n");
914                         tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
915                         l_ptr->fsm_msg_cnt++;
916                         link_set_timer(l_ptr, cont_intv);
917                         break;
918                 default:
919                         err("Unknown link event %u in RU state\n", event);
920                 }
921                 break;
922         case RESET_RESET:
923                 dbg_link("RR/ ");
924                 switch (event) {
925                 case TRAFFIC_MSG_EVT:
926                         dbg_link("TRF-");
927                         /* fall through */
928                 case ACTIVATE_MSG:
929                         other = l_ptr->owner->active_links[0];
930                         if (other && link_working_unknown(other)) {
931                                 dbg_link("ACT\n");
932                                 break;
933                         }
934                         dbg_link("ACT -> WW\n");
935                         l_ptr->state = WORKING_WORKING;
936                         l_ptr->fsm_msg_cnt = 0;
937                         link_activate(l_ptr);
938                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
939                         l_ptr->fsm_msg_cnt++;
940                         link_set_timer(l_ptr, cont_intv);
941                         break;
942                 case RESET_MSG:
943                         dbg_link("RES\n");
944                         break;
945                 case TIMEOUT_EVT:
946                         dbg_link("TIM\n");
947                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
948                         l_ptr->fsm_msg_cnt++;
949                         link_set_timer(l_ptr, cont_intv);
950                         dbg_link("fsm_msg_cnt %u\n", l_ptr->fsm_msg_cnt);
951                         break;
952                 default:
953                         err("Unknown link event %u in RR state\n", event);
954                 }
955                 break;
956         default:
957                 err("Unknown link state %u/%u\n", l_ptr->state, event);
958         }
959 }
960
961 /*
962  * link_bundle_buf(): Append contents of a buffer to
963  * the tail of an existing one.
964  */
965
966 static int link_bundle_buf(struct link *l_ptr,
967                            struct sk_buff *bundler,
968                            struct sk_buff *buf)
969 {
970         struct tipc_msg *bundler_msg = buf_msg(bundler);
971         struct tipc_msg *msg = buf_msg(buf);
972         u32 size = msg_size(msg);
973         u32 bundle_size = msg_size(bundler_msg);
974         u32 to_pos = align(bundle_size);
975         u32 pad = to_pos - bundle_size;
976
977         if (msg_user(bundler_msg) != MSG_BUNDLER)
978                 return 0;
979         if (msg_type(bundler_msg) != OPEN_MSG)
980                 return 0;
981         if (skb_tailroom(bundler) < (pad + size))
982                 return 0;
983         if (l_ptr->max_pkt < (to_pos + size))
984                 return 0;
985
986         skb_put(bundler, pad + size);
987         skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
988         msg_set_size(bundler_msg, to_pos + size);
989         msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
990         dbg("Packed msg # %u(%u octets) into pos %u in buf(#%u)\n",
991             msg_msgcnt(bundler_msg), size, to_pos, msg_seqno(bundler_msg));
992         msg_dbg(msg, "PACKD:");
993         buf_discard(buf);
994         l_ptr->stats.sent_bundled++;
995         return 1;
996 }
997
998 static void link_add_to_outqueue(struct link *l_ptr,
999                                  struct sk_buff *buf,
1000                                  struct tipc_msg *msg)
1001 {
1002         u32 ack = mod(l_ptr->next_in_no - 1);
1003         u32 seqno = mod(l_ptr->next_out_no++);
1004
1005         msg_set_word(msg, 2, ((ack << 16) | seqno));
1006         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1007         buf->next = NULL;
1008         if (l_ptr->first_out) {
1009                 l_ptr->last_out->next = buf;
1010                 l_ptr->last_out = buf;
1011         } else
1012                 l_ptr->first_out = l_ptr->last_out = buf;
1013         l_ptr->out_queue_size++;
1014 }
1015
1016 /*
1017  * tipc_link_send_buf() is the 'full path' for messages, called from
1018  * inside TIPC when the 'fast path' in tipc_send_buf
1019  * has failed, and from link_send()
1020  */
1021
1022 int tipc_link_send_buf(struct link *l_ptr, struct sk_buff *buf)
1023 {
1024         struct tipc_msg *msg = buf_msg(buf);
1025         u32 size = msg_size(msg);
1026         u32 dsz = msg_data_sz(msg);
1027         u32 queue_size = l_ptr->out_queue_size;
1028         u32 imp = tipc_msg_tot_importance(msg);
1029         u32 queue_limit = l_ptr->queue_limit[imp];
1030         u32 max_packet = l_ptr->max_pkt;
1031
1032         msg_set_prevnode(msg, tipc_own_addr);   /* If routed message */
1033
1034         /* Match msg importance against queue limits: */
1035
1036         if (unlikely(queue_size >= queue_limit)) {
1037                 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
1038                         return link_schedule_port(l_ptr, msg_origport(msg),
1039                                                   size);
1040                 }
1041                 msg_dbg(msg, "TIPC: Congestion, throwing away\n");
1042                 buf_discard(buf);
1043                 if (imp > CONN_MANAGER) {
1044                         warn("Resetting link <%s>, send queue full", l_ptr->name);
1045                         tipc_link_reset(l_ptr);
1046                 }
1047                 return dsz;
1048         }
1049
1050         /* Fragmentation needed ? */
1051
1052         if (size > max_packet)
1053                 return tipc_link_send_long_buf(l_ptr, buf);
1054
1055         /* Packet can be queued or sent: */
1056
1057         if (queue_size > l_ptr->stats.max_queue_sz)
1058                 l_ptr->stats.max_queue_sz = queue_size;
1059
1060         if (likely(!tipc_bearer_congested(l_ptr->b_ptr, l_ptr) &&
1061                    !link_congested(l_ptr))) {
1062                 link_add_to_outqueue(l_ptr, buf, msg);
1063
1064                 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr))) {
1065                         l_ptr->unacked_window = 0;
1066                 } else {
1067                         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1068                         l_ptr->stats.bearer_congs++;
1069                         l_ptr->next_out = buf;
1070                 }
1071                 return dsz;
1072         }
1073         /* Congestion: can message be bundled ?: */
1074
1075         if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
1076             (msg_user(msg) != MSG_FRAGMENTER)) {
1077
1078                 /* Try adding message to an existing bundle */
1079
1080                 if (l_ptr->next_out &&
1081                     link_bundle_buf(l_ptr, l_ptr->last_out, buf)) {
1082                         tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1083                         return dsz;
1084                 }
1085
1086                 /* Try creating a new bundle */
1087
1088                 if (size <= max_packet * 2 / 3) {
1089                         struct sk_buff *bundler = buf_acquire(max_packet);
1090                         struct tipc_msg bundler_hdr;
1091
1092                         if (bundler) {
1093                                 tipc_msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
1094                                          INT_H_SIZE, l_ptr->addr);
1095                                 skb_copy_to_linear_data(bundler, &bundler_hdr,
1096                                                         INT_H_SIZE);
1097                                 skb_trim(bundler, INT_H_SIZE);
1098                                 link_bundle_buf(l_ptr, bundler, buf);
1099                                 buf = bundler;
1100                                 msg = buf_msg(buf);
1101                                 l_ptr->stats.sent_bundles++;
1102                         }
1103                 }
1104         }
1105         if (!l_ptr->next_out)
1106                 l_ptr->next_out = buf;
1107         link_add_to_outqueue(l_ptr, buf, msg);
1108         tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1109         return dsz;
1110 }
1111
1112 /*
1113  * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
1114  * not been selected yet, and the the owner node is not locked
1115  * Called by TIPC internal users, e.g. the name distributor
1116  */
1117
1118 int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
1119 {
1120         struct link *l_ptr;
1121         struct tipc_node *n_ptr;
1122         int res = -ELINKCONG;
1123
1124         read_lock_bh(&tipc_net_lock);
1125         n_ptr = tipc_node_select(dest, selector);
1126         if (n_ptr) {
1127                 tipc_node_lock(n_ptr);
1128                 l_ptr = n_ptr->active_links[selector & 1];
1129                 if (l_ptr) {
1130                         dbg("tipc_link_send: found link %x for dest %x\n", l_ptr, dest);
1131                         res = tipc_link_send_buf(l_ptr, buf);
1132                 } else {
1133                         dbg("Attempt to send msg to unreachable node:\n");
1134                         msg_dbg(buf_msg(buf),">>>");
1135                         buf_discard(buf);
1136                 }
1137                 tipc_node_unlock(n_ptr);
1138         } else {
1139                 dbg("Attempt to send msg to unknown node:\n");
1140                 msg_dbg(buf_msg(buf),">>>");
1141                 buf_discard(buf);
1142         }
1143         read_unlock_bh(&tipc_net_lock);
1144         return res;
1145 }
1146
1147 /*
1148  * link_send_buf_fast: Entry for data messages where the
1149  * destination link is known and the header is complete,
1150  * inclusive total message length. Very time critical.
1151  * Link is locked. Returns user data length.
1152  */
1153
1154 static int link_send_buf_fast(struct link *l_ptr, struct sk_buff *buf,
1155                               u32 *used_max_pkt)
1156 {
1157         struct tipc_msg *msg = buf_msg(buf);
1158         int res = msg_data_sz(msg);
1159
1160         if (likely(!link_congested(l_ptr))) {
1161                 if (likely(msg_size(msg) <= l_ptr->max_pkt)) {
1162                         if (likely(list_empty(&l_ptr->b_ptr->cong_links))) {
1163                                 link_add_to_outqueue(l_ptr, buf, msg);
1164                                 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf,
1165                                                             &l_ptr->media_addr))) {
1166                                         l_ptr->unacked_window = 0;
1167                                         msg_dbg(msg,"SENT_FAST:");
1168                                         return res;
1169                                 }
1170                                 dbg("failed sent fast...\n");
1171                                 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1172                                 l_ptr->stats.bearer_congs++;
1173                                 l_ptr->next_out = buf;
1174                                 return res;
1175                         }
1176                 }
1177                 else
1178                         *used_max_pkt = l_ptr->max_pkt;
1179         }
1180         return tipc_link_send_buf(l_ptr, buf);  /* All other cases */
1181 }
1182
1183 /*
1184  * tipc_send_buf_fast: Entry for data messages where the
1185  * destination node is known and the header is complete,
1186  * inclusive total message length.
1187  * Returns user data length.
1188  */
1189 int tipc_send_buf_fast(struct sk_buff *buf, u32 destnode)
1190 {
1191         struct link *l_ptr;
1192         struct tipc_node *n_ptr;
1193         int res;
1194         u32 selector = msg_origport(buf_msg(buf)) & 1;
1195         u32 dummy;
1196
1197         if (destnode == tipc_own_addr)
1198                 return tipc_port_recv_msg(buf);
1199
1200         read_lock_bh(&tipc_net_lock);
1201         n_ptr = tipc_node_select(destnode, selector);
1202         if (likely(n_ptr)) {
1203                 tipc_node_lock(n_ptr);
1204                 l_ptr = n_ptr->active_links[selector];
1205                 dbg("send_fast: buf %x selected %x, destnode = %x\n",
1206                     buf, l_ptr, destnode);
1207                 if (likely(l_ptr)) {
1208                         res = link_send_buf_fast(l_ptr, buf, &dummy);
1209                         tipc_node_unlock(n_ptr);
1210                         read_unlock_bh(&tipc_net_lock);
1211                         return res;
1212                 }
1213                 tipc_node_unlock(n_ptr);
1214         }
1215         read_unlock_bh(&tipc_net_lock);
1216         res = msg_data_sz(buf_msg(buf));
1217         tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1218         return res;
1219 }
1220
1221
1222 /*
1223  * tipc_link_send_sections_fast: Entry for messages where the
1224  * destination processor is known and the header is complete,
1225  * except for total message length.
1226  * Returns user data length or errno.
1227  */
1228 int tipc_link_send_sections_fast(struct port *sender,
1229                                  struct iovec const *msg_sect,
1230                                  const u32 num_sect,
1231                                  u32 destaddr)
1232 {
1233         struct tipc_msg *hdr = &sender->publ.phdr;
1234         struct link *l_ptr;
1235         struct sk_buff *buf;
1236         struct tipc_node *node;
1237         int res;
1238         u32 selector = msg_origport(hdr) & 1;
1239
1240 again:
1241         /*
1242          * Try building message using port's max_pkt hint.
1243          * (Must not hold any locks while building message.)
1244          */
1245
1246         res = tipc_msg_build(hdr, msg_sect, num_sect, sender->publ.max_pkt,
1247                         !sender->user_port, &buf);
1248
1249         read_lock_bh(&tipc_net_lock);
1250         node = tipc_node_select(destaddr, selector);
1251         if (likely(node)) {
1252                 tipc_node_lock(node);
1253                 l_ptr = node->active_links[selector];
1254                 if (likely(l_ptr)) {
1255                         if (likely(buf)) {
1256                                 res = link_send_buf_fast(l_ptr, buf,
1257                                                          &sender->publ.max_pkt);
1258                                 if (unlikely(res < 0))
1259                                         buf_discard(buf);
1260 exit:
1261                                 tipc_node_unlock(node);
1262                                 read_unlock_bh(&tipc_net_lock);
1263                                 return res;
1264                         }
1265
1266                         /* Exit if build request was invalid */
1267
1268                         if (unlikely(res < 0))
1269                                 goto exit;
1270
1271                         /* Exit if link (or bearer) is congested */
1272
1273                         if (link_congested(l_ptr) ||
1274                             !list_empty(&l_ptr->b_ptr->cong_links)) {
1275                                 res = link_schedule_port(l_ptr,
1276                                                          sender->publ.ref, res);
1277                                 goto exit;
1278                         }
1279
1280                         /*
1281                          * Message size exceeds max_pkt hint; update hint,
1282                          * then re-try fast path or fragment the message
1283                          */
1284
1285                         sender->publ.max_pkt = l_ptr->max_pkt;
1286                         tipc_node_unlock(node);
1287                         read_unlock_bh(&tipc_net_lock);
1288
1289
1290                         if ((msg_hdr_sz(hdr) + res) <= sender->publ.max_pkt)
1291                                 goto again;
1292
1293                         return link_send_sections_long(sender, msg_sect,
1294                                                        num_sect, destaddr);
1295                 }
1296                 tipc_node_unlock(node);
1297         }
1298         read_unlock_bh(&tipc_net_lock);
1299
1300         /* Couldn't find a link to the destination node */
1301
1302         if (buf)
1303                 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1304         if (res >= 0)
1305                 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1306                                                  TIPC_ERR_NO_NODE);
1307         return res;
1308 }
1309
1310 /*
1311  * link_send_sections_long(): Entry for long messages where the
1312  * destination node is known and the header is complete,
1313  * inclusive total message length.
1314  * Link and bearer congestion status have been checked to be ok,
1315  * and are ignored if they change.
1316  *
1317  * Note that fragments do not use the full link MTU so that they won't have
1318  * to undergo refragmentation if link changeover causes them to be sent
1319  * over another link with an additional tunnel header added as prefix.
1320  * (Refragmentation will still occur if the other link has a smaller MTU.)
1321  *
1322  * Returns user data length or errno.
1323  */
1324 static int link_send_sections_long(struct port *sender,
1325                                    struct iovec const *msg_sect,
1326                                    u32 num_sect,
1327                                    u32 destaddr)
1328 {
1329         struct link *l_ptr;
1330         struct tipc_node *node;
1331         struct tipc_msg *hdr = &sender->publ.phdr;
1332         u32 dsz = msg_data_sz(hdr);
1333         u32 max_pkt,fragm_sz,rest;
1334         struct tipc_msg fragm_hdr;
1335         struct sk_buff *buf,*buf_chain,*prev;
1336         u32 fragm_crs,fragm_rest,hsz,sect_rest;
1337         const unchar *sect_crs;
1338         int curr_sect;
1339         u32 fragm_no;
1340
1341 again:
1342         fragm_no = 1;
1343         max_pkt = sender->publ.max_pkt - INT_H_SIZE;
1344                 /* leave room for tunnel header in case of link changeover */
1345         fragm_sz = max_pkt - INT_H_SIZE;
1346                 /* leave room for fragmentation header in each fragment */
1347         rest = dsz;
1348         fragm_crs = 0;
1349         fragm_rest = 0;
1350         sect_rest = 0;
1351         sect_crs = NULL;
1352         curr_sect = -1;
1353
1354         /* Prepare reusable fragment header: */
1355
1356         msg_dbg(hdr, ">FRAGMENTING>");
1357         tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1358                  INT_H_SIZE, msg_destnode(hdr));
1359         msg_set_link_selector(&fragm_hdr, sender->publ.ref);
1360         msg_set_size(&fragm_hdr, max_pkt);
1361         msg_set_fragm_no(&fragm_hdr, 1);
1362
1363         /* Prepare header of first fragment: */
1364
1365         buf_chain = buf = buf_acquire(max_pkt);
1366         if (!buf)
1367                 return -ENOMEM;
1368         buf->next = NULL;
1369         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1370         hsz = msg_hdr_sz(hdr);
1371         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1372         msg_dbg(buf_msg(buf), ">BUILD>");
1373
1374         /* Chop up message: */
1375
1376         fragm_crs = INT_H_SIZE + hsz;
1377         fragm_rest = fragm_sz - hsz;
1378
1379         do {            /* For all sections */
1380                 u32 sz;
1381
1382                 if (!sect_rest) {
1383                         sect_rest = msg_sect[++curr_sect].iov_len;
1384                         sect_crs = (const unchar *)msg_sect[curr_sect].iov_base;
1385                 }
1386
1387                 if (sect_rest < fragm_rest)
1388                         sz = sect_rest;
1389                 else
1390                         sz = fragm_rest;
1391
1392                 if (likely(!sender->user_port)) {
1393                         if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1394 error:
1395                                 for (; buf_chain; buf_chain = buf) {
1396                                         buf = buf_chain->next;
1397                                         buf_discard(buf_chain);
1398                                 }
1399                                 return -EFAULT;
1400                         }
1401                 } else
1402                         skb_copy_to_linear_data_offset(buf, fragm_crs,
1403                                                        sect_crs, sz);
1404                 sect_crs += sz;
1405                 sect_rest -= sz;
1406                 fragm_crs += sz;
1407                 fragm_rest -= sz;
1408                 rest -= sz;
1409
1410                 if (!fragm_rest && rest) {
1411
1412                         /* Initiate new fragment: */
1413                         if (rest <= fragm_sz) {
1414                                 fragm_sz = rest;
1415                                 msg_set_type(&fragm_hdr,LAST_FRAGMENT);
1416                         } else {
1417                                 msg_set_type(&fragm_hdr, FRAGMENT);
1418                         }
1419                         msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1420                         msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1421                         prev = buf;
1422                         buf = buf_acquire(fragm_sz + INT_H_SIZE);
1423                         if (!buf)
1424                                 goto error;
1425
1426                         buf->next = NULL;
1427                         prev->next = buf;
1428                         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1429                         fragm_crs = INT_H_SIZE;
1430                         fragm_rest = fragm_sz;
1431                         msg_dbg(buf_msg(buf),"  >BUILD>");
1432                 }
1433         }
1434         while (rest > 0);
1435
1436         /*
1437          * Now we have a buffer chain. Select a link and check
1438          * that packet size is still OK
1439          */
1440         node = tipc_node_select(destaddr, sender->publ.ref & 1);
1441         if (likely(node)) {
1442                 tipc_node_lock(node);
1443                 l_ptr = node->active_links[sender->publ.ref & 1];
1444                 if (!l_ptr) {
1445                         tipc_node_unlock(node);
1446                         goto reject;
1447                 }
1448                 if (l_ptr->max_pkt < max_pkt) {
1449                         sender->publ.max_pkt = l_ptr->max_pkt;
1450                         tipc_node_unlock(node);
1451                         for (; buf_chain; buf_chain = buf) {
1452                                 buf = buf_chain->next;
1453                                 buf_discard(buf_chain);
1454                         }
1455                         goto again;
1456                 }
1457         } else {
1458 reject:
1459                 for (; buf_chain; buf_chain = buf) {
1460                         buf = buf_chain->next;
1461                         buf_discard(buf_chain);
1462                 }
1463                 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1464                                                  TIPC_ERR_NO_NODE);
1465         }
1466
1467         /* Append whole chain to send queue: */
1468
1469         buf = buf_chain;
1470         l_ptr->long_msg_seq_no = mod(l_ptr->long_msg_seq_no + 1);
1471         if (!l_ptr->next_out)
1472                 l_ptr->next_out = buf_chain;
1473         l_ptr->stats.sent_fragmented++;
1474         while (buf) {
1475                 struct sk_buff *next = buf->next;
1476                 struct tipc_msg *msg = buf_msg(buf);
1477
1478                 l_ptr->stats.sent_fragments++;
1479                 msg_set_long_msgno(msg, l_ptr->long_msg_seq_no);
1480                 link_add_to_outqueue(l_ptr, buf, msg);
1481                 msg_dbg(msg, ">ADD>");
1482                 buf = next;
1483         }
1484
1485         /* Send it, if possible: */
1486
1487         tipc_link_push_queue(l_ptr);
1488         tipc_node_unlock(node);
1489         return dsz;
1490 }
1491
1492 /*
1493  * tipc_link_push_packet: Push one unsent packet to the media
1494  */
1495 u32 tipc_link_push_packet(struct link *l_ptr)
1496 {
1497         struct sk_buff *buf = l_ptr->first_out;
1498         u32 r_q_size = l_ptr->retransm_queue_size;
1499         u32 r_q_head = l_ptr->retransm_queue_head;
1500
1501         /* Step to position where retransmission failed, if any,    */
1502         /* consider that buffers may have been released in meantime */
1503
1504         if (r_q_size && buf) {
1505                 u32 last = lesser(mod(r_q_head + r_q_size),
1506                                   link_last_sent(l_ptr));
1507                 u32 first = msg_seqno(buf_msg(buf));
1508
1509                 while (buf && less(first, r_q_head)) {
1510                         first = mod(first + 1);
1511                         buf = buf->next;
1512                 }
1513                 l_ptr->retransm_queue_head = r_q_head = first;
1514                 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1515         }
1516
1517         /* Continue retransmission now, if there is anything: */
1518
1519         if (r_q_size && buf) {
1520                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1521                 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1522                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1523                         msg_dbg(buf_msg(buf), ">DEF-RETR>");
1524                         l_ptr->retransm_queue_head = mod(++r_q_head);
1525                         l_ptr->retransm_queue_size = --r_q_size;
1526                         l_ptr->stats.retransmitted++;
1527                         return 0;
1528                 } else {
1529                         l_ptr->stats.bearer_congs++;
1530                         msg_dbg(buf_msg(buf), "|>DEF-RETR>");
1531                         return PUSH_FAILED;
1532                 }
1533         }
1534
1535         /* Send deferred protocol message, if any: */
1536
1537         buf = l_ptr->proto_msg_queue;
1538         if (buf) {
1539                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1540                 msg_set_bcast_ack(buf_msg(buf),l_ptr->owner->bclink.last_in);
1541                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1542                         msg_dbg(buf_msg(buf), ">DEF-PROT>");
1543                         l_ptr->unacked_window = 0;
1544                         buf_discard(buf);
1545                         l_ptr->proto_msg_queue = NULL;
1546                         return 0;
1547                 } else {
1548                         msg_dbg(buf_msg(buf), "|>DEF-PROT>");
1549                         l_ptr->stats.bearer_congs++;
1550                         return PUSH_FAILED;
1551                 }
1552         }
1553
1554         /* Send one deferred data message, if send window not full: */
1555
1556         buf = l_ptr->next_out;
1557         if (buf) {
1558                 struct tipc_msg *msg = buf_msg(buf);
1559                 u32 next = msg_seqno(msg);
1560                 u32 first = msg_seqno(buf_msg(l_ptr->first_out));
1561
1562                 if (mod(next - first) < l_ptr->queue_limit[0]) {
1563                         msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1564                         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1565                         if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1566                                 if (msg_user(msg) == MSG_BUNDLER)
1567                                         msg_set_type(msg, CLOSED_MSG);
1568                                 msg_dbg(msg, ">PUSH-DATA>");
1569                                 l_ptr->next_out = buf->next;
1570                                 return 0;
1571                         } else {
1572                                 msg_dbg(msg, "|PUSH-DATA|");
1573                                 l_ptr->stats.bearer_congs++;
1574                                 return PUSH_FAILED;
1575                         }
1576                 }
1577         }
1578         return PUSH_FINISHED;
1579 }
1580
1581 /*
1582  * push_queue(): push out the unsent messages of a link where
1583  *               congestion has abated. Node is locked
1584  */
1585 void tipc_link_push_queue(struct link *l_ptr)
1586 {
1587         u32 res;
1588
1589         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr))
1590                 return;
1591
1592         do {
1593                 res = tipc_link_push_packet(l_ptr);
1594         } while (!res);
1595
1596         if (res == PUSH_FAILED)
1597                 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1598 }
1599
1600 static void link_reset_all(unsigned long addr)
1601 {
1602         struct tipc_node *n_ptr;
1603         char addr_string[16];
1604         u32 i;
1605
1606         read_lock_bh(&tipc_net_lock);
1607         n_ptr = tipc_node_find((u32)addr);
1608         if (!n_ptr) {
1609                 read_unlock_bh(&tipc_net_lock);
1610                 return; /* node no longer exists */
1611         }
1612
1613         tipc_node_lock(n_ptr);
1614
1615         warn("Resetting all links to %s\n",
1616              tipc_addr_string_fill(addr_string, n_ptr->addr));
1617
1618         for (i = 0; i < MAX_BEARERS; i++) {
1619                 if (n_ptr->links[i]) {
1620                         link_print(n_ptr->links[i], TIPC_OUTPUT,
1621                                    "Resetting link\n");
1622                         tipc_link_reset(n_ptr->links[i]);
1623                 }
1624         }
1625
1626         tipc_node_unlock(n_ptr);
1627         read_unlock_bh(&tipc_net_lock);
1628 }
1629
1630 static void link_retransmit_failure(struct link *l_ptr, struct sk_buff *buf)
1631 {
1632         struct tipc_msg *msg = buf_msg(buf);
1633
1634         warn("Retransmission failure on link <%s>\n", l_ptr->name);
1635         tipc_msg_dbg(TIPC_OUTPUT, msg, ">RETR-FAIL>");
1636
1637         if (l_ptr->addr) {
1638
1639                 /* Handle failure on standard link */
1640
1641                 link_print(l_ptr, TIPC_OUTPUT, "Resetting link\n");
1642                 tipc_link_reset(l_ptr);
1643
1644         } else {
1645
1646                 /* Handle failure on broadcast link */
1647
1648                 struct tipc_node *n_ptr;
1649                 char addr_string[16];
1650
1651                 tipc_printf(TIPC_OUTPUT, "Msg seq number: %u,  ", msg_seqno(msg));
1652                 tipc_printf(TIPC_OUTPUT, "Outstanding acks: %lu\n",
1653                                      (unsigned long) TIPC_SKB_CB(buf)->handle);
1654
1655                 n_ptr = l_ptr->owner->next;
1656                 tipc_node_lock(n_ptr);
1657
1658                 tipc_addr_string_fill(addr_string, n_ptr->addr);
1659                 tipc_printf(TIPC_OUTPUT, "Multicast link info for %s\n", addr_string);
1660                 tipc_printf(TIPC_OUTPUT, "Supported: %d,  ", n_ptr->bclink.supported);
1661                 tipc_printf(TIPC_OUTPUT, "Acked: %u\n", n_ptr->bclink.acked);
1662                 tipc_printf(TIPC_OUTPUT, "Last in: %u,  ", n_ptr->bclink.last_in);
1663                 tipc_printf(TIPC_OUTPUT, "Gap after: %u,  ", n_ptr->bclink.gap_after);
1664                 tipc_printf(TIPC_OUTPUT, "Gap to: %u\n", n_ptr->bclink.gap_to);
1665                 tipc_printf(TIPC_OUTPUT, "Nack sync: %u\n\n", n_ptr->bclink.nack_sync);
1666
1667                 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1668
1669                 tipc_node_unlock(n_ptr);
1670
1671                 l_ptr->stale_count = 0;
1672         }
1673 }
1674
1675 void tipc_link_retransmit(struct link *l_ptr, struct sk_buff *buf,
1676                           u32 retransmits)
1677 {
1678         struct tipc_msg *msg;
1679
1680         if (!buf)
1681                 return;
1682
1683         msg = buf_msg(buf);
1684
1685         dbg("Retransmitting %u in link %x\n", retransmits, l_ptr);
1686
1687         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1688                 if (l_ptr->retransm_queue_size == 0) {
1689                         msg_dbg(msg, ">NO_RETR->BCONG>");
1690                         dbg_print_link(l_ptr, "   ");
1691                         l_ptr->retransm_queue_head = msg_seqno(msg);
1692                         l_ptr->retransm_queue_size = retransmits;
1693                 } else {
1694                         err("Unexpected retransmit on link %s (qsize=%d)\n",
1695                             l_ptr->name, l_ptr->retransm_queue_size);
1696                 }
1697                 return;
1698         } else {
1699                 /* Detect repeated retransmit failures on uncongested bearer */
1700
1701                 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1702                         if (++l_ptr->stale_count > 100) {
1703                                 link_retransmit_failure(l_ptr, buf);
1704                                 return;
1705                         }
1706                 } else {
1707                         l_ptr->last_retransmitted = msg_seqno(msg);
1708                         l_ptr->stale_count = 1;
1709                 }
1710         }
1711
1712         while (retransmits && (buf != l_ptr->next_out) && buf) {
1713                 msg = buf_msg(buf);
1714                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1715                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1716                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1717                         msg_dbg(buf_msg(buf), ">RETR>");
1718                         buf = buf->next;
1719                         retransmits--;
1720                         l_ptr->stats.retransmitted++;
1721                 } else {
1722                         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1723                         l_ptr->stats.bearer_congs++;
1724                         l_ptr->retransm_queue_head = msg_seqno(buf_msg(buf));
1725                         l_ptr->retransm_queue_size = retransmits;
1726                         return;
1727                 }
1728         }
1729
1730         l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1731 }
1732
1733 /**
1734  * link_insert_deferred_queue - insert deferred messages back into receive chain
1735  */
1736
1737 static struct sk_buff *link_insert_deferred_queue(struct link *l_ptr,
1738                                                   struct sk_buff *buf)
1739 {
1740         u32 seq_no;
1741
1742         if (l_ptr->oldest_deferred_in == NULL)
1743                 return buf;
1744
1745         seq_no = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1746         if (seq_no == mod(l_ptr->next_in_no)) {
1747                 l_ptr->newest_deferred_in->next = buf;
1748                 buf = l_ptr->oldest_deferred_in;
1749                 l_ptr->oldest_deferred_in = NULL;
1750                 l_ptr->deferred_inqueue_sz = 0;
1751         }
1752         return buf;
1753 }
1754
1755 /**
1756  * link_recv_buf_validate - validate basic format of received message
1757  *
1758  * This routine ensures a TIPC message has an acceptable header, and at least
1759  * as much data as the header indicates it should.  The routine also ensures
1760  * that the entire message header is stored in the main fragment of the message
1761  * buffer, to simplify future access to message header fields.
1762  *
1763  * Note: Having extra info present in the message header or data areas is OK.
1764  * TIPC will ignore the excess, under the assumption that it is optional info
1765  * introduced by a later release of the protocol.
1766  */
1767
1768 static int link_recv_buf_validate(struct sk_buff *buf)
1769 {
1770         static u32 min_data_hdr_size[8] = {
1771                 SHORT_H_SIZE, MCAST_H_SIZE, LONG_H_SIZE, DIR_MSG_H_SIZE,
1772                 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1773                 };
1774
1775         struct tipc_msg *msg;
1776         u32 tipc_hdr[2];
1777         u32 size;
1778         u32 hdr_size;
1779         u32 min_hdr_size;
1780
1781         if (unlikely(buf->len < MIN_H_SIZE))
1782                 return 0;
1783
1784         msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1785         if (msg == NULL)
1786                 return 0;
1787
1788         if (unlikely(msg_version(msg) != TIPC_VERSION))
1789                 return 0;
1790
1791         size = msg_size(msg);
1792         hdr_size = msg_hdr_sz(msg);
1793         min_hdr_size = msg_isdata(msg) ?
1794                 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1795
1796         if (unlikely((hdr_size < min_hdr_size) ||
1797                      (size < hdr_size) ||
1798                      (buf->len < size) ||
1799                      (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1800                 return 0;
1801
1802         return pskb_may_pull(buf, hdr_size);
1803 }
1804
1805 /**
1806  * tipc_recv_msg - process TIPC messages arriving from off-node
1807  * @head: pointer to message buffer chain
1808  * @tb_ptr: pointer to bearer message arrived on
1809  *
1810  * Invoked with no locks held.  Bearer pointer must point to a valid bearer
1811  * structure (i.e. cannot be NULL), but bearer can be inactive.
1812  */
1813
1814 void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *tb_ptr)
1815 {
1816         read_lock_bh(&tipc_net_lock);
1817         while (head) {
1818                 struct bearer *b_ptr = (struct bearer *)tb_ptr;
1819                 struct tipc_node *n_ptr;
1820                 struct link *l_ptr;
1821                 struct sk_buff *crs;
1822                 struct sk_buff *buf = head;
1823                 struct tipc_msg *msg;
1824                 u32 seq_no;
1825                 u32 ackd;
1826                 u32 released = 0;
1827                 int type;
1828
1829                 head = head->next;
1830
1831                 /* Ensure bearer is still enabled */
1832
1833                 if (unlikely(!b_ptr->active))
1834                         goto cont;
1835
1836                 /* Ensure message is well-formed */
1837
1838                 if (unlikely(!link_recv_buf_validate(buf)))
1839                         goto cont;
1840
1841                 /* Ensure message data is a single contiguous unit */
1842
1843                 if (unlikely(buf_linearize(buf))) {
1844                         goto cont;
1845                 }
1846
1847                 /* Handle arrival of a non-unicast link message */
1848
1849                 msg = buf_msg(buf);
1850
1851                 if (unlikely(msg_non_seq(msg))) {
1852                         if (msg_user(msg) ==  LINK_CONFIG)
1853                                 tipc_disc_recv_msg(buf, b_ptr);
1854                         else
1855                                 tipc_bclink_recv_pkt(buf);
1856                         continue;
1857                 }
1858
1859                 if (unlikely(!msg_short(msg) &&
1860                              (msg_destnode(msg) != tipc_own_addr)))
1861                         goto cont;
1862
1863                 /* Discard non-routeable messages destined for another node */
1864
1865                 if (unlikely(!msg_isdata(msg) &&
1866                              (msg_destnode(msg) != tipc_own_addr))) {
1867                         if ((msg_user(msg) != CONN_MANAGER) &&
1868                             (msg_user(msg) != MSG_FRAGMENTER))
1869                                 goto cont;
1870                 }
1871
1872                 /* Locate unicast link endpoint that should handle message */
1873
1874                 n_ptr = tipc_node_find(msg_prevnode(msg));
1875                 if (unlikely(!n_ptr))
1876                         goto cont;
1877                 tipc_node_lock(n_ptr);
1878
1879                 l_ptr = n_ptr->links[b_ptr->identity];
1880                 if (unlikely(!l_ptr)) {
1881                         tipc_node_unlock(n_ptr);
1882                         goto cont;
1883                 }
1884
1885                 /* Validate message sequence number info */
1886
1887                 seq_no = msg_seqno(msg);
1888                 ackd = msg_ack(msg);
1889
1890                 /* Release acked messages */
1891
1892                 if (less(n_ptr->bclink.acked, msg_bcast_ack(msg))) {
1893                         if (tipc_node_is_up(n_ptr) && n_ptr->bclink.supported)
1894                                 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1895                 }
1896
1897                 crs = l_ptr->first_out;
1898                 while ((crs != l_ptr->next_out) &&
1899                        less_eq(msg_seqno(buf_msg(crs)), ackd)) {
1900                         struct sk_buff *next = crs->next;
1901
1902                         buf_discard(crs);
1903                         crs = next;
1904                         released++;
1905                 }
1906                 if (released) {
1907                         l_ptr->first_out = crs;
1908                         l_ptr->out_queue_size -= released;
1909                 }
1910
1911                 /* Try sending any messages link endpoint has pending */
1912
1913                 if (unlikely(l_ptr->next_out))
1914                         tipc_link_push_queue(l_ptr);
1915                 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1916                         tipc_link_wakeup_ports(l_ptr, 0);
1917                 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1918                         l_ptr->stats.sent_acks++;
1919                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1920                 }
1921
1922                 /* Now (finally!) process the incoming message */
1923
1924 protocol_check:
1925                 if (likely(link_working_working(l_ptr))) {
1926                         if (likely(seq_no == mod(l_ptr->next_in_no))) {
1927                                 l_ptr->next_in_no++;
1928                                 if (unlikely(l_ptr->oldest_deferred_in))
1929                                         head = link_insert_deferred_queue(l_ptr,
1930                                                                           head);
1931                                 if (likely(msg_is_dest(msg, tipc_own_addr))) {
1932 deliver:
1933                                         if (likely(msg_isdata(msg))) {
1934                                                 tipc_node_unlock(n_ptr);
1935                                                 tipc_port_recv_msg(buf);
1936                                                 continue;
1937                                         }
1938                                         switch (msg_user(msg)) {
1939                                         case MSG_BUNDLER:
1940                                                 l_ptr->stats.recv_bundles++;
1941                                                 l_ptr->stats.recv_bundled +=
1942                                                         msg_msgcnt(msg);
1943                                                 tipc_node_unlock(n_ptr);
1944                                                 tipc_link_recv_bundle(buf);
1945                                                 continue;
1946                                         case ROUTE_DISTRIBUTOR:
1947                                                 tipc_node_unlock(n_ptr);
1948                                                 tipc_cltr_recv_routing_table(buf);
1949                                                 continue;
1950                                         case NAME_DISTRIBUTOR:
1951                                                 tipc_node_unlock(n_ptr);
1952                                                 tipc_named_recv(buf);
1953                                                 continue;
1954                                         case CONN_MANAGER:
1955                                                 tipc_node_unlock(n_ptr);
1956                                                 tipc_port_recv_proto_msg(buf);
1957                                                 continue;
1958                                         case MSG_FRAGMENTER:
1959                                                 l_ptr->stats.recv_fragments++;
1960                                                 if (tipc_link_recv_fragment(&l_ptr->defragm_buf,
1961                                                                             &buf, &msg)) {
1962                                                         l_ptr->stats.recv_fragmented++;
1963                                                         goto deliver;
1964                                                 }
1965                                                 break;
1966                                         case CHANGEOVER_PROTOCOL:
1967                                                 type = msg_type(msg);
1968                                                 if (link_recv_changeover_msg(&l_ptr, &buf)) {
1969                                                         msg = buf_msg(buf);
1970                                                         seq_no = msg_seqno(msg);
1971                                                         if (type == ORIGINAL_MSG)
1972                                                                 goto deliver;
1973                                                         goto protocol_check;
1974                                                 }
1975                                                 break;
1976                                         }
1977                                 }
1978                                 tipc_node_unlock(n_ptr);
1979                                 tipc_net_route_msg(buf);
1980                                 continue;
1981                         }
1982                         link_handle_out_of_seq_msg(l_ptr, buf);
1983                         head = link_insert_deferred_queue(l_ptr, head);
1984                         tipc_node_unlock(n_ptr);
1985                         continue;
1986                 }
1987
1988                 if (msg_user(msg) == LINK_PROTOCOL) {
1989                         link_recv_proto_msg(l_ptr, buf);
1990                         head = link_insert_deferred_queue(l_ptr, head);
1991                         tipc_node_unlock(n_ptr);
1992                         continue;
1993                 }
1994                 msg_dbg(msg,"NSEQ<REC<");
1995                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1996
1997                 if (link_working_working(l_ptr)) {
1998                         /* Re-insert in front of queue */
1999                         msg_dbg(msg,"RECV-REINS:");
2000                         buf->next = head;
2001                         head = buf;
2002                         tipc_node_unlock(n_ptr);
2003                         continue;
2004                 }
2005                 tipc_node_unlock(n_ptr);
2006 cont:
2007                 buf_discard(buf);
2008         }
2009         read_unlock_bh(&tipc_net_lock);
2010 }
2011
2012 /*
2013  * link_defer_buf(): Sort a received out-of-sequence packet
2014  *                   into the deferred reception queue.
2015  * Returns the increase of the queue length,i.e. 0 or 1
2016  */
2017
2018 u32 tipc_link_defer_pkt(struct sk_buff **head,
2019                         struct sk_buff **tail,
2020                         struct sk_buff *buf)
2021 {
2022         struct sk_buff *prev = NULL;
2023         struct sk_buff *crs = *head;
2024         u32 seq_no = msg_seqno(buf_msg(buf));
2025
2026         buf->next = NULL;
2027
2028         /* Empty queue ? */
2029         if (*head == NULL) {
2030                 *head = *tail = buf;
2031                 return 1;
2032         }
2033
2034         /* Last ? */
2035         if (less(msg_seqno(buf_msg(*tail)), seq_no)) {
2036                 (*tail)->next = buf;
2037                 *tail = buf;
2038                 return 1;
2039         }
2040
2041         /* Scan through queue and sort it in */
2042         do {
2043                 struct tipc_msg *msg = buf_msg(crs);
2044
2045                 if (less(seq_no, msg_seqno(msg))) {
2046                         buf->next = crs;
2047                         if (prev)
2048                                 prev->next = buf;
2049                         else
2050                                 *head = buf;
2051                         return 1;
2052                 }
2053                 if (seq_no == msg_seqno(msg)) {
2054                         break;
2055                 }
2056                 prev = crs;
2057                 crs = crs->next;
2058         }
2059         while (crs);
2060
2061         /* Message is a duplicate of an existing message */
2062
2063         buf_discard(buf);
2064         return 0;
2065 }
2066
2067 /**
2068  * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
2069  */
2070
2071 static void link_handle_out_of_seq_msg(struct link *l_ptr,
2072                                        struct sk_buff *buf)
2073 {
2074         u32 seq_no = msg_seqno(buf_msg(buf));
2075
2076         if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
2077                 link_recv_proto_msg(l_ptr, buf);
2078                 return;
2079         }
2080
2081         dbg("rx OOS msg: seq_no %u, expecting %u (%u)\n",
2082             seq_no, mod(l_ptr->next_in_no), l_ptr->next_in_no);
2083
2084         /* Record OOS packet arrival (force mismatch on next timeout) */
2085
2086         l_ptr->checkpoint--;
2087
2088         /*
2089          * Discard packet if a duplicate; otherwise add it to deferred queue
2090          * and notify peer of gap as per protocol specification
2091          */
2092
2093         if (less(seq_no, mod(l_ptr->next_in_no))) {
2094                 l_ptr->stats.duplicates++;
2095                 buf_discard(buf);
2096                 return;
2097         }
2098
2099         if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
2100                                 &l_ptr->newest_deferred_in, buf)) {
2101                 l_ptr->deferred_inqueue_sz++;
2102                 l_ptr->stats.deferred_recv++;
2103                 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
2104                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
2105         } else
2106                 l_ptr->stats.duplicates++;
2107 }
2108
2109 /*
2110  * Send protocol message to the other endpoint.
2111  */
2112 void tipc_link_send_proto_msg(struct link *l_ptr, u32 msg_typ, int probe_msg,
2113                               u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
2114 {
2115         struct sk_buff *buf = NULL;
2116         struct tipc_msg *msg = l_ptr->pmsg;
2117         u32 msg_size = sizeof(l_ptr->proto_msg);
2118
2119         if (link_blocked(l_ptr))
2120                 return;
2121         msg_set_type(msg, msg_typ);
2122         msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
2123         msg_set_bcast_ack(msg, mod(l_ptr->owner->bclink.last_in));
2124         msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
2125
2126         if (msg_typ == STATE_MSG) {
2127                 u32 next_sent = mod(l_ptr->next_out_no);
2128
2129                 if (!tipc_link_is_up(l_ptr))
2130                         return;
2131                 if (l_ptr->next_out)
2132                         next_sent = msg_seqno(buf_msg(l_ptr->next_out));
2133                 msg_set_next_sent(msg, next_sent);
2134                 if (l_ptr->oldest_deferred_in) {
2135                         u32 rec = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
2136                         gap = mod(rec - mod(l_ptr->next_in_no));
2137                 }
2138                 msg_set_seq_gap(msg, gap);
2139                 if (gap)
2140                         l_ptr->stats.sent_nacks++;
2141                 msg_set_link_tolerance(msg, tolerance);
2142                 msg_set_linkprio(msg, priority);
2143                 msg_set_max_pkt(msg, ack_mtu);
2144                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
2145                 msg_set_probe(msg, probe_msg != 0);
2146                 if (probe_msg) {
2147                         u32 mtu = l_ptr->max_pkt;
2148
2149                         if ((mtu < l_ptr->max_pkt_target) &&
2150                             link_working_working(l_ptr) &&
2151                             l_ptr->fsm_msg_cnt) {
2152                                 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2153                                 if (l_ptr->max_pkt_probes == 10) {
2154                                         l_ptr->max_pkt_target = (msg_size - 4);
2155                                         l_ptr->max_pkt_probes = 0;
2156                                         msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2157                                 }
2158                                 l_ptr->max_pkt_probes++;
2159                         }
2160
2161                         l_ptr->stats.sent_probes++;
2162                 }
2163                 l_ptr->stats.sent_states++;
2164         } else {                /* RESET_MSG or ACTIVATE_MSG */
2165                 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
2166                 msg_set_seq_gap(msg, 0);
2167                 msg_set_next_sent(msg, 1);
2168                 msg_set_link_tolerance(msg, l_ptr->tolerance);
2169                 msg_set_linkprio(msg, l_ptr->priority);
2170                 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
2171         }
2172
2173         if (tipc_node_has_redundant_links(l_ptr->owner)) {
2174                 msg_set_redundant_link(msg);
2175         } else {
2176                 msg_clear_redundant_link(msg);
2177         }
2178         msg_set_linkprio(msg, l_ptr->priority);
2179
2180         /* Ensure sequence number will not fit : */
2181
2182         msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
2183
2184         /* Congestion? */
2185
2186         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
2187                 if (!l_ptr->proto_msg_queue) {
2188                         l_ptr->proto_msg_queue =
2189                                 buf_acquire(sizeof(l_ptr->proto_msg));
2190                 }
2191                 buf = l_ptr->proto_msg_queue;
2192                 if (!buf)
2193                         return;
2194                 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2195                 return;
2196         }
2197         msg_set_timestamp(msg, jiffies_to_msecs(jiffies));
2198
2199         /* Message can be sent */
2200
2201         msg_dbg(msg, ">>");
2202
2203         buf = buf_acquire(msg_size);
2204         if (!buf)
2205                 return;
2206
2207         skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2208         msg_set_size(buf_msg(buf), msg_size);
2209
2210         if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
2211                 l_ptr->unacked_window = 0;
2212                 buf_discard(buf);
2213                 return;
2214         }
2215
2216         /* New congestion */
2217         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
2218         l_ptr->proto_msg_queue = buf;
2219         l_ptr->stats.bearer_congs++;
2220 }
2221
2222 /*
2223  * Receive protocol message :
2224  * Note that network plane id propagates through the network, and may
2225  * change at any time. The node with lowest address rules
2226  */
2227
2228 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf)
2229 {
2230         u32 rec_gap = 0;
2231         u32 max_pkt_info;
2232         u32 max_pkt_ack;
2233         u32 msg_tol;
2234         struct tipc_msg *msg = buf_msg(buf);
2235
2236         dbg("AT(%u):", jiffies_to_msecs(jiffies));
2237         msg_dbg(msg, "<<");
2238         if (link_blocked(l_ptr))
2239                 goto exit;
2240
2241         /* record unnumbered packet arrival (force mismatch on next timeout) */
2242
2243         l_ptr->checkpoint--;
2244
2245         if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
2246                 if (tipc_own_addr > msg_prevnode(msg))
2247                         l_ptr->b_ptr->net_plane = msg_net_plane(msg);
2248
2249         l_ptr->owner->permit_changeover = msg_redundant_link(msg);
2250
2251         switch (msg_type(msg)) {
2252
2253         case RESET_MSG:
2254                 if (!link_working_unknown(l_ptr) &&
2255                     (l_ptr->peer_session != INVALID_SESSION)) {
2256                         if (msg_session(msg) == l_ptr->peer_session) {
2257                                 dbg("Duplicate RESET: %u<->%u\n",
2258                                     msg_session(msg), l_ptr->peer_session);
2259                                 break; /* duplicate: ignore */
2260                         }
2261                 }
2262                 /* fall thru' */
2263         case ACTIVATE_MSG:
2264                 /* Update link settings according other endpoint's values */
2265
2266                 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2267
2268                 if ((msg_tol = msg_link_tolerance(msg)) &&
2269                     (msg_tol > l_ptr->tolerance))
2270                         link_set_supervision_props(l_ptr, msg_tol);
2271
2272                 if (msg_linkprio(msg) > l_ptr->priority)
2273                         l_ptr->priority = msg_linkprio(msg);
2274
2275                 max_pkt_info = msg_max_pkt(msg);
2276                 if (max_pkt_info) {
2277                         if (max_pkt_info < l_ptr->max_pkt_target)
2278                                 l_ptr->max_pkt_target = max_pkt_info;
2279                         if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2280                                 l_ptr->max_pkt = l_ptr->max_pkt_target;
2281                 } else {
2282                         l_ptr->max_pkt = l_ptr->max_pkt_target;
2283                 }
2284                 l_ptr->owner->bclink.supported = (max_pkt_info != 0);
2285
2286                 link_state_event(l_ptr, msg_type(msg));
2287
2288                 l_ptr->peer_session = msg_session(msg);
2289                 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2290
2291                 /* Synchronize broadcast sequence numbers */
2292                 if (!tipc_node_has_redundant_links(l_ptr->owner)) {
2293                         l_ptr->owner->bclink.last_in = mod(msg_last_bcast(msg));
2294                 }
2295                 break;
2296         case STATE_MSG:
2297
2298                 if ((msg_tol = msg_link_tolerance(msg)))
2299                         link_set_supervision_props(l_ptr, msg_tol);
2300
2301                 if (msg_linkprio(msg) &&
2302                     (msg_linkprio(msg) != l_ptr->priority)) {
2303                         warn("Resetting link <%s>, priority change %u->%u\n",
2304                              l_ptr->name, l_ptr->priority, msg_linkprio(msg));
2305                         l_ptr->priority = msg_linkprio(msg);
2306                         tipc_link_reset(l_ptr); /* Enforce change to take effect */
2307                         break;
2308                 }
2309                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2310                 l_ptr->stats.recv_states++;
2311                 if (link_reset_unknown(l_ptr))
2312                         break;
2313
2314                 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2315                         rec_gap = mod(msg_next_sent(msg) -
2316                                       mod(l_ptr->next_in_no));
2317                 }
2318
2319                 max_pkt_ack = msg_max_pkt(msg);
2320                 if (max_pkt_ack > l_ptr->max_pkt) {
2321                         dbg("Link <%s> updated MTU %u -> %u\n",
2322                             l_ptr->name, l_ptr->max_pkt, max_pkt_ack);
2323                         l_ptr->max_pkt = max_pkt_ack;
2324                         l_ptr->max_pkt_probes = 0;
2325                 }
2326
2327                 max_pkt_ack = 0;
2328                 if (msg_probe(msg)) {
2329                         l_ptr->stats.recv_probes++;
2330                         if (msg_size(msg) > sizeof(l_ptr->proto_msg)) {
2331                                 max_pkt_ack = msg_size(msg);
2332                         }
2333                 }
2334
2335                 /* Protocol message before retransmits, reduce loss risk */
2336
2337                 tipc_bclink_check_gap(l_ptr->owner, msg_last_bcast(msg));
2338
2339                 if (rec_gap || (msg_probe(msg))) {
2340                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2341                                                  0, rec_gap, 0, 0, max_pkt_ack);
2342                 }
2343                 if (msg_seq_gap(msg)) {
2344                         msg_dbg(msg, "With Gap:");
2345                         l_ptr->stats.recv_nacks++;
2346                         tipc_link_retransmit(l_ptr, l_ptr->first_out,
2347                                              msg_seq_gap(msg));
2348                 }
2349                 break;
2350         default:
2351                 msg_dbg(buf_msg(buf), "<DISCARDING UNKNOWN<");
2352         }
2353 exit:
2354         buf_discard(buf);
2355 }
2356
2357
2358 /*
2359  * tipc_link_tunnel(): Send one message via a link belonging to
2360  * another bearer. Owner node is locked.
2361  */
2362 void tipc_link_tunnel(struct link *l_ptr,
2363                       struct tipc_msg *tunnel_hdr,
2364                       struct tipc_msg  *msg,
2365                       u32 selector)
2366 {
2367         struct link *tunnel;
2368         struct sk_buff *buf;
2369         u32 length = msg_size(msg);
2370
2371         tunnel = l_ptr->owner->active_links[selector & 1];
2372         if (!tipc_link_is_up(tunnel)) {
2373                 warn("Link changeover error, "
2374                      "tunnel link no longer available\n");
2375                 return;
2376         }
2377         msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2378         buf = buf_acquire(length + INT_H_SIZE);
2379         if (!buf) {
2380                 warn("Link changeover error, "
2381                      "unable to send tunnel msg\n");
2382                 return;
2383         }
2384         skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
2385         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
2386         dbg("%c->%c:", l_ptr->b_ptr->net_plane, tunnel->b_ptr->net_plane);
2387         msg_dbg(buf_msg(buf), ">SEND>");
2388         tipc_link_send_buf(tunnel, buf);
2389 }
2390
2391
2392
2393 /*
2394  * changeover(): Send whole message queue via the remaining link
2395  *               Owner node is locked.
2396  */
2397
2398 void tipc_link_changeover(struct link *l_ptr)
2399 {
2400         u32 msgcount = l_ptr->out_queue_size;
2401         struct sk_buff *crs = l_ptr->first_out;
2402         struct link *tunnel = l_ptr->owner->active_links[0];
2403         struct tipc_msg tunnel_hdr;
2404         int split_bundles;
2405
2406         if (!tunnel)
2407                 return;
2408
2409         if (!l_ptr->owner->permit_changeover) {
2410                 warn("Link changeover error, "
2411                      "peer did not permit changeover\n");
2412                 return;
2413         }
2414
2415         tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2416                  ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
2417         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2418         msg_set_msgcnt(&tunnel_hdr, msgcount);
2419         dbg("Link changeover requires %u tunnel messages\n", msgcount);
2420
2421         if (!l_ptr->first_out) {
2422                 struct sk_buff *buf;
2423
2424                 buf = buf_acquire(INT_H_SIZE);
2425                 if (buf) {
2426                         skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2427                         msg_set_size(&tunnel_hdr, INT_H_SIZE);
2428                         dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2429                             tunnel->b_ptr->net_plane);
2430                         msg_dbg(&tunnel_hdr, "EMPTY>SEND>");
2431                         tipc_link_send_buf(tunnel, buf);
2432                 } else {
2433                         warn("Link changeover error, "
2434                              "unable to send changeover msg\n");
2435                 }
2436                 return;
2437         }
2438
2439         split_bundles = (l_ptr->owner->active_links[0] !=
2440                          l_ptr->owner->active_links[1]);
2441
2442         while (crs) {
2443                 struct tipc_msg *msg = buf_msg(crs);
2444
2445                 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2446                         struct tipc_msg *m = msg_get_wrapped(msg);
2447                         unchar* pos = (unchar*)m;
2448
2449                         msgcount = msg_msgcnt(msg);
2450                         while (msgcount--) {
2451                                 msg_set_seqno(m,msg_seqno(msg));
2452                                 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2453                                                  msg_link_selector(m));
2454                                 pos += align(msg_size(m));
2455                                 m = (struct tipc_msg *)pos;
2456                         }
2457                 } else {
2458                         tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2459                                          msg_link_selector(msg));
2460                 }
2461                 crs = crs->next;
2462         }
2463 }
2464
2465 void tipc_link_send_duplicate(struct link *l_ptr, struct link *tunnel)
2466 {
2467         struct sk_buff *iter;
2468         struct tipc_msg tunnel_hdr;
2469
2470         tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2471                  DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
2472         msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2473         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2474         iter = l_ptr->first_out;
2475         while (iter) {
2476                 struct sk_buff *outbuf;
2477                 struct tipc_msg *msg = buf_msg(iter);
2478                 u32 length = msg_size(msg);
2479
2480                 if (msg_user(msg) == MSG_BUNDLER)
2481                         msg_set_type(msg, CLOSED_MSG);
2482                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));   /* Update */
2483                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2484                 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2485                 outbuf = buf_acquire(length + INT_H_SIZE);
2486                 if (outbuf == NULL) {
2487                         warn("Link changeover error, "
2488                              "unable to send duplicate msg\n");
2489                         return;
2490                 }
2491                 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2492                 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2493                                                length);
2494                 dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2495                     tunnel->b_ptr->net_plane);
2496                 msg_dbg(buf_msg(outbuf), ">SEND>");
2497                 tipc_link_send_buf(tunnel, outbuf);
2498                 if (!tipc_link_is_up(l_ptr))
2499                         return;
2500                 iter = iter->next;
2501         }
2502 }
2503
2504
2505
2506 /**
2507  * buf_extract - extracts embedded TIPC message from another message
2508  * @skb: encapsulating message buffer
2509  * @from_pos: offset to extract from
2510  *
2511  * Returns a new message buffer containing an embedded message.  The
2512  * encapsulating message itself is left unchanged.
2513  */
2514
2515 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2516 {
2517         struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2518         u32 size = msg_size(msg);
2519         struct sk_buff *eb;
2520
2521         eb = buf_acquire(size);
2522         if (eb)
2523                 skb_copy_to_linear_data(eb, msg, size);
2524         return eb;
2525 }
2526
2527 /*
2528  *  link_recv_changeover_msg(): Receive tunneled packet sent
2529  *  via other link. Node is locked. Return extracted buffer.
2530  */
2531
2532 static int link_recv_changeover_msg(struct link **l_ptr,
2533                                     struct sk_buff **buf)
2534 {
2535         struct sk_buff *tunnel_buf = *buf;
2536         struct link *dest_link;
2537         struct tipc_msg *msg;
2538         struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2539         u32 msg_typ = msg_type(tunnel_msg);
2540         u32 msg_count = msg_msgcnt(tunnel_msg);
2541
2542         dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2543         if (!dest_link) {
2544                 msg_dbg(tunnel_msg, "NOLINK/<REC<");
2545                 goto exit;
2546         }
2547         if (dest_link == *l_ptr) {
2548                 err("Unexpected changeover message on link <%s>\n",
2549                     (*l_ptr)->name);
2550                 goto exit;
2551         }
2552         dbg("%c<-%c:", dest_link->b_ptr->net_plane,
2553             (*l_ptr)->b_ptr->net_plane);
2554         *l_ptr = dest_link;
2555         msg = msg_get_wrapped(tunnel_msg);
2556
2557         if (msg_typ == DUPLICATE_MSG) {
2558                 if (less(msg_seqno(msg), mod(dest_link->next_in_no))) {
2559                         msg_dbg(tunnel_msg, "DROP/<REC<");
2560                         goto exit;
2561                 }
2562                 *buf = buf_extract(tunnel_buf,INT_H_SIZE);
2563                 if (*buf == NULL) {
2564                         warn("Link changeover error, duplicate msg dropped\n");
2565                         goto exit;
2566                 }
2567                 msg_dbg(tunnel_msg, "TNL<REC<");
2568                 buf_discard(tunnel_buf);
2569                 return 1;
2570         }
2571
2572         /* First original message ?: */
2573
2574         if (tipc_link_is_up(dest_link)) {
2575                 msg_dbg(tunnel_msg, "UP/FIRST/<REC<");
2576                 info("Resetting link <%s>, changeover initiated by peer\n",
2577                      dest_link->name);
2578                 tipc_link_reset(dest_link);
2579                 dest_link->exp_msg_count = msg_count;
2580                 dbg("Expecting %u tunnelled messages\n", msg_count);
2581                 if (!msg_count)
2582                         goto exit;
2583         } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2584                 msg_dbg(tunnel_msg, "BLK/FIRST/<REC<");
2585                 dest_link->exp_msg_count = msg_count;
2586                 dbg("Expecting %u tunnelled messages\n", msg_count);
2587                 if (!msg_count)
2588                         goto exit;
2589         }
2590
2591         /* Receive original message */
2592
2593         if (dest_link->exp_msg_count == 0) {
2594                 warn("Link switchover error, "
2595                      "got too many tunnelled messages\n");
2596                 msg_dbg(tunnel_msg, "OVERDUE/DROP/<REC<");
2597                 dbg_print_link(dest_link, "LINK:");
2598                 goto exit;
2599         }
2600         dest_link->exp_msg_count--;
2601         if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2602                 msg_dbg(tunnel_msg, "DROP/DUPL/<REC<");
2603                 goto exit;
2604         } else {
2605                 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2606                 if (*buf != NULL) {
2607                         msg_dbg(tunnel_msg, "TNL<REC<");
2608                         buf_discard(tunnel_buf);
2609                         return 1;
2610                 } else {
2611                         warn("Link changeover error, original msg dropped\n");
2612                 }
2613         }
2614 exit:
2615         *buf = NULL;
2616         buf_discard(tunnel_buf);
2617         return 0;
2618 }
2619
2620 /*
2621  *  Bundler functionality:
2622  */
2623 void tipc_link_recv_bundle(struct sk_buff *buf)
2624 {
2625         u32 msgcount = msg_msgcnt(buf_msg(buf));
2626         u32 pos = INT_H_SIZE;
2627         struct sk_buff *obuf;
2628
2629         msg_dbg(buf_msg(buf), "<BNDL<: ");
2630         while (msgcount--) {
2631                 obuf = buf_extract(buf, pos);
2632                 if (obuf == NULL) {
2633                         warn("Link unable to unbundle message(s)\n");
2634                         break;
2635                 }
2636                 pos += align(msg_size(buf_msg(obuf)));
2637                 msg_dbg(buf_msg(obuf), "     /");
2638                 tipc_net_route_msg(obuf);
2639         }
2640         buf_discard(buf);
2641 }
2642
2643 /*
2644  *  Fragmentation/defragmentation:
2645  */
2646
2647
2648 /*
2649  * tipc_link_send_long_buf: Entry for buffers needing fragmentation.
2650  * The buffer is complete, inclusive total message length.
2651  * Returns user data length.
2652  */
2653 int tipc_link_send_long_buf(struct link *l_ptr, struct sk_buff *buf)
2654 {
2655         struct tipc_msg *inmsg = buf_msg(buf);
2656         struct tipc_msg fragm_hdr;
2657         u32 insize = msg_size(inmsg);
2658         u32 dsz = msg_data_sz(inmsg);
2659         unchar *crs = buf->data;
2660         u32 rest = insize;
2661         u32 pack_sz = l_ptr->max_pkt;
2662         u32 fragm_sz = pack_sz - INT_H_SIZE;
2663         u32 fragm_no = 1;
2664         u32 destaddr;
2665
2666         if (msg_short(inmsg))
2667                 destaddr = l_ptr->addr;
2668         else
2669                 destaddr = msg_destnode(inmsg);
2670
2671         if (msg_routed(inmsg))
2672                 msg_set_prevnode(inmsg, tipc_own_addr);
2673
2674         /* Prepare reusable fragment header: */
2675
2676         tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2677                  INT_H_SIZE, destaddr);
2678         msg_set_link_selector(&fragm_hdr, msg_link_selector(inmsg));
2679         msg_set_long_msgno(&fragm_hdr, mod(l_ptr->long_msg_seq_no++));
2680         msg_set_fragm_no(&fragm_hdr, fragm_no);
2681         l_ptr->stats.sent_fragmented++;
2682
2683         /* Chop up message: */
2684
2685         while (rest > 0) {
2686                 struct sk_buff *fragm;
2687
2688                 if (rest <= fragm_sz) {
2689                         fragm_sz = rest;
2690                         msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2691                 }
2692                 fragm = buf_acquire(fragm_sz + INT_H_SIZE);
2693                 if (fragm == NULL) {
2694                         warn("Link unable to fragment message\n");
2695                         dsz = -ENOMEM;
2696                         goto exit;
2697                 }
2698                 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2699                 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2700                 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2701                                                fragm_sz);
2702                 /*  Send queued messages first, if any: */
2703
2704                 l_ptr->stats.sent_fragments++;
2705                 tipc_link_send_buf(l_ptr, fragm);
2706                 if (!tipc_link_is_up(l_ptr))
2707                         return dsz;
2708                 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
2709                 rest -= fragm_sz;
2710                 crs += fragm_sz;
2711                 msg_set_type(&fragm_hdr, FRAGMENT);
2712         }
2713 exit:
2714         buf_discard(buf);
2715         return dsz;
2716 }
2717
2718 /*
2719  * A pending message being re-assembled must store certain values
2720  * to handle subsequent fragments correctly. The following functions
2721  * help storing these values in unused, available fields in the
2722  * pending message. This makes dynamic memory allocation unecessary.
2723  */
2724
2725 static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2726 {
2727         msg_set_seqno(buf_msg(buf), seqno);
2728 }
2729
2730 static u32 get_fragm_size(struct sk_buff *buf)
2731 {
2732         return msg_ack(buf_msg(buf));
2733 }
2734
2735 static void set_fragm_size(struct sk_buff *buf, u32 sz)
2736 {
2737         msg_set_ack(buf_msg(buf), sz);
2738 }
2739
2740 static u32 get_expected_frags(struct sk_buff *buf)
2741 {
2742         return msg_bcast_ack(buf_msg(buf));
2743 }
2744
2745 static void set_expected_frags(struct sk_buff *buf, u32 exp)
2746 {
2747         msg_set_bcast_ack(buf_msg(buf), exp);
2748 }
2749
2750 static u32 get_timer_cnt(struct sk_buff *buf)
2751 {
2752         return msg_reroute_cnt(buf_msg(buf));
2753 }
2754
2755 static void incr_timer_cnt(struct sk_buff *buf)
2756 {
2757         msg_incr_reroute_cnt(buf_msg(buf));
2758 }
2759
2760 /*
2761  * tipc_link_recv_fragment(): Called with node lock on. Returns
2762  * the reassembled buffer if message is complete.
2763  */
2764 int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2765                             struct tipc_msg **m)
2766 {
2767         struct sk_buff *prev = NULL;
2768         struct sk_buff *fbuf = *fb;
2769         struct tipc_msg *fragm = buf_msg(fbuf);
2770         struct sk_buff *pbuf = *pending;
2771         u32 long_msg_seq_no = msg_long_msgno(fragm);
2772
2773         *fb = NULL;
2774         msg_dbg(fragm,"FRG<REC<");
2775
2776         /* Is there an incomplete message waiting for this fragment? */
2777
2778         while (pbuf && ((msg_seqno(buf_msg(pbuf)) != long_msg_seq_no) ||
2779                         (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2780                 prev = pbuf;
2781                 pbuf = pbuf->next;
2782         }
2783
2784         if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2785                 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2786                 u32 msg_sz = msg_size(imsg);
2787                 u32 fragm_sz = msg_data_sz(fragm);
2788                 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2789                 u32 max =  TIPC_MAX_USER_MSG_SIZE + LONG_H_SIZE;
2790                 if (msg_type(imsg) == TIPC_MCAST_MSG)
2791                         max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2792                 if (msg_size(imsg) > max) {
2793                         msg_dbg(fragm,"<REC<Oversized: ");
2794                         buf_discard(fbuf);
2795                         return 0;
2796                 }
2797                 pbuf = buf_acquire(msg_size(imsg));
2798                 if (pbuf != NULL) {
2799                         pbuf->next = *pending;
2800                         *pending = pbuf;
2801                         skb_copy_to_linear_data(pbuf, imsg,
2802                                                 msg_data_sz(fragm));
2803                         /*  Prepare buffer for subsequent fragments. */
2804
2805                         set_long_msg_seqno(pbuf, long_msg_seq_no);
2806                         set_fragm_size(pbuf,fragm_sz);
2807                         set_expected_frags(pbuf,exp_fragm_cnt - 1);
2808                 } else {
2809                         warn("Link unable to reassemble fragmented message\n");
2810                 }
2811                 buf_discard(fbuf);
2812                 return 0;
2813         } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2814                 u32 dsz = msg_data_sz(fragm);
2815                 u32 fsz = get_fragm_size(pbuf);
2816                 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2817                 u32 exp_frags = get_expected_frags(pbuf) - 1;
2818                 skb_copy_to_linear_data_offset(pbuf, crs,
2819                                                msg_data(fragm), dsz);
2820                 buf_discard(fbuf);
2821
2822                 /* Is message complete? */
2823
2824                 if (exp_frags == 0) {
2825                         if (prev)
2826                                 prev->next = pbuf->next;
2827                         else
2828                                 *pending = pbuf->next;
2829                         msg_reset_reroute_cnt(buf_msg(pbuf));
2830                         *fb = pbuf;
2831                         *m = buf_msg(pbuf);
2832                         return 1;
2833                 }
2834                 set_expected_frags(pbuf,exp_frags);
2835                 return 0;
2836         }
2837         dbg(" Discarding orphan fragment %x\n",fbuf);
2838         msg_dbg(fragm,"ORPHAN:");
2839         dbg("Pending long buffers:\n");
2840         dbg_print_buf_chain(*pending);
2841         buf_discard(fbuf);
2842         return 0;
2843 }
2844
2845 /**
2846  * link_check_defragm_bufs - flush stale incoming message fragments
2847  * @l_ptr: pointer to link
2848  */
2849
2850 static void link_check_defragm_bufs(struct link *l_ptr)
2851 {
2852         struct sk_buff *prev = NULL;
2853         struct sk_buff *next = NULL;
2854         struct sk_buff *buf = l_ptr->defragm_buf;
2855
2856         if (!buf)
2857                 return;
2858         if (!link_working_working(l_ptr))
2859                 return;
2860         while (buf) {
2861                 u32 cnt = get_timer_cnt(buf);
2862
2863                 next = buf->next;
2864                 if (cnt < 4) {
2865                         incr_timer_cnt(buf);
2866                         prev = buf;
2867                 } else {
2868                         dbg(" Discarding incomplete long buffer\n");
2869                         msg_dbg(buf_msg(buf), "LONG:");
2870                         dbg_print_link(l_ptr, "curr:");
2871                         dbg("Pending long buffers:\n");
2872                         dbg_print_buf_chain(l_ptr->defragm_buf);
2873                         if (prev)
2874                                 prev->next = buf->next;
2875                         else
2876                                 l_ptr->defragm_buf = buf->next;
2877                         buf_discard(buf);
2878                 }
2879                 buf = next;
2880         }
2881 }
2882
2883
2884
2885 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance)
2886 {
2887         l_ptr->tolerance = tolerance;
2888         l_ptr->continuity_interval =
2889                 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2890         l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2891 }
2892
2893
2894 void tipc_link_set_queue_limits(struct link *l_ptr, u32 window)
2895 {
2896         /* Data messages from this node, inclusive FIRST_FRAGM */
2897         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2898         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2899         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2900         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2901         /* Transiting data messages,inclusive FIRST_FRAGM */
2902         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2903         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2904         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2905         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2906         l_ptr->queue_limit[CONN_MANAGER] = 1200;
2907         l_ptr->queue_limit[ROUTE_DISTRIBUTOR] = 1200;
2908         l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2909         l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2910         /* FRAGMENT and LAST_FRAGMENT packets */
2911         l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2912 }
2913
2914 /**
2915  * link_find_link - locate link by name
2916  * @name - ptr to link name string
2917  * @node - ptr to area to be filled with ptr to associated node
2918  *
2919  * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2920  * this also prevents link deletion.
2921  *
2922  * Returns pointer to link (or 0 if invalid link name).
2923  */
2924
2925 static struct link *link_find_link(const char *name, struct tipc_node **node)
2926 {
2927         struct link_name link_name_parts;
2928         struct bearer *b_ptr;
2929         struct link *l_ptr;
2930
2931         if (!link_name_validate(name, &link_name_parts))
2932                 return NULL;
2933
2934         b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2935         if (!b_ptr)
2936                 return NULL;
2937
2938         *node = tipc_node_find(link_name_parts.addr_peer);
2939         if (!*node)
2940                 return NULL;
2941
2942         l_ptr = (*node)->links[b_ptr->identity];
2943         if (!l_ptr || strcmp(l_ptr->name, name))
2944                 return NULL;
2945
2946         return l_ptr;
2947 }
2948
2949 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2950                                      u16 cmd)
2951 {
2952         struct tipc_link_config *args;
2953         u32 new_value;
2954         struct link *l_ptr;
2955         struct tipc_node *node;
2956         int res;
2957
2958         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2959                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2960
2961         args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2962         new_value = ntohl(args->value);
2963
2964         if (!strcmp(args->name, tipc_bclink_name)) {
2965                 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2966                     (tipc_bclink_set_queue_limits(new_value) == 0))
2967                         return tipc_cfg_reply_none();
2968                 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2969                                                    " (cannot change setting on broadcast link)");
2970         }
2971
2972         read_lock_bh(&tipc_net_lock);
2973         l_ptr = link_find_link(args->name, &node);
2974         if (!l_ptr) {
2975                 read_unlock_bh(&tipc_net_lock);
2976                 return tipc_cfg_reply_error_string("link not found");
2977         }
2978
2979         tipc_node_lock(node);
2980         res = -EINVAL;
2981         switch (cmd) {
2982         case TIPC_CMD_SET_LINK_TOL:
2983                 if ((new_value >= TIPC_MIN_LINK_TOL) &&
2984                     (new_value <= TIPC_MAX_LINK_TOL)) {
2985                         link_set_supervision_props(l_ptr, new_value);
2986                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2987                                                  0, 0, new_value, 0, 0);
2988                         res = 0;
2989                 }
2990                 break;
2991         case TIPC_CMD_SET_LINK_PRI:
2992                 if ((new_value >= TIPC_MIN_LINK_PRI) &&
2993                     (new_value <= TIPC_MAX_LINK_PRI)) {
2994                         l_ptr->priority = new_value;
2995                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2996                                                  0, 0, 0, new_value, 0);
2997                         res = 0;
2998                 }
2999                 break;
3000         case TIPC_CMD_SET_LINK_WINDOW:
3001                 if ((new_value >= TIPC_MIN_LINK_WIN) &&
3002                     (new_value <= TIPC_MAX_LINK_WIN)) {
3003                         tipc_link_set_queue_limits(l_ptr, new_value);
3004                         res = 0;
3005                 }
3006                 break;
3007         }
3008         tipc_node_unlock(node);
3009
3010         read_unlock_bh(&tipc_net_lock);
3011         if (res)
3012                 return tipc_cfg_reply_error_string("cannot change link setting");
3013
3014         return tipc_cfg_reply_none();
3015 }
3016
3017 /**
3018  * link_reset_statistics - reset link statistics
3019  * @l_ptr: pointer to link
3020  */
3021
3022 static void link_reset_statistics(struct link *l_ptr)
3023 {
3024         memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
3025         l_ptr->stats.sent_info = l_ptr->next_out_no;
3026         l_ptr->stats.recv_info = l_ptr->next_in_no;
3027 }
3028
3029 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
3030 {
3031         char *link_name;
3032         struct link *l_ptr;
3033         struct tipc_node *node;
3034
3035         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3036                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3037
3038         link_name = (char *)TLV_DATA(req_tlv_area);
3039         if (!strcmp(link_name, tipc_bclink_name)) {
3040                 if (tipc_bclink_reset_stats())
3041                         return tipc_cfg_reply_error_string("link not found");
3042                 return tipc_cfg_reply_none();
3043         }
3044
3045         read_lock_bh(&tipc_net_lock);
3046         l_ptr = link_find_link(link_name, &node);
3047         if (!l_ptr) {
3048                 read_unlock_bh(&tipc_net_lock);
3049                 return tipc_cfg_reply_error_string("link not found");
3050         }
3051
3052         tipc_node_lock(node);
3053         link_reset_statistics(l_ptr);
3054         tipc_node_unlock(node);
3055         read_unlock_bh(&tipc_net_lock);
3056         return tipc_cfg_reply_none();
3057 }
3058
3059 /**
3060  * percent - convert count to a percentage of total (rounding up or down)
3061  */
3062
3063 static u32 percent(u32 count, u32 total)
3064 {
3065         return (count * 100 + (total / 2)) / total;
3066 }
3067
3068 /**
3069  * tipc_link_stats - print link statistics
3070  * @name: link name
3071  * @buf: print buffer area
3072  * @buf_size: size of print buffer area
3073  *
3074  * Returns length of print buffer data string (or 0 if error)
3075  */
3076
3077 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
3078 {
3079         struct print_buf pb;
3080         struct link *l_ptr;
3081         struct tipc_node *node;
3082         char *status;
3083         u32 profile_total = 0;
3084
3085         if (!strcmp(name, tipc_bclink_name))
3086                 return tipc_bclink_stats(buf, buf_size);
3087
3088         tipc_printbuf_init(&pb, buf, buf_size);
3089
3090         read_lock_bh(&tipc_net_lock);
3091         l_ptr = link_find_link(name, &node);
3092         if (!l_ptr) {
3093                 read_unlock_bh(&tipc_net_lock);
3094                 return 0;
3095         }
3096         tipc_node_lock(node);
3097
3098         if (tipc_link_is_active(l_ptr))
3099                 status = "ACTIVE";
3100         else if (tipc_link_is_up(l_ptr))
3101                 status = "STANDBY";
3102         else
3103                 status = "DEFUNCT";
3104         tipc_printf(&pb, "Link <%s>\n"
3105                          "  %s  MTU:%u  Priority:%u  Tolerance:%u ms"
3106                          "  Window:%u packets\n",
3107                     l_ptr->name, status, l_ptr->max_pkt,
3108                     l_ptr->priority, l_ptr->tolerance, l_ptr->queue_limit[0]);
3109         tipc_printf(&pb, "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
3110                     l_ptr->next_in_no - l_ptr->stats.recv_info,
3111                     l_ptr->stats.recv_fragments,
3112                     l_ptr->stats.recv_fragmented,
3113                     l_ptr->stats.recv_bundles,
3114                     l_ptr->stats.recv_bundled);
3115         tipc_printf(&pb, "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
3116                     l_ptr->next_out_no - l_ptr->stats.sent_info,
3117                     l_ptr->stats.sent_fragments,
3118                     l_ptr->stats.sent_fragmented,
3119                     l_ptr->stats.sent_bundles,
3120                     l_ptr->stats.sent_bundled);
3121         profile_total = l_ptr->stats.msg_length_counts;
3122         if (!profile_total)
3123                 profile_total = 1;
3124         tipc_printf(&pb, "  TX profile sample:%u packets  average:%u octets\n"
3125                          "  0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
3126                          "-16354:%u%% -32768:%u%% -66000:%u%%\n",
3127                     l_ptr->stats.msg_length_counts,
3128                     l_ptr->stats.msg_lengths_total / profile_total,
3129                     percent(l_ptr->stats.msg_length_profile[0], profile_total),
3130                     percent(l_ptr->stats.msg_length_profile[1], profile_total),
3131                     percent(l_ptr->stats.msg_length_profile[2], profile_total),
3132                     percent(l_ptr->stats.msg_length_profile[3], profile_total),
3133                     percent(l_ptr->stats.msg_length_profile[4], profile_total),
3134                     percent(l_ptr->stats.msg_length_profile[5], profile_total),
3135                     percent(l_ptr->stats.msg_length_profile[6], profile_total));
3136         tipc_printf(&pb, "  RX states:%u probes:%u naks:%u defs:%u dups:%u\n",
3137                     l_ptr->stats.recv_states,
3138                     l_ptr->stats.recv_probes,
3139                     l_ptr->stats.recv_nacks,
3140                     l_ptr->stats.deferred_recv,
3141                     l_ptr->stats.duplicates);
3142         tipc_printf(&pb, "  TX states:%u probes:%u naks:%u acks:%u dups:%u\n",
3143                     l_ptr->stats.sent_states,
3144                     l_ptr->stats.sent_probes,
3145                     l_ptr->stats.sent_nacks,
3146                     l_ptr->stats.sent_acks,
3147                     l_ptr->stats.retransmitted);
3148         tipc_printf(&pb, "  Congestion bearer:%u link:%u  Send queue max:%u avg:%u\n",
3149                     l_ptr->stats.bearer_congs,
3150                     l_ptr->stats.link_congs,
3151                     l_ptr->stats.max_queue_sz,
3152                     l_ptr->stats.queue_sz_counts
3153                     ? (l_ptr->stats.accu_queue_sz / l_ptr->stats.queue_sz_counts)
3154                     : 0);
3155
3156         tipc_node_unlock(node);
3157         read_unlock_bh(&tipc_net_lock);
3158         return tipc_printbuf_validate(&pb);
3159 }
3160
3161 #define MAX_LINK_STATS_INFO 2000
3162
3163 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
3164 {
3165         struct sk_buff *buf;
3166         struct tlv_desc *rep_tlv;
3167         int str_len;
3168
3169         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3170                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3171
3172         buf = tipc_cfg_reply_alloc(TLV_SPACE(MAX_LINK_STATS_INFO));
3173         if (!buf)
3174                 return NULL;
3175
3176         rep_tlv = (struct tlv_desc *)buf->data;
3177
3178         str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
3179                                   (char *)TLV_DATA(rep_tlv), MAX_LINK_STATS_INFO);
3180         if (!str_len) {
3181                 buf_discard(buf);
3182                 return tipc_cfg_reply_error_string("link not found");
3183         }
3184
3185         skb_put(buf, TLV_SPACE(str_len));
3186         TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
3187
3188         return buf;
3189 }
3190
3191 #if 0
3192 int link_control(const char *name, u32 op, u32 val)
3193 {
3194         int res = -EINVAL;
3195         struct link *l_ptr;
3196         u32 bearer_id;
3197         struct tipc_node * node;
3198         u32 a;
3199
3200         a = link_name2addr(name, &bearer_id);
3201         read_lock_bh(&tipc_net_lock);
3202         node = tipc_node_find(a);
3203         if (node) {
3204                 tipc_node_lock(node);
3205                 l_ptr = node->links[bearer_id];
3206                 if (l_ptr) {
3207                         if (op == TIPC_REMOVE_LINK) {
3208                                 struct bearer *b_ptr = l_ptr->b_ptr;
3209                                 spin_lock_bh(&b_ptr->publ.lock);
3210                                 tipc_link_delete(l_ptr);
3211                                 spin_unlock_bh(&b_ptr->publ.lock);
3212                         }
3213                         if (op == TIPC_CMD_BLOCK_LINK) {
3214                                 tipc_link_reset(l_ptr);
3215                                 l_ptr->blocked = 1;
3216                         }
3217                         if (op == TIPC_CMD_UNBLOCK_LINK) {
3218                                 l_ptr->blocked = 0;
3219                         }
3220                         res = 0;
3221                 }
3222                 tipc_node_unlock(node);
3223         }
3224         read_unlock_bh(&tipc_net_lock);
3225         return res;
3226 }
3227 #endif
3228
3229 /**
3230  * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
3231  * @dest: network address of destination node
3232  * @selector: used to select from set of active links
3233  *
3234  * If no active link can be found, uses default maximum packet size.
3235  */
3236
3237 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
3238 {
3239         struct tipc_node *n_ptr;
3240         struct link *l_ptr;
3241         u32 res = MAX_PKT_DEFAULT;
3242
3243         if (dest == tipc_own_addr)
3244                 return MAX_MSG_SIZE;
3245
3246         read_lock_bh(&tipc_net_lock);
3247         n_ptr = tipc_node_select(dest, selector);
3248         if (n_ptr) {
3249                 tipc_node_lock(n_ptr);
3250                 l_ptr = n_ptr->active_links[selector & 1];
3251                 if (l_ptr)
3252                         res = l_ptr->max_pkt;
3253                 tipc_node_unlock(n_ptr);
3254         }
3255         read_unlock_bh(&tipc_net_lock);
3256         return res;
3257 }
3258
3259 #if 0
3260 static void link_dump_rec_queue(struct link *l_ptr)
3261 {
3262         struct sk_buff *crs;
3263
3264         if (!l_ptr->oldest_deferred_in) {
3265                 info("Reception queue empty\n");
3266                 return;
3267         }
3268         info("Contents of Reception queue:\n");
3269         crs = l_ptr->oldest_deferred_in;
3270         while (crs) {
3271                 if (crs->data == (void *)0x0000a3a3) {
3272                         info("buffer %x invalid\n", crs);
3273                         return;
3274                 }
3275                 msg_dbg(buf_msg(crs), "In rec queue:\n");
3276                 crs = crs->next;
3277         }
3278 }
3279 #endif
3280
3281 static void link_dump_send_queue(struct link *l_ptr)
3282 {
3283         if (l_ptr->next_out) {
3284                 info("\nContents of unsent queue:\n");
3285                 dbg_print_buf_chain(l_ptr->next_out);
3286         }
3287         info("\nContents of send queue:\n");
3288         if (l_ptr->first_out) {
3289                 dbg_print_buf_chain(l_ptr->first_out);
3290         }
3291         info("Empty send queue\n");
3292 }
3293
3294 static void link_print(struct link *l_ptr, struct print_buf *buf,
3295                        const char *str)
3296 {
3297         tipc_printf(buf, str);
3298         if (link_reset_reset(l_ptr) || link_reset_unknown(l_ptr))
3299                 return;
3300         tipc_printf(buf, "Link %x<%s>:",
3301                     l_ptr->addr, l_ptr->b_ptr->publ.name);
3302         tipc_printf(buf, ": NXO(%u):", mod(l_ptr->next_out_no));
3303         tipc_printf(buf, "NXI(%u):", mod(l_ptr->next_in_no));
3304         tipc_printf(buf, "SQUE");
3305         if (l_ptr->first_out) {
3306                 tipc_printf(buf, "[%u..", msg_seqno(buf_msg(l_ptr->first_out)));
3307                 if (l_ptr->next_out)
3308                         tipc_printf(buf, "%u..",
3309                                     msg_seqno(buf_msg(l_ptr->next_out)));
3310                 tipc_printf(buf, "%u]", msg_seqno(buf_msg(l_ptr->last_out)));
3311                 if ((mod(msg_seqno(buf_msg(l_ptr->last_out)) -
3312                          msg_seqno(buf_msg(l_ptr->first_out)))
3313                      != (l_ptr->out_queue_size - 1)) ||
3314                     (l_ptr->last_out->next != NULL)) {
3315                         tipc_printf(buf, "\nSend queue inconsistency\n");
3316                         tipc_printf(buf, "first_out= %x ", l_ptr->first_out);
3317                         tipc_printf(buf, "next_out= %x ", l_ptr->next_out);
3318                         tipc_printf(buf, "last_out= %x ", l_ptr->last_out);
3319                         link_dump_send_queue(l_ptr);
3320                 }
3321         } else
3322                 tipc_printf(buf, "[]");
3323         tipc_printf(buf, "SQSIZ(%u)", l_ptr->out_queue_size);
3324         if (l_ptr->oldest_deferred_in) {
3325                 u32 o = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
3326                 u32 n = msg_seqno(buf_msg(l_ptr->newest_deferred_in));
3327                 tipc_printf(buf, ":RQUE[%u..%u]", o, n);
3328                 if (l_ptr->deferred_inqueue_sz != mod((n + 1) - o)) {
3329                         tipc_printf(buf, ":RQSIZ(%u)",
3330                                     l_ptr->deferred_inqueue_sz);
3331                 }
3332         }
3333         if (link_working_unknown(l_ptr))
3334                 tipc_printf(buf, ":WU");
3335         if (link_reset_reset(l_ptr))
3336                 tipc_printf(buf, ":RR");
3337         if (link_reset_unknown(l_ptr))
3338                 tipc_printf(buf, ":RU");
3339         if (link_working_working(l_ptr))
3340                 tipc_printf(buf, ":WW");
3341         tipc_printf(buf, "\n");
3342 }
3343