7e0036f5a364c3664d12050583fae8ea3df1c2fc
[cascardo/linux.git] / net / tipc / link.c
1 /*
2  * net/tipc/link.c: TIPC link code
3  *
4  * Copyright (c) 1996-2007, 2012-2015, Ericsson AB
5  * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36
37 #include "core.h"
38 #include "link.h"
39 #include "bcast.h"
40 #include "socket.h"
41 #include "name_distr.h"
42 #include "discover.h"
43 #include "netlink.h"
44
45 #include <linux/pkt_sched.h>
46
47 /*
48  * Error message prefixes
49  */
50 static const char *link_co_err = "Link changeover error, ";
51 static const char *link_rst_msg = "Resetting link ";
52 static const char *link_unk_evt = "Unknown link event ";
53
54 static const struct nla_policy tipc_nl_link_policy[TIPC_NLA_LINK_MAX + 1] = {
55         [TIPC_NLA_LINK_UNSPEC]          = { .type = NLA_UNSPEC },
56         [TIPC_NLA_LINK_NAME] = {
57                 .type = NLA_STRING,
58                 .len = TIPC_MAX_LINK_NAME
59         },
60         [TIPC_NLA_LINK_MTU]             = { .type = NLA_U32 },
61         [TIPC_NLA_LINK_BROADCAST]       = { .type = NLA_FLAG },
62         [TIPC_NLA_LINK_UP]              = { .type = NLA_FLAG },
63         [TIPC_NLA_LINK_ACTIVE]          = { .type = NLA_FLAG },
64         [TIPC_NLA_LINK_PROP]            = { .type = NLA_NESTED },
65         [TIPC_NLA_LINK_STATS]           = { .type = NLA_NESTED },
66         [TIPC_NLA_LINK_RX]              = { .type = NLA_U32 },
67         [TIPC_NLA_LINK_TX]              = { .type = NLA_U32 }
68 };
69
70 /* Properties valid for media, bearar and link */
71 static const struct nla_policy tipc_nl_prop_policy[TIPC_NLA_PROP_MAX + 1] = {
72         [TIPC_NLA_PROP_UNSPEC]          = { .type = NLA_UNSPEC },
73         [TIPC_NLA_PROP_PRIO]            = { .type = NLA_U32 },
74         [TIPC_NLA_PROP_TOL]             = { .type = NLA_U32 },
75         [TIPC_NLA_PROP_WIN]             = { .type = NLA_U32 }
76 };
77
78 /*
79  * Out-of-range value for link session numbers
80  */
81 #define INVALID_SESSION 0x10000
82
83 /*
84  * Link state events:
85  */
86 #define  STARTING_EVT    856384768      /* link processing trigger */
87 #define  TRAFFIC_MSG_EVT 560815u        /* rx'd ??? */
88 #define  TIMEOUT_EVT     560817u        /* link timer expired */
89
90 /*
91  * The following two 'message types' is really just implementation
92  * data conveniently stored in the message header.
93  * They must not be considered part of the protocol
94  */
95 #define OPEN_MSG   0
96 #define CLOSED_MSG 1
97
98 /*
99  * State value stored in 'exp_msg_count'
100  */
101 #define START_CHANGEOVER 100000u
102
103 static void link_handle_out_of_seq_msg(struct tipc_link *link,
104                                        struct sk_buff *skb);
105 static void tipc_link_proto_rcv(struct tipc_link *link,
106                                 struct sk_buff *skb);
107 static int  tipc_link_tunnel_rcv(struct tipc_node *node,
108                                  struct sk_buff **skb);
109 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tol);
110 static void link_state_event(struct tipc_link *l_ptr, u32 event);
111 static void link_reset_statistics(struct tipc_link *l_ptr);
112 static void link_print(struct tipc_link *l_ptr, const char *str);
113 static void tipc_link_sync_xmit(struct tipc_link *l);
114 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf);
115 static void tipc_link_input(struct tipc_link *l, struct sk_buff *skb);
116 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb);
117
118 /*
119  *  Simple link routines
120  */
121 static unsigned int align(unsigned int i)
122 {
123         return (i + 3) & ~3u;
124 }
125
126 static void tipc_link_release(struct kref *kref)
127 {
128         kfree(container_of(kref, struct tipc_link, ref));
129 }
130
131 static void tipc_link_get(struct tipc_link *l_ptr)
132 {
133         kref_get(&l_ptr->ref);
134 }
135
136 static void tipc_link_put(struct tipc_link *l_ptr)
137 {
138         kref_put(&l_ptr->ref, tipc_link_release);
139 }
140
141 static void link_init_max_pkt(struct tipc_link *l_ptr)
142 {
143         struct tipc_node *node = l_ptr->owner;
144         struct tipc_net *tn = net_generic(node->net, tipc_net_id);
145         struct tipc_bearer *b_ptr;
146         u32 max_pkt;
147
148         rcu_read_lock();
149         b_ptr = rcu_dereference_rtnl(tn->bearer_list[l_ptr->bearer_id]);
150         if (!b_ptr) {
151                 rcu_read_unlock();
152                 return;
153         }
154         max_pkt = (b_ptr->mtu & ~3);
155         rcu_read_unlock();
156
157         if (max_pkt > MAX_MSG_SIZE)
158                 max_pkt = MAX_MSG_SIZE;
159
160         l_ptr->max_pkt_target = max_pkt;
161         if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
162                 l_ptr->max_pkt = l_ptr->max_pkt_target;
163         else
164                 l_ptr->max_pkt = MAX_PKT_DEFAULT;
165
166         l_ptr->max_pkt_probes = 0;
167 }
168
169 /*
170  *  Simple non-static link routines (i.e. referenced outside this file)
171  */
172 int tipc_link_is_up(struct tipc_link *l_ptr)
173 {
174         if (!l_ptr)
175                 return 0;
176         return link_working_working(l_ptr) || link_working_unknown(l_ptr);
177 }
178
179 int tipc_link_is_active(struct tipc_link *l_ptr)
180 {
181         return  (l_ptr->owner->active_links[0] == l_ptr) ||
182                 (l_ptr->owner->active_links[1] == l_ptr);
183 }
184
185 /**
186  * link_timeout - handle expiration of link timer
187  * @l_ptr: pointer to link
188  */
189 static void link_timeout(unsigned long data)
190 {
191         struct tipc_link *l_ptr = (struct tipc_link *)data;
192         struct sk_buff *skb;
193
194         tipc_node_lock(l_ptr->owner);
195
196         /* update counters used in statistical profiling of send traffic */
197         l_ptr->stats.accu_queue_sz += skb_queue_len(&l_ptr->transmq);
198         l_ptr->stats.queue_sz_counts++;
199
200         skb = skb_peek(&l_ptr->transmq);
201         if (skb) {
202                 struct tipc_msg *msg = buf_msg(skb);
203                 u32 length = msg_size(msg);
204
205                 if ((msg_user(msg) == MSG_FRAGMENTER) &&
206                     (msg_type(msg) == FIRST_FRAGMENT)) {
207                         length = msg_size(msg_get_wrapped(msg));
208                 }
209                 if (length) {
210                         l_ptr->stats.msg_lengths_total += length;
211                         l_ptr->stats.msg_length_counts++;
212                         if (length <= 64)
213                                 l_ptr->stats.msg_length_profile[0]++;
214                         else if (length <= 256)
215                                 l_ptr->stats.msg_length_profile[1]++;
216                         else if (length <= 1024)
217                                 l_ptr->stats.msg_length_profile[2]++;
218                         else if (length <= 4096)
219                                 l_ptr->stats.msg_length_profile[3]++;
220                         else if (length <= 16384)
221                                 l_ptr->stats.msg_length_profile[4]++;
222                         else if (length <= 32768)
223                                 l_ptr->stats.msg_length_profile[5]++;
224                         else
225                                 l_ptr->stats.msg_length_profile[6]++;
226                 }
227         }
228
229         /* do all other link processing performed on a periodic basis */
230         link_state_event(l_ptr, TIMEOUT_EVT);
231
232         if (skb_queue_len(&l_ptr->backlogq))
233                 tipc_link_push_packets(l_ptr);
234
235         tipc_node_unlock(l_ptr->owner);
236         tipc_link_put(l_ptr);
237 }
238
239 static void link_set_timer(struct tipc_link *link, unsigned long time)
240 {
241         if (!mod_timer(&link->timer, jiffies + time))
242                 tipc_link_get(link);
243 }
244
245 /**
246  * tipc_link_create - create a new link
247  * @n_ptr: pointer to associated node
248  * @b_ptr: pointer to associated bearer
249  * @media_addr: media address to use when sending messages over link
250  *
251  * Returns pointer to link.
252  */
253 struct tipc_link *tipc_link_create(struct tipc_node *n_ptr,
254                                    struct tipc_bearer *b_ptr,
255                                    const struct tipc_media_addr *media_addr)
256 {
257         struct tipc_net *tn = net_generic(n_ptr->net, tipc_net_id);
258         struct tipc_link *l_ptr;
259         struct tipc_msg *msg;
260         char *if_name;
261         char addr_string[16];
262         u32 peer = n_ptr->addr;
263
264         if (n_ptr->link_cnt >= MAX_BEARERS) {
265                 tipc_addr_string_fill(addr_string, n_ptr->addr);
266                 pr_err("Attempt to establish %uth link to %s. Max %u allowed.\n",
267                         n_ptr->link_cnt, addr_string, MAX_BEARERS);
268                 return NULL;
269         }
270
271         if (n_ptr->links[b_ptr->identity]) {
272                 tipc_addr_string_fill(addr_string, n_ptr->addr);
273                 pr_err("Attempt to establish second link on <%s> to %s\n",
274                        b_ptr->name, addr_string);
275                 return NULL;
276         }
277
278         l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
279         if (!l_ptr) {
280                 pr_warn("Link creation failed, no memory\n");
281                 return NULL;
282         }
283         kref_init(&l_ptr->ref);
284         l_ptr->addr = peer;
285         if_name = strchr(b_ptr->name, ':') + 1;
286         sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
287                 tipc_zone(tn->own_addr), tipc_cluster(tn->own_addr),
288                 tipc_node(tn->own_addr),
289                 if_name,
290                 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
291                 /* note: peer i/f name is updated by reset/activate message */
292         memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
293         l_ptr->owner = n_ptr;
294         l_ptr->checkpoint = 1;
295         l_ptr->peer_session = INVALID_SESSION;
296         l_ptr->bearer_id = b_ptr->identity;
297         link_set_supervision_props(l_ptr, b_ptr->tolerance);
298         l_ptr->state = RESET_UNKNOWN;
299
300         l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
301         msg = l_ptr->pmsg;
302         tipc_msg_init(tn->own_addr, msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE,
303                       l_ptr->addr);
304         msg_set_size(msg, sizeof(l_ptr->proto_msg));
305         msg_set_session(msg, (tn->random & 0xffff));
306         msg_set_bearer_id(msg, b_ptr->identity);
307         strcpy((char *)msg_data(msg), if_name);
308
309         l_ptr->priority = b_ptr->priority;
310         tipc_link_set_queue_limits(l_ptr, b_ptr->window);
311
312         l_ptr->net_plane = b_ptr->net_plane;
313         link_init_max_pkt(l_ptr);
314
315         l_ptr->next_out_no = 1;
316         __skb_queue_head_init(&l_ptr->transmq);
317         __skb_queue_head_init(&l_ptr->backlogq);
318         __skb_queue_head_init(&l_ptr->deferdq);
319         skb_queue_head_init(&l_ptr->wakeupq);
320         skb_queue_head_init(&l_ptr->inputq);
321         skb_queue_head_init(&l_ptr->namedq);
322         link_reset_statistics(l_ptr);
323         tipc_node_attach_link(n_ptr, l_ptr);
324         setup_timer(&l_ptr->timer, link_timeout, (unsigned long)l_ptr);
325         link_state_event(l_ptr, STARTING_EVT);
326
327         return l_ptr;
328 }
329
330 /**
331  * link_delete - Conditional deletion of link.
332  *               If timer still running, real delete is done when it expires
333  * @link: link to be deleted
334  */
335 void tipc_link_delete(struct tipc_link *link)
336 {
337         tipc_link_reset_fragments(link);
338         tipc_node_detach_link(link->owner, link);
339         tipc_link_put(link);
340 }
341
342 void tipc_link_delete_list(struct net *net, unsigned int bearer_id,
343                            bool shutting_down)
344 {
345         struct tipc_net *tn = net_generic(net, tipc_net_id);
346         struct tipc_link *link;
347         struct tipc_node *node;
348         bool del_link;
349
350         rcu_read_lock();
351         list_for_each_entry_rcu(node, &tn->node_list, list) {
352                 tipc_node_lock(node);
353                 link = node->links[bearer_id];
354                 if (!link) {
355                         tipc_node_unlock(node);
356                         continue;
357                 }
358                 del_link = !tipc_link_is_up(link) && !link->exp_msg_count;
359                 tipc_link_reset(link);
360                 if (del_timer(&link->timer))
361                         tipc_link_put(link);
362                 link->flags |= LINK_STOPPED;
363                 /* Delete link now, or when failover is finished: */
364                 if (shutting_down || !tipc_node_is_up(node) || del_link)
365                         tipc_link_delete(link);
366                 tipc_node_unlock(node);
367         }
368         rcu_read_unlock();
369 }
370
371 /**
372  * link_schedule_user - schedule user for wakeup after congestion
373  * @link: congested link
374  * @oport: sending port
375  * @chain_sz: size of buffer chain that was attempted sent
376  * @imp: importance of message attempted sent
377  * Create pseudo msg to send back to user when congestion abates
378  */
379 static bool link_schedule_user(struct tipc_link *link, u32 oport,
380                                uint chain_sz, uint imp)
381 {
382         struct sk_buff *buf;
383
384         buf = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
385                               link_own_addr(link), link_own_addr(link),
386                               oport, 0, 0);
387         if (!buf)
388                 return false;
389         TIPC_SKB_CB(buf)->chain_sz = chain_sz;
390         TIPC_SKB_CB(buf)->chain_imp = imp;
391         skb_queue_tail(&link->wakeupq, buf);
392         link->stats.link_congs++;
393         return true;
394 }
395
396 /**
397  * link_prepare_wakeup - prepare users for wakeup after congestion
398  * @link: congested link
399  * Move a number of waiting users, as permitted by available space in
400  * the send queue, from link wait queue to node wait queue for wakeup
401  */
402 void link_prepare_wakeup(struct tipc_link *link)
403 {
404         uint pend_qsz = skb_queue_len(&link->backlogq);
405         struct sk_buff *skb, *tmp;
406
407         skb_queue_walk_safe(&link->wakeupq, skb, tmp) {
408                 if (pend_qsz >= link->queue_limit[TIPC_SKB_CB(skb)->chain_imp])
409                         break;
410                 pend_qsz += TIPC_SKB_CB(skb)->chain_sz;
411                 skb_unlink(skb, &link->wakeupq);
412                 skb_queue_tail(&link->inputq, skb);
413                 link->owner->inputq = &link->inputq;
414                 link->owner->action_flags |= TIPC_MSG_EVT;
415         }
416 }
417
418 /**
419  * tipc_link_reset_fragments - purge link's inbound message fragments queue
420  * @l_ptr: pointer to link
421  */
422 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
423 {
424         kfree_skb(l_ptr->reasm_buf);
425         l_ptr->reasm_buf = NULL;
426 }
427
428 /**
429  * tipc_link_purge_queues - purge all pkt queues associated with link
430  * @l_ptr: pointer to link
431  */
432 void tipc_link_purge_queues(struct tipc_link *l_ptr)
433 {
434         __skb_queue_purge(&l_ptr->deferdq);
435         __skb_queue_purge(&l_ptr->transmq);
436         __skb_queue_purge(&l_ptr->backlogq);
437         tipc_link_reset_fragments(l_ptr);
438 }
439
440 void tipc_link_reset(struct tipc_link *l_ptr)
441 {
442         u32 prev_state = l_ptr->state;
443         u32 checkpoint = l_ptr->next_in_no;
444         int was_active_link = tipc_link_is_active(l_ptr);
445         struct tipc_node *owner = l_ptr->owner;
446
447         msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
448
449         /* Link is down, accept any session */
450         l_ptr->peer_session = INVALID_SESSION;
451
452         /* Prepare for max packet size negotiation */
453         link_init_max_pkt(l_ptr);
454
455         l_ptr->state = RESET_UNKNOWN;
456
457         if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
458                 return;
459
460         tipc_node_link_down(l_ptr->owner, l_ptr);
461         tipc_bearer_remove_dest(owner->net, l_ptr->bearer_id, l_ptr->addr);
462
463         if (was_active_link && tipc_node_active_links(l_ptr->owner)) {
464                 l_ptr->reset_checkpoint = checkpoint;
465                 l_ptr->exp_msg_count = START_CHANGEOVER;
466         }
467
468         /* Clean up all queues, except inputq: */
469         __skb_queue_purge(&l_ptr->transmq);
470         __skb_queue_purge(&l_ptr->backlogq);
471         __skb_queue_purge(&l_ptr->deferdq);
472         if (!owner->inputq)
473                 owner->inputq = &l_ptr->inputq;
474         skb_queue_splice_init(&l_ptr->wakeupq, owner->inputq);
475         if (!skb_queue_empty(owner->inputq))
476                 owner->action_flags |= TIPC_MSG_EVT;
477         l_ptr->rcv_unacked = 0;
478         l_ptr->checkpoint = 1;
479         l_ptr->next_out_no = 1;
480         l_ptr->fsm_msg_cnt = 0;
481         l_ptr->stale_count = 0;
482         link_reset_statistics(l_ptr);
483 }
484
485 void tipc_link_reset_list(struct net *net, unsigned int bearer_id)
486 {
487         struct tipc_net *tn = net_generic(net, tipc_net_id);
488         struct tipc_link *l_ptr;
489         struct tipc_node *n_ptr;
490
491         rcu_read_lock();
492         list_for_each_entry_rcu(n_ptr, &tn->node_list, list) {
493                 tipc_node_lock(n_ptr);
494                 l_ptr = n_ptr->links[bearer_id];
495                 if (l_ptr)
496                         tipc_link_reset(l_ptr);
497                 tipc_node_unlock(n_ptr);
498         }
499         rcu_read_unlock();
500 }
501
502 static void link_activate(struct tipc_link *link)
503 {
504         struct tipc_node *node = link->owner;
505
506         link->next_in_no = 1;
507         link->stats.recv_info = 1;
508         tipc_node_link_up(node, link);
509         tipc_bearer_add_dest(node->net, link->bearer_id, link->addr);
510 }
511
512 /**
513  * link_state_event - link finite state machine
514  * @l_ptr: pointer to link
515  * @event: state machine event to process
516  */
517 static void link_state_event(struct tipc_link *l_ptr, unsigned int event)
518 {
519         struct tipc_link *other;
520         unsigned long cont_intv = l_ptr->cont_intv;
521
522         if (l_ptr->flags & LINK_STOPPED)
523                 return;
524
525         if (!(l_ptr->flags & LINK_STARTED) && (event != STARTING_EVT))
526                 return;         /* Not yet. */
527
528         /* Check whether changeover is going on */
529         if (l_ptr->exp_msg_count) {
530                 if (event == TIMEOUT_EVT)
531                         link_set_timer(l_ptr, cont_intv);
532                 return;
533         }
534
535         switch (l_ptr->state) {
536         case WORKING_WORKING:
537                 switch (event) {
538                 case TRAFFIC_MSG_EVT:
539                 case ACTIVATE_MSG:
540                         break;
541                 case TIMEOUT_EVT:
542                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
543                                 l_ptr->checkpoint = l_ptr->next_in_no;
544                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
545                                         tipc_link_proto_xmit(l_ptr, STATE_MSG,
546                                                              0, 0, 0, 0, 0);
547                                         l_ptr->fsm_msg_cnt++;
548                                 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
549                                         tipc_link_proto_xmit(l_ptr, STATE_MSG,
550                                                              1, 0, 0, 0, 0);
551                                         l_ptr->fsm_msg_cnt++;
552                                 }
553                                 link_set_timer(l_ptr, cont_intv);
554                                 break;
555                         }
556                         l_ptr->state = WORKING_UNKNOWN;
557                         l_ptr->fsm_msg_cnt = 0;
558                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
559                         l_ptr->fsm_msg_cnt++;
560                         link_set_timer(l_ptr, cont_intv / 4);
561                         break;
562                 case RESET_MSG:
563                         pr_debug("%s<%s>, requested by peer\n",
564                                  link_rst_msg, l_ptr->name);
565                         tipc_link_reset(l_ptr);
566                         l_ptr->state = RESET_RESET;
567                         l_ptr->fsm_msg_cnt = 0;
568                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
569                                              0, 0, 0, 0, 0);
570                         l_ptr->fsm_msg_cnt++;
571                         link_set_timer(l_ptr, cont_intv);
572                         break;
573                 default:
574                         pr_debug("%s%u in WW state\n", link_unk_evt, event);
575                 }
576                 break;
577         case WORKING_UNKNOWN:
578                 switch (event) {
579                 case TRAFFIC_MSG_EVT:
580                 case ACTIVATE_MSG:
581                         l_ptr->state = WORKING_WORKING;
582                         l_ptr->fsm_msg_cnt = 0;
583                         link_set_timer(l_ptr, cont_intv);
584                         break;
585                 case RESET_MSG:
586                         pr_debug("%s<%s>, requested by peer while probing\n",
587                                  link_rst_msg, l_ptr->name);
588                         tipc_link_reset(l_ptr);
589                         l_ptr->state = RESET_RESET;
590                         l_ptr->fsm_msg_cnt = 0;
591                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
592                                              0, 0, 0, 0, 0);
593                         l_ptr->fsm_msg_cnt++;
594                         link_set_timer(l_ptr, cont_intv);
595                         break;
596                 case TIMEOUT_EVT:
597                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
598                                 l_ptr->state = WORKING_WORKING;
599                                 l_ptr->fsm_msg_cnt = 0;
600                                 l_ptr->checkpoint = l_ptr->next_in_no;
601                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
602                                         tipc_link_proto_xmit(l_ptr, STATE_MSG,
603                                                              0, 0, 0, 0, 0);
604                                         l_ptr->fsm_msg_cnt++;
605                                 }
606                                 link_set_timer(l_ptr, cont_intv);
607                         } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
608                                 tipc_link_proto_xmit(l_ptr, STATE_MSG,
609                                                      1, 0, 0, 0, 0);
610                                 l_ptr->fsm_msg_cnt++;
611                                 link_set_timer(l_ptr, cont_intv / 4);
612                         } else {        /* Link has failed */
613                                 pr_debug("%s<%s>, peer not responding\n",
614                                          link_rst_msg, l_ptr->name);
615                                 tipc_link_reset(l_ptr);
616                                 l_ptr->state = RESET_UNKNOWN;
617                                 l_ptr->fsm_msg_cnt = 0;
618                                 tipc_link_proto_xmit(l_ptr, RESET_MSG,
619                                                      0, 0, 0, 0, 0);
620                                 l_ptr->fsm_msg_cnt++;
621                                 link_set_timer(l_ptr, cont_intv);
622                         }
623                         break;
624                 default:
625                         pr_err("%s%u in WU state\n", link_unk_evt, event);
626                 }
627                 break;
628         case RESET_UNKNOWN:
629                 switch (event) {
630                 case TRAFFIC_MSG_EVT:
631                         break;
632                 case ACTIVATE_MSG:
633                         other = l_ptr->owner->active_links[0];
634                         if (other && link_working_unknown(other))
635                                 break;
636                         l_ptr->state = WORKING_WORKING;
637                         l_ptr->fsm_msg_cnt = 0;
638                         link_activate(l_ptr);
639                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
640                         l_ptr->fsm_msg_cnt++;
641                         if (l_ptr->owner->working_links == 1)
642                                 tipc_link_sync_xmit(l_ptr);
643                         link_set_timer(l_ptr, cont_intv);
644                         break;
645                 case RESET_MSG:
646                         l_ptr->state = RESET_RESET;
647                         l_ptr->fsm_msg_cnt = 0;
648                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
649                                              1, 0, 0, 0, 0);
650                         l_ptr->fsm_msg_cnt++;
651                         link_set_timer(l_ptr, cont_intv);
652                         break;
653                 case STARTING_EVT:
654                         l_ptr->flags |= LINK_STARTED;
655                         l_ptr->fsm_msg_cnt++;
656                         link_set_timer(l_ptr, cont_intv);
657                         break;
658                 case TIMEOUT_EVT:
659                         tipc_link_proto_xmit(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
660                         l_ptr->fsm_msg_cnt++;
661                         link_set_timer(l_ptr, cont_intv);
662                         break;
663                 default:
664                         pr_err("%s%u in RU state\n", link_unk_evt, event);
665                 }
666                 break;
667         case RESET_RESET:
668                 switch (event) {
669                 case TRAFFIC_MSG_EVT:
670                 case ACTIVATE_MSG:
671                         other = l_ptr->owner->active_links[0];
672                         if (other && link_working_unknown(other))
673                                 break;
674                         l_ptr->state = WORKING_WORKING;
675                         l_ptr->fsm_msg_cnt = 0;
676                         link_activate(l_ptr);
677                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
678                         l_ptr->fsm_msg_cnt++;
679                         if (l_ptr->owner->working_links == 1)
680                                 tipc_link_sync_xmit(l_ptr);
681                         link_set_timer(l_ptr, cont_intv);
682                         break;
683                 case RESET_MSG:
684                         break;
685                 case TIMEOUT_EVT:
686                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
687                                              0, 0, 0, 0, 0);
688                         l_ptr->fsm_msg_cnt++;
689                         link_set_timer(l_ptr, cont_intv);
690                         break;
691                 default:
692                         pr_err("%s%u in RR state\n", link_unk_evt, event);
693                 }
694                 break;
695         default:
696                 pr_err("Unknown link state %u/%u\n", l_ptr->state, event);
697         }
698 }
699
700 /* tipc_link_cong: determine return value and how to treat the
701  * sent buffer during link congestion.
702  * - For plain, errorless user data messages we keep the buffer and
703  *   return -ELINKONG.
704  * - For all other messages we discard the buffer and return -EHOSTUNREACH
705  * - For TIPC internal messages we also reset the link
706  */
707 static int tipc_link_cong(struct tipc_link *link, struct sk_buff_head *list)
708 {
709         struct sk_buff *skb = skb_peek(list);
710         struct tipc_msg *msg = buf_msg(skb);
711         uint imp = tipc_msg_tot_importance(msg);
712         u32 oport = msg_tot_origport(msg);
713
714         if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
715                 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
716                 tipc_link_reset(link);
717                 goto drop;
718         }
719         if (unlikely(msg_errcode(msg)))
720                 goto drop;
721         if (unlikely(msg_reroute_cnt(msg)))
722                 goto drop;
723         if (TIPC_SKB_CB(skb)->wakeup_pending)
724                 return -ELINKCONG;
725         if (link_schedule_user(link, oport, skb_queue_len(list), imp))
726                 return -ELINKCONG;
727 drop:
728         __skb_queue_purge(list);
729         return -EHOSTUNREACH;
730 }
731
732 /**
733  * __tipc_link_xmit(): same as tipc_link_xmit, but destlink is known & locked
734  * @link: link to use
735  * @list: chain of buffers containing message
736  *
737  * Consumes the buffer chain, except when returning -ELINKCONG
738  * Returns 0 if success, otherwise errno: -ELINKCONG, -EMSGSIZE (plain socket
739  * user data messages) or -EHOSTUNREACH (all other messages/senders)
740  * Only the socket functions tipc_send_stream() and tipc_send_packet() need
741  * to act on the return value, since they may need to do more send attempts.
742  */
743 int __tipc_link_xmit(struct net *net, struct tipc_link *link,
744                      struct sk_buff_head *list)
745 {
746         struct tipc_msg *msg = buf_msg(skb_peek(list));
747         unsigned int maxwin = link->window;
748         uint imp = tipc_msg_tot_importance(msg);
749         uint mtu = link->max_pkt;
750         uint ack = mod(link->next_in_no - 1);
751         uint seqno = link->next_out_no;
752         uint bc_last_in = link->owner->bclink.last_in;
753         struct tipc_media_addr *addr = &link->media_addr;
754         struct sk_buff_head *transmq = &link->transmq;
755         struct sk_buff_head *backlogq = &link->backlogq;
756         struct sk_buff *skb, *tmp;
757
758         /* Match queue limits against msg importance: */
759         if (unlikely(skb_queue_len(backlogq) >= link->queue_limit[imp]))
760                 return tipc_link_cong(link, list);
761
762         /* Has valid packet limit been used ? */
763         if (unlikely(msg_size(msg) > mtu)) {
764                 __skb_queue_purge(list);
765                 return -EMSGSIZE;
766         }
767
768         /* Prepare each packet for sending, and add to relevant queue: */
769         skb_queue_walk_safe(list, skb, tmp) {
770                 __skb_unlink(skb, list);
771                 msg = buf_msg(skb);
772                 msg_set_seqno(msg, seqno);
773                 msg_set_ack(msg, ack);
774                 msg_set_bcast_ack(msg, bc_last_in);
775
776                 if (likely(skb_queue_len(transmq) < maxwin)) {
777                         __skb_queue_tail(transmq, skb);
778                         tipc_bearer_send(net, link->bearer_id, skb, addr);
779                         link->rcv_unacked = 0;
780                         seqno++;
781                         continue;
782                 }
783                 if (tipc_msg_bundle(skb_peek_tail(backlogq), skb, mtu)) {
784                         link->stats.sent_bundled++;
785                         continue;
786                 }
787                 if (tipc_msg_make_bundle(&skb, mtu, link->addr)) {
788                         link->stats.sent_bundled++;
789                         link->stats.sent_bundles++;
790                 }
791                 __skb_queue_tail(backlogq, skb);
792                 seqno++;
793         }
794         link->next_out_no = seqno;
795         return 0;
796 }
797
798 static void skb2list(struct sk_buff *skb, struct sk_buff_head *list)
799 {
800         skb_queue_head_init(list);
801         __skb_queue_tail(list, skb);
802 }
803
804 static int __tipc_link_xmit_skb(struct tipc_link *link, struct sk_buff *skb)
805 {
806         struct sk_buff_head head;
807
808         skb2list(skb, &head);
809         return __tipc_link_xmit(link->owner->net, link, &head);
810 }
811
812 int tipc_link_xmit_skb(struct net *net, struct sk_buff *skb, u32 dnode,
813                        u32 selector)
814 {
815         struct sk_buff_head head;
816
817         skb2list(skb, &head);
818         return tipc_link_xmit(net, &head, dnode, selector);
819 }
820
821 /**
822  * tipc_link_xmit() is the general link level function for message sending
823  * @net: the applicable net namespace
824  * @list: chain of buffers containing message
825  * @dsz: amount of user data to be sent
826  * @dnode: address of destination node
827  * @selector: a number used for deterministic link selection
828  * Consumes the buffer chain, except when returning -ELINKCONG
829  * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
830  */
831 int tipc_link_xmit(struct net *net, struct sk_buff_head *list, u32 dnode,
832                    u32 selector)
833 {
834         struct tipc_link *link = NULL;
835         struct tipc_node *node;
836         int rc = -EHOSTUNREACH;
837
838         node = tipc_node_find(net, dnode);
839         if (node) {
840                 tipc_node_lock(node);
841                 link = node->active_links[selector & 1];
842                 if (link)
843                         rc = __tipc_link_xmit(net, link, list);
844                 tipc_node_unlock(node);
845         }
846         if (link)
847                 return rc;
848
849         if (likely(in_own_node(net, dnode)))
850                 return tipc_sk_rcv(net, list);
851
852         __skb_queue_purge(list);
853         return rc;
854 }
855
856 /*
857  * tipc_link_sync_xmit - synchronize broadcast link endpoints.
858  *
859  * Give a newly added peer node the sequence number where it should
860  * start receiving and acking broadcast packets.
861  *
862  * Called with node locked
863  */
864 static void tipc_link_sync_xmit(struct tipc_link *link)
865 {
866         struct sk_buff *skb;
867         struct tipc_msg *msg;
868
869         skb = tipc_buf_acquire(INT_H_SIZE);
870         if (!skb)
871                 return;
872
873         msg = buf_msg(skb);
874         tipc_msg_init(link_own_addr(link), msg, BCAST_PROTOCOL, STATE_MSG,
875                       INT_H_SIZE, link->addr);
876         msg_set_last_bcast(msg, link->owner->bclink.acked);
877         __tipc_link_xmit_skb(link, skb);
878 }
879
880 /*
881  * tipc_link_sync_rcv - synchronize broadcast link endpoints.
882  * Receive the sequence number where we should start receiving and
883  * acking broadcast packets from a newly added peer node, and open
884  * up for reception of such packets.
885  *
886  * Called with node locked
887  */
888 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf)
889 {
890         struct tipc_msg *msg = buf_msg(buf);
891
892         n->bclink.last_sent = n->bclink.last_in = msg_last_bcast(msg);
893         n->bclink.recv_permitted = true;
894         kfree_skb(buf);
895 }
896
897 /*
898  * tipc_link_push_packets - push unsent packets to bearer
899  *
900  * Push out the unsent messages of a link where congestion
901  * has abated. Node is locked.
902  *
903  * Called with node locked
904  */
905 void tipc_link_push_packets(struct tipc_link *link)
906 {
907         struct sk_buff *skb;
908         struct tipc_msg *msg;
909         unsigned int ack = mod(link->next_in_no - 1);
910
911         while (skb_queue_len(&link->transmq) < link->window) {
912                 skb = __skb_dequeue(&link->backlogq);
913                 if (!skb)
914                         break;
915                 msg = buf_msg(skb);
916                 msg_set_ack(msg, ack);
917                 msg_set_bcast_ack(msg, link->owner->bclink.last_in);
918                 link->rcv_unacked = 0;
919                 __skb_queue_tail(&link->transmq, skb);
920                 tipc_bearer_send(link->owner->net, link->bearer_id,
921                                  skb, &link->media_addr);
922         }
923 }
924
925 void tipc_link_reset_all(struct tipc_node *node)
926 {
927         char addr_string[16];
928         u32 i;
929
930         tipc_node_lock(node);
931
932         pr_warn("Resetting all links to %s\n",
933                 tipc_addr_string_fill(addr_string, node->addr));
934
935         for (i = 0; i < MAX_BEARERS; i++) {
936                 if (node->links[i]) {
937                         link_print(node->links[i], "Resetting link\n");
938                         tipc_link_reset(node->links[i]);
939                 }
940         }
941
942         tipc_node_unlock(node);
943 }
944
945 static void link_retransmit_failure(struct tipc_link *l_ptr,
946                                     struct sk_buff *buf)
947 {
948         struct tipc_msg *msg = buf_msg(buf);
949         struct net *net = l_ptr->owner->net;
950
951         pr_warn("Retransmission failure on link <%s>\n", l_ptr->name);
952
953         if (l_ptr->addr) {
954                 /* Handle failure on standard link */
955                 link_print(l_ptr, "Resetting link\n");
956                 tipc_link_reset(l_ptr);
957
958         } else {
959                 /* Handle failure on broadcast link */
960                 struct tipc_node *n_ptr;
961                 char addr_string[16];
962
963                 pr_info("Msg seq number: %u,  ", msg_seqno(msg));
964                 pr_cont("Outstanding acks: %lu\n",
965                         (unsigned long) TIPC_SKB_CB(buf)->handle);
966
967                 n_ptr = tipc_bclink_retransmit_to(net);
968                 tipc_node_lock(n_ptr);
969
970                 tipc_addr_string_fill(addr_string, n_ptr->addr);
971                 pr_info("Broadcast link info for %s\n", addr_string);
972                 pr_info("Reception permitted: %d,  Acked: %u\n",
973                         n_ptr->bclink.recv_permitted,
974                         n_ptr->bclink.acked);
975                 pr_info("Last in: %u,  Oos state: %u,  Last sent: %u\n",
976                         n_ptr->bclink.last_in,
977                         n_ptr->bclink.oos_state,
978                         n_ptr->bclink.last_sent);
979
980                 tipc_node_unlock(n_ptr);
981
982                 tipc_bclink_set_flags(net, TIPC_BCLINK_RESET);
983                 l_ptr->stale_count = 0;
984         }
985 }
986
987 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *skb,
988                           u32 retransmits)
989 {
990         struct tipc_msg *msg;
991
992         if (!skb)
993                 return;
994
995         msg = buf_msg(skb);
996
997         /* Detect repeated retransmit failures */
998         if (l_ptr->last_retransmitted == msg_seqno(msg)) {
999                 if (++l_ptr->stale_count > 100) {
1000                         link_retransmit_failure(l_ptr, skb);
1001                         return;
1002                 }
1003         } else {
1004                 l_ptr->last_retransmitted = msg_seqno(msg);
1005                 l_ptr->stale_count = 1;
1006         }
1007
1008         skb_queue_walk_from(&l_ptr->transmq, skb) {
1009                 if (!retransmits)
1010                         break;
1011                 msg = buf_msg(skb);
1012                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1013                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1014                 tipc_bearer_send(l_ptr->owner->net, l_ptr->bearer_id, skb,
1015                                  &l_ptr->media_addr);
1016                 retransmits--;
1017                 l_ptr->stats.retransmitted++;
1018         }
1019 }
1020
1021 static void link_retrieve_defq(struct tipc_link *link,
1022                                struct sk_buff_head *list)
1023 {
1024         u32 seq_no;
1025
1026         if (skb_queue_empty(&link->deferdq))
1027                 return;
1028
1029         seq_no = buf_seqno(skb_peek(&link->deferdq));
1030         if (seq_no == mod(link->next_in_no))
1031                 skb_queue_splice_tail_init(&link->deferdq, list);
1032 }
1033
1034 /**
1035  * tipc_rcv - process TIPC packets/messages arriving from off-node
1036  * @net: the applicable net namespace
1037  * @skb: TIPC packet
1038  * @b_ptr: pointer to bearer message arrived on
1039  *
1040  * Invoked with no locks held.  Bearer pointer must point to a valid bearer
1041  * structure (i.e. cannot be NULL), but bearer can be inactive.
1042  */
1043 void tipc_rcv(struct net *net, struct sk_buff *skb, struct tipc_bearer *b_ptr)
1044 {
1045         struct tipc_net *tn = net_generic(net, tipc_net_id);
1046         struct sk_buff_head head;
1047         struct tipc_node *n_ptr;
1048         struct tipc_link *l_ptr;
1049         struct sk_buff *skb1, *tmp;
1050         struct tipc_msg *msg;
1051         u32 seq_no;
1052         u32 ackd;
1053         u32 released;
1054
1055         skb2list(skb, &head);
1056
1057         while ((skb = __skb_dequeue(&head))) {
1058                 /* Ensure message is well-formed */
1059                 if (unlikely(!tipc_msg_validate(skb)))
1060                         goto discard;
1061
1062                 /* Handle arrival of a non-unicast link message */
1063                 msg = buf_msg(skb);
1064                 if (unlikely(msg_non_seq(msg))) {
1065                         if (msg_user(msg) ==  LINK_CONFIG)
1066                                 tipc_disc_rcv(net, skb, b_ptr);
1067                         else
1068                                 tipc_bclink_rcv(net, skb);
1069                         continue;
1070                 }
1071
1072                 /* Discard unicast link messages destined for another node */
1073                 if (unlikely(!msg_short(msg) &&
1074                              (msg_destnode(msg) != tn->own_addr)))
1075                         goto discard;
1076
1077                 /* Locate neighboring node that sent message */
1078                 n_ptr = tipc_node_find(net, msg_prevnode(msg));
1079                 if (unlikely(!n_ptr))
1080                         goto discard;
1081                 tipc_node_lock(n_ptr);
1082
1083                 /* Locate unicast link endpoint that should handle message */
1084                 l_ptr = n_ptr->links[b_ptr->identity];
1085                 if (unlikely(!l_ptr))
1086                         goto unlock;
1087
1088                 /* Verify that communication with node is currently allowed */
1089                 if ((n_ptr->action_flags & TIPC_WAIT_PEER_LINKS_DOWN) &&
1090                     msg_user(msg) == LINK_PROTOCOL &&
1091                     (msg_type(msg) == RESET_MSG ||
1092                     msg_type(msg) == ACTIVATE_MSG) &&
1093                     !msg_redundant_link(msg))
1094                         n_ptr->action_flags &= ~TIPC_WAIT_PEER_LINKS_DOWN;
1095
1096                 if (tipc_node_blocked(n_ptr))
1097                         goto unlock;
1098
1099                 /* Validate message sequence number info */
1100                 seq_no = msg_seqno(msg);
1101                 ackd = msg_ack(msg);
1102
1103                 /* Release acked messages */
1104                 if (unlikely(n_ptr->bclink.acked != msg_bcast_ack(msg)))
1105                         tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1106
1107                 released = 0;
1108                 skb_queue_walk_safe(&l_ptr->transmq, skb1, tmp) {
1109                         if (more(buf_seqno(skb1), ackd))
1110                                 break;
1111                          __skb_unlink(skb1, &l_ptr->transmq);
1112                          kfree_skb(skb1);
1113                          released = 1;
1114                 }
1115
1116                 /* Try sending any messages link endpoint has pending */
1117                 if (unlikely(skb_queue_len(&l_ptr->backlogq)))
1118                         tipc_link_push_packets(l_ptr);
1119
1120                 if (released && !skb_queue_empty(&l_ptr->wakeupq))
1121                         link_prepare_wakeup(l_ptr);
1122
1123                 /* Process the incoming packet */
1124                 if (unlikely(!link_working_working(l_ptr))) {
1125                         if (msg_user(msg) == LINK_PROTOCOL) {
1126                                 tipc_link_proto_rcv(l_ptr, skb);
1127                                 link_retrieve_defq(l_ptr, &head);
1128                                 skb = NULL;
1129                                 goto unlock;
1130                         }
1131
1132                         /* Traffic message. Conditionally activate link */
1133                         link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1134
1135                         if (link_working_working(l_ptr)) {
1136                                 /* Re-insert buffer in front of queue */
1137                                 __skb_queue_head(&head, skb);
1138                                 skb = NULL;
1139                                 goto unlock;
1140                         }
1141                         goto unlock;
1142                 }
1143
1144                 /* Link is now in state WORKING_WORKING */
1145                 if (unlikely(seq_no != mod(l_ptr->next_in_no))) {
1146                         link_handle_out_of_seq_msg(l_ptr, skb);
1147                         link_retrieve_defq(l_ptr, &head);
1148                         skb = NULL;
1149                         goto unlock;
1150                 }
1151                 l_ptr->next_in_no++;
1152                 if (unlikely(!skb_queue_empty(&l_ptr->deferdq)))
1153                         link_retrieve_defq(l_ptr, &head);
1154                 if (unlikely(++l_ptr->rcv_unacked >= TIPC_MIN_LINK_WIN)) {
1155                         l_ptr->stats.sent_acks++;
1156                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1157                 }
1158                 tipc_link_input(l_ptr, skb);
1159                 skb = NULL;
1160 unlock:
1161                 tipc_node_unlock(n_ptr);
1162 discard:
1163                 if (unlikely(skb))
1164                         kfree_skb(skb);
1165         }
1166 }
1167
1168 /* tipc_data_input - deliver data and name distr msgs to upper layer
1169  *
1170  * Consumes buffer if message is of right type
1171  * Node lock must be held
1172  */
1173 static bool tipc_data_input(struct tipc_link *link, struct sk_buff *skb)
1174 {
1175         struct tipc_node *node = link->owner;
1176         struct tipc_msg *msg = buf_msg(skb);
1177         u32 dport = msg_destport(msg);
1178
1179         switch (msg_user(msg)) {
1180         case TIPC_LOW_IMPORTANCE:
1181         case TIPC_MEDIUM_IMPORTANCE:
1182         case TIPC_HIGH_IMPORTANCE:
1183         case TIPC_CRITICAL_IMPORTANCE:
1184         case CONN_MANAGER:
1185                 if (tipc_skb_queue_tail(&link->inputq, skb, dport)) {
1186                         node->inputq = &link->inputq;
1187                         node->action_flags |= TIPC_MSG_EVT;
1188                 }
1189                 return true;
1190         case NAME_DISTRIBUTOR:
1191                 node->bclink.recv_permitted = true;
1192                 node->namedq = &link->namedq;
1193                 skb_queue_tail(&link->namedq, skb);
1194                 if (skb_queue_len(&link->namedq) == 1)
1195                         node->action_flags |= TIPC_NAMED_MSG_EVT;
1196                 return true;
1197         case MSG_BUNDLER:
1198         case CHANGEOVER_PROTOCOL:
1199         case MSG_FRAGMENTER:
1200         case BCAST_PROTOCOL:
1201                 return false;
1202         default:
1203                 pr_warn("Dropping received illegal msg type\n");
1204                 kfree_skb(skb);
1205                 return false;
1206         };
1207 }
1208
1209 /* tipc_link_input - process packet that has passed link protocol check
1210  *
1211  * Consumes buffer
1212  * Node lock must be held
1213  */
1214 static void tipc_link_input(struct tipc_link *link, struct sk_buff *skb)
1215 {
1216         struct tipc_node *node = link->owner;
1217         struct tipc_msg *msg = buf_msg(skb);
1218         struct sk_buff *iskb;
1219         int pos = 0;
1220
1221         if (likely(tipc_data_input(link, skb)))
1222                 return;
1223
1224         switch (msg_user(msg)) {
1225         case CHANGEOVER_PROTOCOL:
1226                 if (!tipc_link_tunnel_rcv(node, &skb))
1227                         break;
1228                 if (msg_user(buf_msg(skb)) != MSG_BUNDLER) {
1229                         tipc_data_input(link, skb);
1230                         break;
1231                 }
1232         case MSG_BUNDLER:
1233                 link->stats.recv_bundles++;
1234                 link->stats.recv_bundled += msg_msgcnt(msg);
1235
1236                 while (tipc_msg_extract(skb, &iskb, &pos))
1237                         tipc_data_input(link, iskb);
1238                 break;
1239         case MSG_FRAGMENTER:
1240                 link->stats.recv_fragments++;
1241                 if (tipc_buf_append(&link->reasm_buf, &skb)) {
1242                         link->stats.recv_fragmented++;
1243                         tipc_data_input(link, skb);
1244                 } else if (!link->reasm_buf) {
1245                         tipc_link_reset(link);
1246                 }
1247                 break;
1248         case BCAST_PROTOCOL:
1249                 tipc_link_sync_rcv(node, skb);
1250                 break;
1251         default:
1252                 break;
1253         };
1254 }
1255
1256 /**
1257  * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1258  *
1259  * Returns increase in queue length (i.e. 0 or 1)
1260  */
1261 u32 tipc_link_defer_pkt(struct sk_buff_head *list, struct sk_buff *skb)
1262 {
1263         struct sk_buff *skb1;
1264         u32 seq_no = buf_seqno(skb);
1265
1266         /* Empty queue ? */
1267         if (skb_queue_empty(list)) {
1268                 __skb_queue_tail(list, skb);
1269                 return 1;
1270         }
1271
1272         /* Last ? */
1273         if (less(buf_seqno(skb_peek_tail(list)), seq_no)) {
1274                 __skb_queue_tail(list, skb);
1275                 return 1;
1276         }
1277
1278         /* Locate insertion point in queue, then insert; discard if duplicate */
1279         skb_queue_walk(list, skb1) {
1280                 u32 curr_seqno = buf_seqno(skb1);
1281
1282                 if (seq_no == curr_seqno) {
1283                         kfree_skb(skb);
1284                         return 0;
1285                 }
1286
1287                 if (less(seq_no, curr_seqno))
1288                         break;
1289         }
1290
1291         __skb_queue_before(list, skb1, skb);
1292         return 1;
1293 }
1294
1295 /*
1296  * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
1297  */
1298 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
1299                                        struct sk_buff *buf)
1300 {
1301         u32 seq_no = buf_seqno(buf);
1302
1303         if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
1304                 tipc_link_proto_rcv(l_ptr, buf);
1305                 return;
1306         }
1307
1308         /* Record OOS packet arrival (force mismatch on next timeout) */
1309         l_ptr->checkpoint--;
1310
1311         /*
1312          * Discard packet if a duplicate; otherwise add it to deferred queue
1313          * and notify peer of gap as per protocol specification
1314          */
1315         if (less(seq_no, mod(l_ptr->next_in_no))) {
1316                 l_ptr->stats.duplicates++;
1317                 kfree_skb(buf);
1318                 return;
1319         }
1320
1321         if (tipc_link_defer_pkt(&l_ptr->deferdq, buf)) {
1322                 l_ptr->stats.deferred_recv++;
1323                 if ((skb_queue_len(&l_ptr->deferdq) % TIPC_MIN_LINK_WIN) == 1)
1324                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1325         } else {
1326                 l_ptr->stats.duplicates++;
1327         }
1328 }
1329
1330 /*
1331  * Send protocol message to the other endpoint.
1332  */
1333 void tipc_link_proto_xmit(struct tipc_link *l_ptr, u32 msg_typ, int probe_msg,
1334                           u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
1335 {
1336         struct sk_buff *buf = NULL;
1337         struct tipc_msg *msg = l_ptr->pmsg;
1338         u32 msg_size = sizeof(l_ptr->proto_msg);
1339         int r_flag;
1340
1341         /* Don't send protocol message during link changeover */
1342         if (l_ptr->exp_msg_count)
1343                 return;
1344
1345         /* Abort non-RESET send if communication with node is prohibited */
1346         if ((tipc_node_blocked(l_ptr->owner)) && (msg_typ != RESET_MSG))
1347                 return;
1348
1349         /* Create protocol message with "out-of-sequence" sequence number */
1350         msg_set_type(msg, msg_typ);
1351         msg_set_net_plane(msg, l_ptr->net_plane);
1352         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1353         msg_set_last_bcast(msg, tipc_bclink_get_last_sent(l_ptr->owner->net));
1354
1355         if (msg_typ == STATE_MSG) {
1356                 u32 next_sent = mod(l_ptr->next_out_no);
1357
1358                 if (!tipc_link_is_up(l_ptr))
1359                         return;
1360                 if (skb_queue_len(&l_ptr->backlogq))
1361                         next_sent = buf_seqno(skb_peek(&l_ptr->backlogq));
1362                 msg_set_next_sent(msg, next_sent);
1363                 if (!skb_queue_empty(&l_ptr->deferdq)) {
1364                         u32 rec = buf_seqno(skb_peek(&l_ptr->deferdq));
1365                         gap = mod(rec - mod(l_ptr->next_in_no));
1366                 }
1367                 msg_set_seq_gap(msg, gap);
1368                 if (gap)
1369                         l_ptr->stats.sent_nacks++;
1370                 msg_set_link_tolerance(msg, tolerance);
1371                 msg_set_linkprio(msg, priority);
1372                 msg_set_max_pkt(msg, ack_mtu);
1373                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1374                 msg_set_probe(msg, probe_msg != 0);
1375                 if (probe_msg) {
1376                         u32 mtu = l_ptr->max_pkt;
1377
1378                         if ((mtu < l_ptr->max_pkt_target) &&
1379                             link_working_working(l_ptr) &&
1380                             l_ptr->fsm_msg_cnt) {
1381                                 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1382                                 if (l_ptr->max_pkt_probes == 10) {
1383                                         l_ptr->max_pkt_target = (msg_size - 4);
1384                                         l_ptr->max_pkt_probes = 0;
1385                                         msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1386                                 }
1387                                 l_ptr->max_pkt_probes++;
1388                         }
1389
1390                         l_ptr->stats.sent_probes++;
1391                 }
1392                 l_ptr->stats.sent_states++;
1393         } else {                /* RESET_MSG or ACTIVATE_MSG */
1394                 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
1395                 msg_set_seq_gap(msg, 0);
1396                 msg_set_next_sent(msg, 1);
1397                 msg_set_probe(msg, 0);
1398                 msg_set_link_tolerance(msg, l_ptr->tolerance);
1399                 msg_set_linkprio(msg, l_ptr->priority);
1400                 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
1401         }
1402
1403         r_flag = (l_ptr->owner->working_links > tipc_link_is_up(l_ptr));
1404         msg_set_redundant_link(msg, r_flag);
1405         msg_set_linkprio(msg, l_ptr->priority);
1406         msg_set_size(msg, msg_size);
1407
1408         msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
1409
1410         buf = tipc_buf_acquire(msg_size);
1411         if (!buf)
1412                 return;
1413
1414         skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1415         buf->priority = TC_PRIO_CONTROL;
1416         tipc_bearer_send(l_ptr->owner->net, l_ptr->bearer_id, buf,
1417                          &l_ptr->media_addr);
1418         l_ptr->rcv_unacked = 0;
1419         kfree_skb(buf);
1420 }
1421
1422 /*
1423  * Receive protocol message :
1424  * Note that network plane id propagates through the network, and may
1425  * change at any time. The node with lowest address rules
1426  */
1427 static void tipc_link_proto_rcv(struct tipc_link *l_ptr,
1428                                 struct sk_buff *buf)
1429 {
1430         u32 rec_gap = 0;
1431         u32 max_pkt_info;
1432         u32 max_pkt_ack;
1433         u32 msg_tol;
1434         struct tipc_msg *msg = buf_msg(buf);
1435
1436         /* Discard protocol message during link changeover */
1437         if (l_ptr->exp_msg_count)
1438                 goto exit;
1439
1440         if (l_ptr->net_plane != msg_net_plane(msg))
1441                 if (link_own_addr(l_ptr) > msg_prevnode(msg))
1442                         l_ptr->net_plane = msg_net_plane(msg);
1443
1444         switch (msg_type(msg)) {
1445
1446         case RESET_MSG:
1447                 if (!link_working_unknown(l_ptr) &&
1448                     (l_ptr->peer_session != INVALID_SESSION)) {
1449                         if (less_eq(msg_session(msg), l_ptr->peer_session))
1450                                 break; /* duplicate or old reset: ignore */
1451                 }
1452
1453                 if (!msg_redundant_link(msg) && (link_working_working(l_ptr) ||
1454                                 link_working_unknown(l_ptr))) {
1455                         /*
1456                          * peer has lost contact -- don't allow peer's links
1457                          * to reactivate before we recognize loss & clean up
1458                          */
1459                         l_ptr->owner->action_flags |= TIPC_WAIT_OWN_LINKS_DOWN;
1460                 }
1461
1462                 link_state_event(l_ptr, RESET_MSG);
1463
1464                 /* fall thru' */
1465         case ACTIVATE_MSG:
1466                 /* Update link settings according other endpoint's values */
1467                 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
1468
1469                 msg_tol = msg_link_tolerance(msg);
1470                 if (msg_tol > l_ptr->tolerance)
1471                         link_set_supervision_props(l_ptr, msg_tol);
1472
1473                 if (msg_linkprio(msg) > l_ptr->priority)
1474                         l_ptr->priority = msg_linkprio(msg);
1475
1476                 max_pkt_info = msg_max_pkt(msg);
1477                 if (max_pkt_info) {
1478                         if (max_pkt_info < l_ptr->max_pkt_target)
1479                                 l_ptr->max_pkt_target = max_pkt_info;
1480                         if (l_ptr->max_pkt > l_ptr->max_pkt_target)
1481                                 l_ptr->max_pkt = l_ptr->max_pkt_target;
1482                 } else {
1483                         l_ptr->max_pkt = l_ptr->max_pkt_target;
1484                 }
1485
1486                 /* Synchronize broadcast link info, if not done previously */
1487                 if (!tipc_node_is_up(l_ptr->owner)) {
1488                         l_ptr->owner->bclink.last_sent =
1489                                 l_ptr->owner->bclink.last_in =
1490                                 msg_last_bcast(msg);
1491                         l_ptr->owner->bclink.oos_state = 0;
1492                 }
1493
1494                 l_ptr->peer_session = msg_session(msg);
1495                 l_ptr->peer_bearer_id = msg_bearer_id(msg);
1496
1497                 if (msg_type(msg) == ACTIVATE_MSG)
1498                         link_state_event(l_ptr, ACTIVATE_MSG);
1499                 break;
1500         case STATE_MSG:
1501
1502                 msg_tol = msg_link_tolerance(msg);
1503                 if (msg_tol)
1504                         link_set_supervision_props(l_ptr, msg_tol);
1505
1506                 if (msg_linkprio(msg) &&
1507                     (msg_linkprio(msg) != l_ptr->priority)) {
1508                         pr_debug("%s<%s>, priority change %u->%u\n",
1509                                  link_rst_msg, l_ptr->name,
1510                                  l_ptr->priority, msg_linkprio(msg));
1511                         l_ptr->priority = msg_linkprio(msg);
1512                         tipc_link_reset(l_ptr); /* Enforce change to take effect */
1513                         break;
1514                 }
1515
1516                 /* Record reception; force mismatch at next timeout: */
1517                 l_ptr->checkpoint--;
1518
1519                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1520                 l_ptr->stats.recv_states++;
1521                 if (link_reset_unknown(l_ptr))
1522                         break;
1523
1524                 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
1525                         rec_gap = mod(msg_next_sent(msg) -
1526                                       mod(l_ptr->next_in_no));
1527                 }
1528
1529                 max_pkt_ack = msg_max_pkt(msg);
1530                 if (max_pkt_ack > l_ptr->max_pkt) {
1531                         l_ptr->max_pkt = max_pkt_ack;
1532                         l_ptr->max_pkt_probes = 0;
1533                 }
1534
1535                 max_pkt_ack = 0;
1536                 if (msg_probe(msg)) {
1537                         l_ptr->stats.recv_probes++;
1538                         if (msg_size(msg) > sizeof(l_ptr->proto_msg))
1539                                 max_pkt_ack = msg_size(msg);
1540                 }
1541
1542                 /* Protocol message before retransmits, reduce loss risk */
1543                 if (l_ptr->owner->bclink.recv_permitted)
1544                         tipc_bclink_update_link_state(l_ptr->owner,
1545                                                       msg_last_bcast(msg));
1546
1547                 if (rec_gap || (msg_probe(msg))) {
1548                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, rec_gap, 0,
1549                                              0, max_pkt_ack);
1550                 }
1551                 if (msg_seq_gap(msg)) {
1552                         l_ptr->stats.recv_nacks++;
1553                         tipc_link_retransmit(l_ptr, skb_peek(&l_ptr->transmq),
1554                                              msg_seq_gap(msg));
1555                 }
1556                 break;
1557         }
1558 exit:
1559         kfree_skb(buf);
1560 }
1561
1562
1563 /* tipc_link_tunnel_xmit(): Tunnel one packet via a link belonging to
1564  * a different bearer. Owner node is locked.
1565  */
1566 static void tipc_link_tunnel_xmit(struct tipc_link *l_ptr,
1567                                   struct tipc_msg *tunnel_hdr,
1568                                   struct tipc_msg *msg,
1569                                   u32 selector)
1570 {
1571         struct tipc_link *tunnel;
1572         struct sk_buff *skb;
1573         u32 length = msg_size(msg);
1574
1575         tunnel = l_ptr->owner->active_links[selector & 1];
1576         if (!tipc_link_is_up(tunnel)) {
1577                 pr_warn("%stunnel link no longer available\n", link_co_err);
1578                 return;
1579         }
1580         msg_set_size(tunnel_hdr, length + INT_H_SIZE);
1581         skb = tipc_buf_acquire(length + INT_H_SIZE);
1582         if (!skb) {
1583                 pr_warn("%sunable to send tunnel msg\n", link_co_err);
1584                 return;
1585         }
1586         skb_copy_to_linear_data(skb, tunnel_hdr, INT_H_SIZE);
1587         skb_copy_to_linear_data_offset(skb, INT_H_SIZE, msg, length);
1588         __tipc_link_xmit_skb(tunnel, skb);
1589 }
1590
1591
1592 /* tipc_link_failover_send_queue(): A link has gone down, but a second
1593  * link is still active. We can do failover. Tunnel the failing link's
1594  * whole send queue via the remaining link. This way, we don't lose
1595  * any packets, and sequence order is preserved for subsequent traffic
1596  * sent over the remaining link. Owner node is locked.
1597  */
1598 void tipc_link_failover_send_queue(struct tipc_link *l_ptr)
1599 {
1600         int msgcount;
1601         struct tipc_link *tunnel = l_ptr->owner->active_links[0];
1602         struct tipc_msg tunnel_hdr;
1603         struct sk_buff *skb;
1604         int split_bundles;
1605
1606         if (!tunnel)
1607                 return;
1608
1609         tipc_msg_init(link_own_addr(l_ptr), &tunnel_hdr, CHANGEOVER_PROTOCOL,
1610                       ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
1611         skb_queue_splice_tail_init(&l_ptr->backlogq, &l_ptr->transmq);
1612         msgcount = skb_queue_len(&l_ptr->transmq);
1613         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
1614         msg_set_msgcnt(&tunnel_hdr, msgcount);
1615
1616         if (skb_queue_empty(&l_ptr->transmq)) {
1617                 skb = tipc_buf_acquire(INT_H_SIZE);
1618                 if (skb) {
1619                         skb_copy_to_linear_data(skb, &tunnel_hdr, INT_H_SIZE);
1620                         msg_set_size(&tunnel_hdr, INT_H_SIZE);
1621                         __tipc_link_xmit_skb(tunnel, skb);
1622                 } else {
1623                         pr_warn("%sunable to send changeover msg\n",
1624                                 link_co_err);
1625                 }
1626                 return;
1627         }
1628
1629         split_bundles = (l_ptr->owner->active_links[0] !=
1630                          l_ptr->owner->active_links[1]);
1631
1632         skb_queue_walk(&l_ptr->transmq, skb) {
1633                 struct tipc_msg *msg = buf_msg(skb);
1634
1635                 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
1636                         struct tipc_msg *m = msg_get_wrapped(msg);
1637                         unchar *pos = (unchar *)m;
1638
1639                         msgcount = msg_msgcnt(msg);
1640                         while (msgcount--) {
1641                                 msg_set_seqno(m, msg_seqno(msg));
1642                                 tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, m,
1643                                                       msg_link_selector(m));
1644                                 pos += align(msg_size(m));
1645                                 m = (struct tipc_msg *)pos;
1646                         }
1647                 } else {
1648                         tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, msg,
1649                                               msg_link_selector(msg));
1650                 }
1651         }
1652 }
1653
1654 /* tipc_link_dup_queue_xmit(): A second link has become active. Tunnel a
1655  * duplicate of the first link's send queue via the new link. This way, we
1656  * are guaranteed that currently queued packets from a socket are delivered
1657  * before future traffic from the same socket, even if this is using the
1658  * new link. The last arriving copy of each duplicate packet is dropped at
1659  * the receiving end by the regular protocol check, so packet cardinality
1660  * and sequence order is preserved per sender/receiver socket pair.
1661  * Owner node is locked.
1662  */
1663 void tipc_link_dup_queue_xmit(struct tipc_link *link,
1664                               struct tipc_link *tnl)
1665 {
1666         struct sk_buff *skb;
1667         struct tipc_msg tnl_hdr;
1668         struct sk_buff_head *queue = &link->transmq;
1669         int mcnt;
1670
1671         tipc_msg_init(link_own_addr(link), &tnl_hdr, CHANGEOVER_PROTOCOL,
1672                       DUPLICATE_MSG, INT_H_SIZE, link->addr);
1673         mcnt = skb_queue_len(&link->transmq) + skb_queue_len(&link->backlogq);
1674         msg_set_msgcnt(&tnl_hdr, mcnt);
1675         msg_set_bearer_id(&tnl_hdr, link->peer_bearer_id);
1676
1677 tunnel_queue:
1678         skb_queue_walk(queue, skb) {
1679                 struct sk_buff *outskb;
1680                 struct tipc_msg *msg = buf_msg(skb);
1681                 u32 len = msg_size(msg);
1682
1683                 msg_set_ack(msg, mod(link->next_in_no - 1));
1684                 msg_set_bcast_ack(msg, link->owner->bclink.last_in);
1685                 msg_set_size(&tnl_hdr, len + INT_H_SIZE);
1686                 outskb = tipc_buf_acquire(len + INT_H_SIZE);
1687                 if (outskb == NULL) {
1688                         pr_warn("%sunable to send duplicate msg\n",
1689                                 link_co_err);
1690                         return;
1691                 }
1692                 skb_copy_to_linear_data(outskb, &tnl_hdr, INT_H_SIZE);
1693                 skb_copy_to_linear_data_offset(outskb, INT_H_SIZE,
1694                                                skb->data, len);
1695                 __tipc_link_xmit_skb(tnl, outskb);
1696                 if (!tipc_link_is_up(link))
1697                         return;
1698         }
1699         if (queue == &link->backlogq)
1700                 return;
1701         queue = &link->backlogq;
1702         goto tunnel_queue;
1703 }
1704
1705 /* tipc_link_dup_rcv(): Receive a tunnelled DUPLICATE_MSG packet.
1706  * Owner node is locked.
1707  */
1708 static void tipc_link_dup_rcv(struct tipc_link *link,
1709                               struct sk_buff *skb)
1710 {
1711         struct sk_buff *iskb;
1712         int pos = 0;
1713
1714         if (!tipc_link_is_up(link))
1715                 return;
1716
1717         if (!tipc_msg_extract(skb, &iskb, &pos)) {
1718                 pr_warn("%sfailed to extract inner dup pkt\n", link_co_err);
1719                 return;
1720         }
1721         /* Append buffer to deferred queue, if applicable: */
1722         link_handle_out_of_seq_msg(link, iskb);
1723 }
1724
1725 /*  tipc_link_failover_rcv(): Receive a tunnelled ORIGINAL_MSG packet
1726  *  Owner node is locked.
1727  */
1728 static struct sk_buff *tipc_link_failover_rcv(struct tipc_link *l_ptr,
1729                                               struct sk_buff *t_buf)
1730 {
1731         struct tipc_msg *t_msg = buf_msg(t_buf);
1732         struct sk_buff *buf = NULL;
1733         struct tipc_msg *msg;
1734         int pos = 0;
1735
1736         if (tipc_link_is_up(l_ptr))
1737                 tipc_link_reset(l_ptr);
1738
1739         /* First failover packet? */
1740         if (l_ptr->exp_msg_count == START_CHANGEOVER)
1741                 l_ptr->exp_msg_count = msg_msgcnt(t_msg);
1742
1743         /* Should there be an inner packet? */
1744         if (l_ptr->exp_msg_count) {
1745                 l_ptr->exp_msg_count--;
1746                 if (!tipc_msg_extract(t_buf, &buf, &pos)) {
1747                         pr_warn("%sno inner failover pkt\n", link_co_err);
1748                         goto exit;
1749                 }
1750                 msg = buf_msg(buf);
1751
1752                 if (less(msg_seqno(msg), l_ptr->reset_checkpoint)) {
1753                         kfree_skb(buf);
1754                         buf = NULL;
1755                         goto exit;
1756                 }
1757                 if (msg_user(msg) == MSG_FRAGMENTER) {
1758                         l_ptr->stats.recv_fragments++;
1759                         tipc_buf_append(&l_ptr->reasm_buf, &buf);
1760                 }
1761         }
1762 exit:
1763         if ((!l_ptr->exp_msg_count) && (l_ptr->flags & LINK_STOPPED))
1764                 tipc_link_delete(l_ptr);
1765         return buf;
1766 }
1767
1768 /*  tipc_link_tunnel_rcv(): Receive a tunnelled packet, sent
1769  *  via other link as result of a failover (ORIGINAL_MSG) or
1770  *  a new active link (DUPLICATE_MSG). Failover packets are
1771  *  returned to the active link for delivery upwards.
1772  *  Owner node is locked.
1773  */
1774 static int tipc_link_tunnel_rcv(struct tipc_node *n_ptr,
1775                                 struct sk_buff **buf)
1776 {
1777         struct sk_buff *t_buf = *buf;
1778         struct tipc_link *l_ptr;
1779         struct tipc_msg *t_msg = buf_msg(t_buf);
1780         u32 bearer_id = msg_bearer_id(t_msg);
1781
1782         *buf = NULL;
1783
1784         if (bearer_id >= MAX_BEARERS)
1785                 goto exit;
1786
1787         l_ptr = n_ptr->links[bearer_id];
1788         if (!l_ptr)
1789                 goto exit;
1790
1791         if (msg_type(t_msg) == DUPLICATE_MSG)
1792                 tipc_link_dup_rcv(l_ptr, t_buf);
1793         else if (msg_type(t_msg) == ORIGINAL_MSG)
1794                 *buf = tipc_link_failover_rcv(l_ptr, t_buf);
1795         else
1796                 pr_warn("%sunknown tunnel pkt received\n", link_co_err);
1797 exit:
1798         kfree_skb(t_buf);
1799         return *buf != NULL;
1800 }
1801
1802 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tol)
1803 {
1804         unsigned long intv = ((tol / 4) > 500) ? 500 : tol / 4;
1805
1806         if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1807                 return;
1808
1809         l_ptr->tolerance = tol;
1810         l_ptr->cont_intv = msecs_to_jiffies(intv);
1811         l_ptr->abort_limit = tol / (jiffies_to_msecs(l_ptr->cont_intv) / 4);
1812 }
1813
1814 void tipc_link_set_queue_limits(struct tipc_link *l_ptr, u32 window)
1815 {
1816         l_ptr->window = window;
1817
1818         /* Data messages from this node, inclusive FIRST_FRAGM */
1819         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
1820         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
1821         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
1822         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
1823         /* Transiting data messages,inclusive FIRST_FRAGM */
1824         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
1825         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
1826         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
1827         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
1828         l_ptr->queue_limit[CONN_MANAGER] = 1200;
1829         l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
1830         l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
1831         /* FRAGMENT and LAST_FRAGMENT packets */
1832         l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
1833 }
1834
1835 /* tipc_link_find_owner - locate owner node of link by link's name
1836  * @net: the applicable net namespace
1837  * @name: pointer to link name string
1838  * @bearer_id: pointer to index in 'node->links' array where the link was found.
1839  *
1840  * Returns pointer to node owning the link, or 0 if no matching link is found.
1841  */
1842 static struct tipc_node *tipc_link_find_owner(struct net *net,
1843                                               const char *link_name,
1844                                               unsigned int *bearer_id)
1845 {
1846         struct tipc_net *tn = net_generic(net, tipc_net_id);
1847         struct tipc_link *l_ptr;
1848         struct tipc_node *n_ptr;
1849         struct tipc_node *found_node = NULL;
1850         int i;
1851
1852         *bearer_id = 0;
1853         rcu_read_lock();
1854         list_for_each_entry_rcu(n_ptr, &tn->node_list, list) {
1855                 tipc_node_lock(n_ptr);
1856                 for (i = 0; i < MAX_BEARERS; i++) {
1857                         l_ptr = n_ptr->links[i];
1858                         if (l_ptr && !strcmp(l_ptr->name, link_name)) {
1859                                 *bearer_id = i;
1860                                 found_node = n_ptr;
1861                                 break;
1862                         }
1863                 }
1864                 tipc_node_unlock(n_ptr);
1865                 if (found_node)
1866                         break;
1867         }
1868         rcu_read_unlock();
1869
1870         return found_node;
1871 }
1872
1873 /**
1874  * link_reset_statistics - reset link statistics
1875  * @l_ptr: pointer to link
1876  */
1877 static void link_reset_statistics(struct tipc_link *l_ptr)
1878 {
1879         memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
1880         l_ptr->stats.sent_info = l_ptr->next_out_no;
1881         l_ptr->stats.recv_info = l_ptr->next_in_no;
1882 }
1883
1884 static void link_print(struct tipc_link *l_ptr, const char *str)
1885 {
1886         struct tipc_net *tn = net_generic(l_ptr->owner->net, tipc_net_id);
1887         struct tipc_bearer *b_ptr;
1888
1889         rcu_read_lock();
1890         b_ptr = rcu_dereference_rtnl(tn->bearer_list[l_ptr->bearer_id]);
1891         if (b_ptr)
1892                 pr_info("%s Link %x<%s>:", str, l_ptr->addr, b_ptr->name);
1893         rcu_read_unlock();
1894
1895         if (link_working_unknown(l_ptr))
1896                 pr_cont(":WU\n");
1897         else if (link_reset_reset(l_ptr))
1898                 pr_cont(":RR\n");
1899         else if (link_reset_unknown(l_ptr))
1900                 pr_cont(":RU\n");
1901         else if (link_working_working(l_ptr))
1902                 pr_cont(":WW\n");
1903         else
1904                 pr_cont("\n");
1905 }
1906
1907 /* Parse and validate nested (link) properties valid for media, bearer and link
1908  */
1909 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1910 {
1911         int err;
1912
1913         err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1914                                tipc_nl_prop_policy);
1915         if (err)
1916                 return err;
1917
1918         if (props[TIPC_NLA_PROP_PRIO]) {
1919                 u32 prio;
1920
1921                 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1922                 if (prio > TIPC_MAX_LINK_PRI)
1923                         return -EINVAL;
1924         }
1925
1926         if (props[TIPC_NLA_PROP_TOL]) {
1927                 u32 tol;
1928
1929                 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1930                 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1931                         return -EINVAL;
1932         }
1933
1934         if (props[TIPC_NLA_PROP_WIN]) {
1935                 u32 win;
1936
1937                 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1938                 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1939                         return -EINVAL;
1940         }
1941
1942         return 0;
1943 }
1944
1945 int tipc_nl_link_set(struct sk_buff *skb, struct genl_info *info)
1946 {
1947         int err;
1948         int res = 0;
1949         int bearer_id;
1950         char *name;
1951         struct tipc_link *link;
1952         struct tipc_node *node;
1953         struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
1954         struct net *net = sock_net(skb->sk);
1955
1956         if (!info->attrs[TIPC_NLA_LINK])
1957                 return -EINVAL;
1958
1959         err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
1960                                info->attrs[TIPC_NLA_LINK],
1961                                tipc_nl_link_policy);
1962         if (err)
1963                 return err;
1964
1965         if (!attrs[TIPC_NLA_LINK_NAME])
1966                 return -EINVAL;
1967
1968         name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
1969
1970         node = tipc_link_find_owner(net, name, &bearer_id);
1971         if (!node)
1972                 return -EINVAL;
1973
1974         tipc_node_lock(node);
1975
1976         link = node->links[bearer_id];
1977         if (!link) {
1978                 res = -EINVAL;
1979                 goto out;
1980         }
1981
1982         if (attrs[TIPC_NLA_LINK_PROP]) {
1983                 struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
1984
1985                 err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP],
1986                                               props);
1987                 if (err) {
1988                         res = err;
1989                         goto out;
1990                 }
1991
1992                 if (props[TIPC_NLA_PROP_TOL]) {
1993                         u32 tol;
1994
1995                         tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1996                         link_set_supervision_props(link, tol);
1997                         tipc_link_proto_xmit(link, STATE_MSG, 0, 0, tol, 0, 0);
1998                 }
1999                 if (props[TIPC_NLA_PROP_PRIO]) {
2000                         u32 prio;
2001
2002                         prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2003                         link->priority = prio;
2004                         tipc_link_proto_xmit(link, STATE_MSG, 0, 0, 0, prio, 0);
2005                 }
2006                 if (props[TIPC_NLA_PROP_WIN]) {
2007                         u32 win;
2008
2009                         win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2010                         tipc_link_set_queue_limits(link, win);
2011                 }
2012         }
2013
2014 out:
2015         tipc_node_unlock(node);
2016
2017         return res;
2018 }
2019
2020 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2021 {
2022         int i;
2023         struct nlattr *stats;
2024
2025         struct nla_map {
2026                 u32 key;
2027                 u32 val;
2028         };
2029
2030         struct nla_map map[] = {
2031                 {TIPC_NLA_STATS_RX_INFO, s->recv_info},
2032                 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2033                 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2034                 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2035                 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2036                 {TIPC_NLA_STATS_TX_INFO, s->sent_info},
2037                 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2038                 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2039                 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2040                 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2041                 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2042                         s->msg_length_counts : 1},
2043                 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2044                 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2045                 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2046                 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2047                 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2048                 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2049                 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2050                 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2051                 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2052                 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
2053                 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2054                 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2055                 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2056                 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
2057                 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2058                 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2059                 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2060                 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2061                 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2062                 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2063                 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2064                 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2065                         (s->accu_queue_sz / s->queue_sz_counts) : 0}
2066         };
2067
2068         stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
2069         if (!stats)
2070                 return -EMSGSIZE;
2071
2072         for (i = 0; i <  ARRAY_SIZE(map); i++)
2073                 if (nla_put_u32(skb, map[i].key, map[i].val))
2074                         goto msg_full;
2075
2076         nla_nest_end(skb, stats);
2077
2078         return 0;
2079 msg_full:
2080         nla_nest_cancel(skb, stats);
2081
2082         return -EMSGSIZE;
2083 }
2084
2085 /* Caller should hold appropriate locks to protect the link */
2086 static int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2087                               struct tipc_link *link)
2088 {
2089         int err;
2090         void *hdr;
2091         struct nlattr *attrs;
2092         struct nlattr *prop;
2093         struct tipc_net *tn = net_generic(net, tipc_net_id);
2094
2095         hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2096                           NLM_F_MULTI, TIPC_NL_LINK_GET);
2097         if (!hdr)
2098                 return -EMSGSIZE;
2099
2100         attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
2101         if (!attrs)
2102                 goto msg_full;
2103
2104         if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2105                 goto attr_msg_full;
2106         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
2107                         tipc_cluster_mask(tn->own_addr)))
2108                 goto attr_msg_full;
2109         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->max_pkt))
2110                 goto attr_msg_full;
2111         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->next_in_no))
2112                 goto attr_msg_full;
2113         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->next_out_no))
2114                 goto attr_msg_full;
2115
2116         if (tipc_link_is_up(link))
2117                 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2118                         goto attr_msg_full;
2119         if (tipc_link_is_active(link))
2120                 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2121                         goto attr_msg_full;
2122
2123         prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2124         if (!prop)
2125                 goto attr_msg_full;
2126         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2127                 goto prop_msg_full;
2128         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2129                 goto prop_msg_full;
2130         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2131                         link->queue_limit[TIPC_LOW_IMPORTANCE]))
2132                 goto prop_msg_full;
2133         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2134                 goto prop_msg_full;
2135         nla_nest_end(msg->skb, prop);
2136
2137         err = __tipc_nl_add_stats(msg->skb, &link->stats);
2138         if (err)
2139                 goto attr_msg_full;
2140
2141         nla_nest_end(msg->skb, attrs);
2142         genlmsg_end(msg->skb, hdr);
2143
2144         return 0;
2145
2146 prop_msg_full:
2147         nla_nest_cancel(msg->skb, prop);
2148 attr_msg_full:
2149         nla_nest_cancel(msg->skb, attrs);
2150 msg_full:
2151         genlmsg_cancel(msg->skb, hdr);
2152
2153         return -EMSGSIZE;
2154 }
2155
2156 /* Caller should hold node lock  */
2157 static int __tipc_nl_add_node_links(struct net *net, struct tipc_nl_msg *msg,
2158                                     struct tipc_node *node, u32 *prev_link)
2159 {
2160         u32 i;
2161         int err;
2162
2163         for (i = *prev_link; i < MAX_BEARERS; i++) {
2164                 *prev_link = i;
2165
2166                 if (!node->links[i])
2167                         continue;
2168
2169                 err = __tipc_nl_add_link(net, msg, node->links[i]);
2170                 if (err)
2171                         return err;
2172         }
2173         *prev_link = 0;
2174
2175         return 0;
2176 }
2177
2178 int tipc_nl_link_dump(struct sk_buff *skb, struct netlink_callback *cb)
2179 {
2180         struct net *net = sock_net(skb->sk);
2181         struct tipc_net *tn = net_generic(net, tipc_net_id);
2182         struct tipc_node *node;
2183         struct tipc_nl_msg msg;
2184         u32 prev_node = cb->args[0];
2185         u32 prev_link = cb->args[1];
2186         int done = cb->args[2];
2187         int err;
2188
2189         if (done)
2190                 return 0;
2191
2192         msg.skb = skb;
2193         msg.portid = NETLINK_CB(cb->skb).portid;
2194         msg.seq = cb->nlh->nlmsg_seq;
2195
2196         rcu_read_lock();
2197
2198         if (prev_node) {
2199                 node = tipc_node_find(net, prev_node);
2200                 if (!node) {
2201                         /* We never set seq or call nl_dump_check_consistent()
2202                          * this means that setting prev_seq here will cause the
2203                          * consistence check to fail in the netlink callback
2204                          * handler. Resulting in the last NLMSG_DONE message
2205                          * having the NLM_F_DUMP_INTR flag set.
2206                          */
2207                         cb->prev_seq = 1;
2208                         goto out;
2209                 }
2210
2211                 list_for_each_entry_continue_rcu(node, &tn->node_list,
2212                                                  list) {
2213                         tipc_node_lock(node);
2214                         err = __tipc_nl_add_node_links(net, &msg, node,
2215                                                        &prev_link);
2216                         tipc_node_unlock(node);
2217                         if (err)
2218                                 goto out;
2219
2220                         prev_node = node->addr;
2221                 }
2222         } else {
2223                 err = tipc_nl_add_bc_link(net, &msg);
2224                 if (err)
2225                         goto out;
2226
2227                 list_for_each_entry_rcu(node, &tn->node_list, list) {
2228                         tipc_node_lock(node);
2229                         err = __tipc_nl_add_node_links(net, &msg, node,
2230                                                        &prev_link);
2231                         tipc_node_unlock(node);
2232                         if (err)
2233                                 goto out;
2234
2235                         prev_node = node->addr;
2236                 }
2237         }
2238         done = 1;
2239 out:
2240         rcu_read_unlock();
2241
2242         cb->args[0] = prev_node;
2243         cb->args[1] = prev_link;
2244         cb->args[2] = done;
2245
2246         return skb->len;
2247 }
2248
2249 int tipc_nl_link_get(struct sk_buff *skb, struct genl_info *info)
2250 {
2251         struct net *net = genl_info_net(info);
2252         struct sk_buff *ans_skb;
2253         struct tipc_nl_msg msg;
2254         struct tipc_link *link;
2255         struct tipc_node *node;
2256         char *name;
2257         int bearer_id;
2258         int err;
2259
2260         if (!info->attrs[TIPC_NLA_LINK_NAME])
2261                 return -EINVAL;
2262
2263         name = nla_data(info->attrs[TIPC_NLA_LINK_NAME]);
2264         node = tipc_link_find_owner(net, name, &bearer_id);
2265         if (!node)
2266                 return -EINVAL;
2267
2268         ans_skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
2269         if (!ans_skb)
2270                 return -ENOMEM;
2271
2272         msg.skb = ans_skb;
2273         msg.portid = info->snd_portid;
2274         msg.seq = info->snd_seq;
2275
2276         tipc_node_lock(node);
2277         link = node->links[bearer_id];
2278         if (!link) {
2279                 err = -EINVAL;
2280                 goto err_out;
2281         }
2282
2283         err = __tipc_nl_add_link(net, &msg, link);
2284         if (err)
2285                 goto err_out;
2286
2287         tipc_node_unlock(node);
2288
2289         return genlmsg_reply(ans_skb, info);
2290
2291 err_out:
2292         tipc_node_unlock(node);
2293         nlmsg_free(ans_skb);
2294
2295         return err;
2296 }
2297
2298 int tipc_nl_link_reset_stats(struct sk_buff *skb, struct genl_info *info)
2299 {
2300         int err;
2301         char *link_name;
2302         unsigned int bearer_id;
2303         struct tipc_link *link;
2304         struct tipc_node *node;
2305         struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
2306         struct net *net = sock_net(skb->sk);
2307
2308         if (!info->attrs[TIPC_NLA_LINK])
2309                 return -EINVAL;
2310
2311         err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
2312                                info->attrs[TIPC_NLA_LINK],
2313                                tipc_nl_link_policy);
2314         if (err)
2315                 return err;
2316
2317         if (!attrs[TIPC_NLA_LINK_NAME])
2318                 return -EINVAL;
2319
2320         link_name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
2321
2322         if (strcmp(link_name, tipc_bclink_name) == 0) {
2323                 err = tipc_bclink_reset_stats(net);
2324                 if (err)
2325                         return err;
2326                 return 0;
2327         }
2328
2329         node = tipc_link_find_owner(net, link_name, &bearer_id);
2330         if (!node)
2331                 return -EINVAL;
2332
2333         tipc_node_lock(node);
2334
2335         link = node->links[bearer_id];
2336         if (!link) {
2337                 tipc_node_unlock(node);
2338                 return -EINVAL;
2339         }
2340
2341         link_reset_statistics(link);
2342
2343         tipc_node_unlock(node);
2344
2345         return 0;
2346 }