819fb7163fa239aba605d06433981274af91caf3
[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 "subscr.h"
39 #include "link.h"
40 #include "bcast.h"
41 #include "socket.h"
42 #include "name_distr.h"
43 #include "discover.h"
44 #include "netlink.h"
45
46 #include <linux/pkt_sched.h>
47
48 /*
49  * Error message prefixes
50  */
51 static const char *link_co_err = "Link tunneling error, ";
52 static const char *link_rst_msg = "Resetting link ";
53 static const char tipc_bclink_name[] = "broadcast-link";
54
55 static const struct nla_policy tipc_nl_link_policy[TIPC_NLA_LINK_MAX + 1] = {
56         [TIPC_NLA_LINK_UNSPEC]          = { .type = NLA_UNSPEC },
57         [TIPC_NLA_LINK_NAME] = {
58                 .type = NLA_STRING,
59                 .len = TIPC_MAX_LINK_NAME
60         },
61         [TIPC_NLA_LINK_MTU]             = { .type = NLA_U32 },
62         [TIPC_NLA_LINK_BROADCAST]       = { .type = NLA_FLAG },
63         [TIPC_NLA_LINK_UP]              = { .type = NLA_FLAG },
64         [TIPC_NLA_LINK_ACTIVE]          = { .type = NLA_FLAG },
65         [TIPC_NLA_LINK_PROP]            = { .type = NLA_NESTED },
66         [TIPC_NLA_LINK_STATS]           = { .type = NLA_NESTED },
67         [TIPC_NLA_LINK_RX]              = { .type = NLA_U32 },
68         [TIPC_NLA_LINK_TX]              = { .type = NLA_U32 }
69 };
70
71 /* Properties valid for media, bearar and link */
72 static const struct nla_policy tipc_nl_prop_policy[TIPC_NLA_PROP_MAX + 1] = {
73         [TIPC_NLA_PROP_UNSPEC]          = { .type = NLA_UNSPEC },
74         [TIPC_NLA_PROP_PRIO]            = { .type = NLA_U32 },
75         [TIPC_NLA_PROP_TOL]             = { .type = NLA_U32 },
76         [TIPC_NLA_PROP_WIN]             = { .type = NLA_U32 }
77 };
78
79 /* Send states for broadcast NACKs
80  */
81 enum {
82         BC_NACK_SND_CONDITIONAL,
83         BC_NACK_SND_UNCONDITIONAL,
84         BC_NACK_SND_SUPPRESS,
85 };
86
87 /*
88  * Interval between NACKs when packets arrive out of order
89  */
90 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
91 /*
92  * Out-of-range value for link session numbers
93  */
94 #define WILDCARD_SESSION 0x10000
95
96 /* Link FSM states:
97  */
98 enum {
99         LINK_ESTABLISHED     = 0xe,
100         LINK_ESTABLISHING    = 0xe  << 4,
101         LINK_RESET           = 0x1  << 8,
102         LINK_RESETTING       = 0x2  << 12,
103         LINK_PEER_RESET      = 0xd  << 16,
104         LINK_FAILINGOVER     = 0xf  << 20,
105         LINK_SYNCHING        = 0xc  << 24
106 };
107
108 /* Link FSM state checking routines
109  */
110 static int link_is_up(struct tipc_link *l)
111 {
112         return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
113 }
114
115 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
116                                struct sk_buff_head *xmitq);
117 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
118                                       u16 rcvgap, int tolerance, int priority,
119                                       struct sk_buff_head *xmitq);
120 static void link_reset_statistics(struct tipc_link *l_ptr);
121 static void link_print(struct tipc_link *l_ptr, const char *str);
122 static void tipc_link_build_nack_msg(struct tipc_link *l,
123                                      struct sk_buff_head *xmitq);
124 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
125                                         struct sk_buff_head *xmitq);
126 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
127
128 /*
129  *  Simple non-static link routines (i.e. referenced outside this file)
130  */
131 bool tipc_link_is_up(struct tipc_link *l)
132 {
133         return link_is_up(l);
134 }
135
136 bool tipc_link_peer_is_down(struct tipc_link *l)
137 {
138         return l->state == LINK_PEER_RESET;
139 }
140
141 bool tipc_link_is_reset(struct tipc_link *l)
142 {
143         return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
144 }
145
146 bool tipc_link_is_establishing(struct tipc_link *l)
147 {
148         return l->state == LINK_ESTABLISHING;
149 }
150
151 bool tipc_link_is_synching(struct tipc_link *l)
152 {
153         return l->state == LINK_SYNCHING;
154 }
155
156 bool tipc_link_is_failingover(struct tipc_link *l)
157 {
158         return l->state == LINK_FAILINGOVER;
159 }
160
161 bool tipc_link_is_blocked(struct tipc_link *l)
162 {
163         return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
164 }
165
166 bool link_is_bc_sndlink(struct tipc_link *l)
167 {
168         return !l->bc_sndlink;
169 }
170
171 bool link_is_bc_rcvlink(struct tipc_link *l)
172 {
173         return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
174 }
175
176 int tipc_link_is_active(struct tipc_link *l)
177 {
178         return l->active;
179 }
180
181 void tipc_link_set_active(struct tipc_link *l, bool active)
182 {
183         l->active = active;
184 }
185
186 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
187                            struct tipc_link *uc_l,
188                            struct sk_buff_head *xmitq)
189 {
190         struct tipc_link *rcv_l = uc_l->bc_rcvlink;
191
192         snd_l->ackers++;
193         rcv_l->acked = snd_l->snd_nxt - 1;
194         tipc_link_build_bc_init_msg(uc_l, xmitq);
195 }
196
197 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
198                               struct tipc_link *rcv_l,
199                               struct sk_buff_head *xmitq)
200 {
201         u16 ack = snd_l->snd_nxt - 1;
202
203         snd_l->ackers--;
204         tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
205         tipc_link_reset(rcv_l);
206         rcv_l->state = LINK_RESET;
207         if (!snd_l->ackers) {
208                 tipc_link_reset(snd_l);
209                 __skb_queue_purge(xmitq);
210         }
211 }
212
213 int tipc_link_bc_peers(struct tipc_link *l)
214 {
215         return l->ackers;
216 }
217
218 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
219 {
220         l->mtu = mtu;
221 }
222
223 int tipc_link_mtu(struct tipc_link *l)
224 {
225         return l->mtu;
226 }
227
228 static u32 link_own_addr(struct tipc_link *l)
229 {
230         return msg_prevnode(l->pmsg);
231 }
232
233 /**
234  * tipc_link_create - create a new link
235  * @n: pointer to associated node
236  * @if_name: associated interface name
237  * @bearer_id: id (index) of associated bearer
238  * @tolerance: link tolerance to be used by link
239  * @net_plane: network plane (A,B,c..) this link belongs to
240  * @mtu: mtu to be advertised by link
241  * @priority: priority to be used by link
242  * @window: send window to be used by link
243  * @session: session to be used by link
244  * @ownnode: identity of own node
245  * @peer: node id of peer node
246  * @peer_caps: bitmap describing peer node capabilities
247  * @maddr: media address to be used
248  * @bc_sndlink: the namespace global link used for broadcast sending
249  * @bc_rcvlink: the peer specific link used for broadcast reception
250  * @inputq: queue to put messages ready for delivery
251  * @namedq: queue to put binding table update messages ready for delivery
252  * @link: return value, pointer to put the created link
253  *
254  * Returns true if link was created, otherwise false
255  */
256 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
257                       int tolerance, char net_plane, u32 mtu, int priority,
258                       int window, u32 session, u32 ownnode, u32 peer,
259                       u16 peer_caps,
260                       struct tipc_media_addr *maddr,
261                       struct tipc_link *bc_sndlink,
262                       struct tipc_link *bc_rcvlink,
263                       struct sk_buff_head *inputq,
264                       struct sk_buff_head *namedq,
265                       struct tipc_link **link)
266 {
267         struct tipc_link *l;
268         struct tipc_msg *hdr;
269
270         l = kzalloc(sizeof(*l), GFP_ATOMIC);
271         if (!l)
272                 return false;
273         *link = l;
274         l->pmsg = (struct tipc_msg *)&l->proto_msg;
275         hdr = l->pmsg;
276         tipc_msg_init(ownnode, hdr, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, peer);
277         msg_set_size(hdr, sizeof(l->proto_msg));
278         msg_set_session(hdr, session);
279         msg_set_bearer_id(hdr, l->bearer_id);
280
281         /* Note: peer i/f name is completed by reset/activate message */
282         sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
283                 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
284                 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
285         strcpy((char *)msg_data(hdr), if_name);
286
287         l->addr = peer;
288         l->peer_caps = peer_caps;
289         l->media_addr = maddr;
290         l->net = net;
291         l->peer_session = WILDCARD_SESSION;
292         l->bearer_id = bearer_id;
293         l->tolerance = tolerance;
294         l->net_plane = net_plane;
295         l->advertised_mtu = mtu;
296         l->mtu = mtu;
297         l->priority = priority;
298         tipc_link_set_queue_limits(l, window);
299         l->ackers = 1;
300         l->bc_sndlink = bc_sndlink;
301         l->bc_rcvlink = bc_rcvlink;
302         l->inputq = inputq;
303         l->namedq = namedq;
304         l->state = LINK_RESETTING;
305         __skb_queue_head_init(&l->transmq);
306         __skb_queue_head_init(&l->backlogq);
307         __skb_queue_head_init(&l->deferdq);
308         skb_queue_head_init(&l->wakeupq);
309         skb_queue_head_init(l->inputq);
310         return true;
311 }
312
313 /**
314  * tipc_link_bc_create - create new link to be used for broadcast
315  * @n: pointer to associated node
316  * @mtu: mtu to be used
317  * @window: send window to be used
318  * @inputq: queue to put messages ready for delivery
319  * @namedq: queue to put binding table update messages ready for delivery
320  * @link: return value, pointer to put the created link
321  *
322  * Returns true if link was created, otherwise false
323  */
324 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
325                          int mtu, int window, u16 peer_caps,
326                          struct sk_buff_head *inputq,
327                          struct sk_buff_head *namedq,
328                          struct tipc_link *bc_sndlink,
329                          struct tipc_link **link)
330 {
331         struct tipc_link *l;
332
333         if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
334                               0, ownnode, peer, peer_caps, NULL, bc_sndlink,
335                               NULL, inputq, namedq, link))
336                 return false;
337
338         l = *link;
339         strcpy(l->name, tipc_bclink_name);
340         tipc_link_reset(l);
341         l->state = LINK_RESET;
342         l->ackers = 0;
343         l->bc_rcvlink = l;
344
345         /* Broadcast send link is always up */
346         if (link_is_bc_sndlink(l))
347                 l->state = LINK_ESTABLISHED;
348
349         return true;
350 }
351
352 /**
353  * tipc_link_fsm_evt - link finite state machine
354  * @l: pointer to link
355  * @evt: state machine event to be processed
356  */
357 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
358 {
359         int rc = 0;
360
361         switch (l->state) {
362         case LINK_RESETTING:
363                 switch (evt) {
364                 case LINK_PEER_RESET_EVT:
365                         l->state = LINK_PEER_RESET;
366                         break;
367                 case LINK_RESET_EVT:
368                         l->state = LINK_RESET;
369                         break;
370                 case LINK_FAILURE_EVT:
371                 case LINK_FAILOVER_BEGIN_EVT:
372                 case LINK_ESTABLISH_EVT:
373                 case LINK_FAILOVER_END_EVT:
374                 case LINK_SYNCH_BEGIN_EVT:
375                 case LINK_SYNCH_END_EVT:
376                 default:
377                         goto illegal_evt;
378                 }
379                 break;
380         case LINK_RESET:
381                 switch (evt) {
382                 case LINK_PEER_RESET_EVT:
383                         l->state = LINK_ESTABLISHING;
384                         break;
385                 case LINK_FAILOVER_BEGIN_EVT:
386                         l->state = LINK_FAILINGOVER;
387                 case LINK_FAILURE_EVT:
388                 case LINK_RESET_EVT:
389                 case LINK_ESTABLISH_EVT:
390                 case LINK_FAILOVER_END_EVT:
391                         break;
392                 case LINK_SYNCH_BEGIN_EVT:
393                 case LINK_SYNCH_END_EVT:
394                 default:
395                         goto illegal_evt;
396                 }
397                 break;
398         case LINK_PEER_RESET:
399                 switch (evt) {
400                 case LINK_RESET_EVT:
401                         l->state = LINK_ESTABLISHING;
402                         break;
403                 case LINK_PEER_RESET_EVT:
404                 case LINK_ESTABLISH_EVT:
405                 case LINK_FAILURE_EVT:
406                         break;
407                 case LINK_SYNCH_BEGIN_EVT:
408                 case LINK_SYNCH_END_EVT:
409                 case LINK_FAILOVER_BEGIN_EVT:
410                 case LINK_FAILOVER_END_EVT:
411                 default:
412                         goto illegal_evt;
413                 }
414                 break;
415         case LINK_FAILINGOVER:
416                 switch (evt) {
417                 case LINK_FAILOVER_END_EVT:
418                         l->state = LINK_RESET;
419                         break;
420                 case LINK_PEER_RESET_EVT:
421                 case LINK_RESET_EVT:
422                 case LINK_ESTABLISH_EVT:
423                 case LINK_FAILURE_EVT:
424                         break;
425                 case LINK_FAILOVER_BEGIN_EVT:
426                 case LINK_SYNCH_BEGIN_EVT:
427                 case LINK_SYNCH_END_EVT:
428                 default:
429                         goto illegal_evt;
430                 }
431                 break;
432         case LINK_ESTABLISHING:
433                 switch (evt) {
434                 case LINK_ESTABLISH_EVT:
435                         l->state = LINK_ESTABLISHED;
436                         break;
437                 case LINK_FAILOVER_BEGIN_EVT:
438                         l->state = LINK_FAILINGOVER;
439                         break;
440                 case LINK_RESET_EVT:
441                         l->state = LINK_RESET;
442                         break;
443                 case LINK_FAILURE_EVT:
444                 case LINK_PEER_RESET_EVT:
445                 case LINK_SYNCH_BEGIN_EVT:
446                 case LINK_FAILOVER_END_EVT:
447                         break;
448                 case LINK_SYNCH_END_EVT:
449                 default:
450                         goto illegal_evt;
451                 }
452                 break;
453         case LINK_ESTABLISHED:
454                 switch (evt) {
455                 case LINK_PEER_RESET_EVT:
456                         l->state = LINK_PEER_RESET;
457                         rc |= TIPC_LINK_DOWN_EVT;
458                         break;
459                 case LINK_FAILURE_EVT:
460                         l->state = LINK_RESETTING;
461                         rc |= TIPC_LINK_DOWN_EVT;
462                         break;
463                 case LINK_RESET_EVT:
464                         l->state = LINK_RESET;
465                         break;
466                 case LINK_ESTABLISH_EVT:
467                 case LINK_SYNCH_END_EVT:
468                         break;
469                 case LINK_SYNCH_BEGIN_EVT:
470                         l->state = LINK_SYNCHING;
471                         break;
472                 case LINK_FAILOVER_BEGIN_EVT:
473                 case LINK_FAILOVER_END_EVT:
474                 default:
475                         goto illegal_evt;
476                 }
477                 break;
478         case LINK_SYNCHING:
479                 switch (evt) {
480                 case LINK_PEER_RESET_EVT:
481                         l->state = LINK_PEER_RESET;
482                         rc |= TIPC_LINK_DOWN_EVT;
483                         break;
484                 case LINK_FAILURE_EVT:
485                         l->state = LINK_RESETTING;
486                         rc |= TIPC_LINK_DOWN_EVT;
487                         break;
488                 case LINK_RESET_EVT:
489                         l->state = LINK_RESET;
490                         break;
491                 case LINK_ESTABLISH_EVT:
492                 case LINK_SYNCH_BEGIN_EVT:
493                         break;
494                 case LINK_SYNCH_END_EVT:
495                         l->state = LINK_ESTABLISHED;
496                         break;
497                 case LINK_FAILOVER_BEGIN_EVT:
498                 case LINK_FAILOVER_END_EVT:
499                 default:
500                         goto illegal_evt;
501                 }
502                 break;
503         default:
504                 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
505         }
506         return rc;
507 illegal_evt:
508         pr_err("Illegal FSM event %x in state %x on link %s\n",
509                evt, l->state, l->name);
510         return rc;
511 }
512
513 /* link_profile_stats - update statistical profiling of traffic
514  */
515 static void link_profile_stats(struct tipc_link *l)
516 {
517         struct sk_buff *skb;
518         struct tipc_msg *msg;
519         int length;
520
521         /* Update counters used in statistical profiling of send traffic */
522         l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
523         l->stats.queue_sz_counts++;
524
525         skb = skb_peek(&l->transmq);
526         if (!skb)
527                 return;
528         msg = buf_msg(skb);
529         length = msg_size(msg);
530
531         if (msg_user(msg) == MSG_FRAGMENTER) {
532                 if (msg_type(msg) != FIRST_FRAGMENT)
533                         return;
534                 length = msg_size(msg_get_wrapped(msg));
535         }
536         l->stats.msg_lengths_total += length;
537         l->stats.msg_length_counts++;
538         if (length <= 64)
539                 l->stats.msg_length_profile[0]++;
540         else if (length <= 256)
541                 l->stats.msg_length_profile[1]++;
542         else if (length <= 1024)
543                 l->stats.msg_length_profile[2]++;
544         else if (length <= 4096)
545                 l->stats.msg_length_profile[3]++;
546         else if (length <= 16384)
547                 l->stats.msg_length_profile[4]++;
548         else if (length <= 32768)
549                 l->stats.msg_length_profile[5]++;
550         else
551                 l->stats.msg_length_profile[6]++;
552 }
553
554 /* tipc_link_timeout - perform periodic task as instructed from node timeout
555  */
556 /* tipc_link_timeout - perform periodic task as instructed from node timeout
557  */
558 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
559 {
560         int rc = 0;
561         int mtyp = STATE_MSG;
562         bool xmit = false;
563         bool prb = false;
564         u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
565         u16 bc_acked = l->bc_rcvlink->acked;
566         bool bc_up = link_is_up(l->bc_rcvlink);
567
568         link_profile_stats(l);
569
570         switch (l->state) {
571         case LINK_ESTABLISHED:
572         case LINK_SYNCHING:
573                 if (!l->silent_intv_cnt) {
574                         if (bc_up && (bc_acked != bc_snt))
575                                 xmit = true;
576                 } else if (l->silent_intv_cnt <= l->abort_limit) {
577                         xmit = true;
578                         prb = true;
579                 } else {
580                         rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
581                 }
582                 l->silent_intv_cnt++;
583                 break;
584         case LINK_RESET:
585                 xmit = true;
586                 mtyp = RESET_MSG;
587                 break;
588         case LINK_ESTABLISHING:
589                 xmit = true;
590                 mtyp = ACTIVATE_MSG;
591                 break;
592         case LINK_PEER_RESET:
593         case LINK_RESETTING:
594         case LINK_FAILINGOVER:
595                 break;
596         default:
597                 break;
598         }
599
600         if (xmit)
601                 tipc_link_build_proto_msg(l, mtyp, prb, 0, 0, 0, xmitq);
602
603         return rc;
604 }
605
606 /**
607  * link_schedule_user - schedule a message sender for wakeup after congestion
608  * @link: congested link
609  * @list: message that was attempted sent
610  * Create pseudo msg to send back to user when congestion abates
611  * Does not consume buffer list
612  */
613 static int link_schedule_user(struct tipc_link *link, struct sk_buff_head *list)
614 {
615         struct tipc_msg *msg = buf_msg(skb_peek(list));
616         int imp = msg_importance(msg);
617         u32 oport = msg_origport(msg);
618         u32 addr = link_own_addr(link);
619         struct sk_buff *skb;
620
621         /* This really cannot happen...  */
622         if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
623                 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
624                 return -ENOBUFS;
625         }
626         /* Non-blocking sender: */
627         if (TIPC_SKB_CB(skb_peek(list))->wakeup_pending)
628                 return -ELINKCONG;
629
630         /* Create and schedule wakeup pseudo message */
631         skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
632                               addr, addr, oport, 0, 0);
633         if (!skb)
634                 return -ENOBUFS;
635         TIPC_SKB_CB(skb)->chain_sz = skb_queue_len(list);
636         TIPC_SKB_CB(skb)->chain_imp = imp;
637         skb_queue_tail(&link->wakeupq, skb);
638         link->stats.link_congs++;
639         return -ELINKCONG;
640 }
641
642 /**
643  * link_prepare_wakeup - prepare users for wakeup after congestion
644  * @link: congested link
645  * Move a number of waiting users, as permitted by available space in
646  * the send queue, from link wait queue to node wait queue for wakeup
647  */
648 void link_prepare_wakeup(struct tipc_link *l)
649 {
650         int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
651         int imp, lim;
652         struct sk_buff *skb, *tmp;
653
654         skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
655                 imp = TIPC_SKB_CB(skb)->chain_imp;
656                 lim = l->window + l->backlog[imp].limit;
657                 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
658                 if ((pnd[imp] + l->backlog[imp].len) >= lim)
659                         break;
660                 skb_unlink(skb, &l->wakeupq);
661                 skb_queue_tail(l->inputq, skb);
662         }
663 }
664
665 /**
666  * tipc_link_reset_fragments - purge link's inbound message fragments queue
667  * @l_ptr: pointer to link
668  */
669 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
670 {
671         kfree_skb(l_ptr->reasm_buf);
672         l_ptr->reasm_buf = NULL;
673 }
674
675 void tipc_link_purge_backlog(struct tipc_link *l)
676 {
677         __skb_queue_purge(&l->backlogq);
678         l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
679         l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
680         l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
681         l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
682         l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
683 }
684
685 /**
686  * tipc_link_purge_queues - purge all pkt queues associated with link
687  * @l_ptr: pointer to link
688  */
689 void tipc_link_purge_queues(struct tipc_link *l_ptr)
690 {
691         __skb_queue_purge(&l_ptr->deferdq);
692         __skb_queue_purge(&l_ptr->transmq);
693         tipc_link_purge_backlog(l_ptr);
694         tipc_link_reset_fragments(l_ptr);
695 }
696
697 void tipc_link_reset(struct tipc_link *l)
698 {
699         /* Link is down, accept any session */
700         l->peer_session = WILDCARD_SESSION;
701
702         /* If peer is up, it only accepts an incremented session number */
703         msg_set_session(l->pmsg, msg_session(l->pmsg) + 1);
704
705         /* Prepare for renewed mtu size negotiation */
706         l->mtu = l->advertised_mtu;
707
708         /* Clean up all queues: */
709         __skb_queue_purge(&l->transmq);
710         __skb_queue_purge(&l->deferdq);
711         skb_queue_splice_init(&l->wakeupq, l->inputq);
712
713         tipc_link_purge_backlog(l);
714         kfree_skb(l->reasm_buf);
715         kfree_skb(l->failover_reasm_skb);
716         l->reasm_buf = NULL;
717         l->failover_reasm_skb = NULL;
718         l->rcv_unacked = 0;
719         l->snd_nxt = 1;
720         l->rcv_nxt = 1;
721         l->acked = 0;
722         l->silent_intv_cnt = 0;
723         l->stats.recv_info = 0;
724         l->stale_count = 0;
725         l->bc_peer_is_up = false;
726         link_reset_statistics(l);
727 }
728
729 /**
730  * __tipc_link_xmit(): same as tipc_link_xmit, but destlink is known & locked
731  * @link: link to use
732  * @list: chain of buffers containing message
733  *
734  * Consumes the buffer chain, except when returning an error code,
735  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
736  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
737  */
738 int __tipc_link_xmit(struct net *net, struct tipc_link *link,
739                      struct sk_buff_head *list)
740 {
741         struct tipc_msg *msg = buf_msg(skb_peek(list));
742         unsigned int maxwin = link->window;
743         unsigned int i, imp = msg_importance(msg);
744         uint mtu = link->mtu;
745         u16 ack = mod(link->rcv_nxt - 1);
746         u16 seqno = link->snd_nxt;
747         u16 bc_ack = link->bc_rcvlink->rcv_nxt - 1;
748         struct tipc_media_addr *addr = link->media_addr;
749         struct sk_buff_head *transmq = &link->transmq;
750         struct sk_buff_head *backlogq = &link->backlogq;
751         struct sk_buff *skb, *bskb;
752
753         /* Match msg importance against this and all higher backlog limits: */
754         for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
755                 if (unlikely(link->backlog[i].len >= link->backlog[i].limit))
756                         return link_schedule_user(link, list);
757         }
758         if (unlikely(msg_size(msg) > mtu))
759                 return -EMSGSIZE;
760
761         /* Prepare each packet for sending, and add to relevant queue: */
762         while (skb_queue_len(list)) {
763                 skb = skb_peek(list);
764                 msg = buf_msg(skb);
765                 msg_set_seqno(msg, seqno);
766                 msg_set_ack(msg, ack);
767                 msg_set_bcast_ack(msg, bc_ack);
768
769                 if (likely(skb_queue_len(transmq) < maxwin)) {
770                         __skb_dequeue(list);
771                         __skb_queue_tail(transmq, skb);
772                         tipc_bearer_send(net, link->bearer_id, skb, addr);
773                         link->rcv_unacked = 0;
774                         seqno++;
775                         continue;
776                 }
777                 if (tipc_msg_bundle(skb_peek_tail(backlogq), msg, mtu)) {
778                         kfree_skb(__skb_dequeue(list));
779                         link->stats.sent_bundled++;
780                         continue;
781                 }
782                 if (tipc_msg_make_bundle(&bskb, msg, mtu, link->addr)) {
783                         kfree_skb(__skb_dequeue(list));
784                         __skb_queue_tail(backlogq, bskb);
785                         link->backlog[msg_importance(buf_msg(bskb))].len++;
786                         link->stats.sent_bundled++;
787                         link->stats.sent_bundles++;
788                         continue;
789                 }
790                 link->backlog[imp].len += skb_queue_len(list);
791                 skb_queue_splice_tail_init(list, backlogq);
792         }
793         link->snd_nxt = seqno;
794         return 0;
795 }
796
797 /**
798  * tipc_link_xmit(): enqueue buffer list according to queue situation
799  * @link: link to use
800  * @list: chain of buffers containing message
801  * @xmitq: returned list of packets to be sent by caller
802  *
803  * Consumes the buffer chain, except when returning -ELINKCONG,
804  * since the caller then may want to make more send attempts.
805  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
806  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
807  */
808 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
809                    struct sk_buff_head *xmitq)
810 {
811         struct tipc_msg *hdr = buf_msg(skb_peek(list));
812         unsigned int maxwin = l->window;
813         unsigned int i, imp = msg_importance(hdr);
814         unsigned int mtu = l->mtu;
815         u16 ack = l->rcv_nxt - 1;
816         u16 seqno = l->snd_nxt;
817         u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
818         struct sk_buff_head *transmq = &l->transmq;
819         struct sk_buff_head *backlogq = &l->backlogq;
820         struct sk_buff *skb, *_skb, *bskb;
821
822         /* Match msg importance against this and all higher backlog limits: */
823         for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
824                 if (unlikely(l->backlog[i].len >= l->backlog[i].limit))
825                         return link_schedule_user(l, list);
826         }
827         if (unlikely(msg_size(hdr) > mtu))
828                 return -EMSGSIZE;
829
830         /* Prepare each packet for sending, and add to relevant queue: */
831         while (skb_queue_len(list)) {
832                 skb = skb_peek(list);
833                 hdr = buf_msg(skb);
834                 msg_set_seqno(hdr, seqno);
835                 msg_set_ack(hdr, ack);
836                 msg_set_bcast_ack(hdr, bc_ack);
837
838                 if (likely(skb_queue_len(transmq) < maxwin)) {
839                         _skb = skb_clone(skb, GFP_ATOMIC);
840                         if (!_skb)
841                                 return -ENOBUFS;
842                         __skb_dequeue(list);
843                         __skb_queue_tail(transmq, skb);
844                         __skb_queue_tail(xmitq, _skb);
845                         TIPC_SKB_CB(skb)->ackers = l->ackers;
846                         l->rcv_unacked = 0;
847                         seqno++;
848                         continue;
849                 }
850                 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
851                         kfree_skb(__skb_dequeue(list));
852                         l->stats.sent_bundled++;
853                         continue;
854                 }
855                 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
856                         kfree_skb(__skb_dequeue(list));
857                         __skb_queue_tail(backlogq, bskb);
858                         l->backlog[msg_importance(buf_msg(bskb))].len++;
859                         l->stats.sent_bundled++;
860                         l->stats.sent_bundles++;
861                         continue;
862                 }
863                 l->backlog[imp].len += skb_queue_len(list);
864                 skb_queue_splice_tail_init(list, backlogq);
865         }
866         l->snd_nxt = seqno;
867         return 0;
868 }
869
870 /*
871  * tipc_link_push_packets - push unsent packets to bearer
872  *
873  * Push out the unsent messages of a link where congestion
874  * has abated. Node is locked.
875  *
876  * Called with node locked
877  */
878 void tipc_link_push_packets(struct tipc_link *link)
879 {
880         struct sk_buff *skb;
881         struct tipc_msg *msg;
882         u16 seqno = link->snd_nxt;
883         u16 ack = mod(link->rcv_nxt - 1);
884
885         while (skb_queue_len(&link->transmq) < link->window) {
886                 skb = __skb_dequeue(&link->backlogq);
887                 if (!skb)
888                         break;
889                 TIPC_SKB_CB(skb)->ackers = link->ackers;
890                 msg = buf_msg(skb);
891                 link->backlog[msg_importance(msg)].len--;
892                 msg_set_ack(msg, ack);
893                 msg_set_seqno(msg, seqno);
894                 seqno = mod(seqno + 1);
895                 /* msg_set_bcast_ack(msg, link->owner->bclink.last_in); */
896                 link->rcv_unacked = 0;
897                 __skb_queue_tail(&link->transmq, skb);
898                 tipc_bearer_send(link->net, link->bearer_id,
899                                  skb, link->media_addr);
900         }
901         link->snd_nxt = seqno;
902 }
903
904 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
905 {
906         struct sk_buff *skb, *_skb;
907         struct tipc_msg *hdr;
908         u16 seqno = l->snd_nxt;
909         u16 ack = l->rcv_nxt - 1;
910         u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
911
912         while (skb_queue_len(&l->transmq) < l->window) {
913                 skb = skb_peek(&l->backlogq);
914                 if (!skb)
915                         break;
916                 _skb = skb_clone(skb, GFP_ATOMIC);
917                 if (!_skb)
918                         break;
919                 __skb_dequeue(&l->backlogq);
920                 hdr = buf_msg(skb);
921                 l->backlog[msg_importance(hdr)].len--;
922                 __skb_queue_tail(&l->transmq, skb);
923                 __skb_queue_tail(xmitq, _skb);
924                 TIPC_SKB_CB(skb)->ackers = l->ackers;
925                 msg_set_seqno(hdr, seqno);
926                 msg_set_ack(hdr, ack);
927                 msg_set_bcast_ack(hdr, bc_ack);
928                 l->rcv_unacked = 0;
929                 seqno++;
930         }
931         l->snd_nxt = seqno;
932 }
933
934 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
935 {
936         struct tipc_msg *hdr = buf_msg(skb);
937
938         pr_warn("Retransmission failure on link <%s>\n", l->name);
939         link_print(l, "Resetting link ");
940         pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
941                 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
942         pr_info("sqno %u, prev: %x, src: %x\n",
943                 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
944 }
945
946 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *skb,
947                           u32 retransmits)
948 {
949         struct tipc_msg *msg;
950
951         if (!skb)
952                 return;
953
954         msg = buf_msg(skb);
955
956         /* Detect repeated retransmit failures */
957         if (l_ptr->last_retransm == msg_seqno(msg)) {
958                 if (++l_ptr->stale_count > 100) {
959                         link_retransmit_failure(l_ptr, skb);
960                         return;
961                 }
962         } else {
963                 l_ptr->last_retransm = msg_seqno(msg);
964                 l_ptr->stale_count = 1;
965         }
966
967         skb_queue_walk_from(&l_ptr->transmq, skb) {
968                 if (!retransmits)
969                         break;
970                 msg = buf_msg(skb);
971                 msg_set_ack(msg, mod(l_ptr->rcv_nxt - 1));
972                 /* msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in); */
973                 tipc_bearer_send(l_ptr->net, l_ptr->bearer_id, skb,
974                                  l_ptr->media_addr);
975                 retransmits--;
976                 l_ptr->stats.retransmitted++;
977         }
978 }
979
980 int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
981                       struct sk_buff_head *xmitq)
982 {
983         struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
984         struct tipc_msg *hdr;
985         u16 ack = l->rcv_nxt - 1;
986         u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
987
988         if (!skb)
989                 return 0;
990
991         /* Detect repeated retransmit failures on same packet */
992         if (likely(l->last_retransm != buf_seqno(skb))) {
993                 l->last_retransm = buf_seqno(skb);
994                 l->stale_count = 1;
995         } else if (++l->stale_count > 100) {
996                 link_retransmit_failure(l, skb);
997                 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
998         }
999
1000         /* Move forward to where retransmission should start */
1001         skb_queue_walk(&l->transmq, skb) {
1002                 if (!less(buf_seqno(skb), from))
1003                         break;
1004         }
1005
1006         skb_queue_walk_from(&l->transmq, skb) {
1007                 if (more(buf_seqno(skb), to))
1008                         break;
1009                 hdr = buf_msg(skb);
1010                 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1011                 if (!_skb)
1012                         return 0;
1013                 hdr = buf_msg(_skb);
1014                 msg_set_ack(hdr, ack);
1015                 msg_set_bcast_ack(hdr, bc_ack);
1016                 _skb->priority = TC_PRIO_CONTROL;
1017                 __skb_queue_tail(xmitq, _skb);
1018                 l->stats.retransmitted++;
1019         }
1020         return 0;
1021 }
1022
1023 /* tipc_data_input - deliver data and name distr msgs to upper layer
1024  *
1025  * Consumes buffer if message is of right type
1026  * Node lock must be held
1027  */
1028 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1029                             struct sk_buff_head *inputq)
1030 {
1031         switch (msg_user(buf_msg(skb))) {
1032         case TIPC_LOW_IMPORTANCE:
1033         case TIPC_MEDIUM_IMPORTANCE:
1034         case TIPC_HIGH_IMPORTANCE:
1035         case TIPC_CRITICAL_IMPORTANCE:
1036         case CONN_MANAGER:
1037                 skb_queue_tail(inputq, skb);
1038                 return true;
1039         case NAME_DISTRIBUTOR:
1040                 l->bc_rcvlink->state = LINK_ESTABLISHED;
1041                 skb_queue_tail(l->namedq, skb);
1042                 return true;
1043         case MSG_BUNDLER:
1044         case TUNNEL_PROTOCOL:
1045         case MSG_FRAGMENTER:
1046         case BCAST_PROTOCOL:
1047                 return false;
1048         default:
1049                 pr_warn("Dropping received illegal msg type\n");
1050                 kfree_skb(skb);
1051                 return false;
1052         };
1053 }
1054
1055 /* tipc_link_input - process packet that has passed link protocol check
1056  *
1057  * Consumes buffer
1058  */
1059 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1060                            struct sk_buff_head *inputq)
1061 {
1062         struct tipc_msg *hdr = buf_msg(skb);
1063         struct sk_buff **reasm_skb = &l->reasm_buf;
1064         struct sk_buff *iskb;
1065         struct sk_buff_head tmpq;
1066         int usr = msg_user(hdr);
1067         int rc = 0;
1068         int pos = 0;
1069         int ipos = 0;
1070
1071         if (unlikely(usr == TUNNEL_PROTOCOL)) {
1072                 if (msg_type(hdr) == SYNCH_MSG) {
1073                         __skb_queue_purge(&l->deferdq);
1074                         goto drop;
1075                 }
1076                 if (!tipc_msg_extract(skb, &iskb, &ipos))
1077                         return rc;
1078                 kfree_skb(skb);
1079                 skb = iskb;
1080                 hdr = buf_msg(skb);
1081                 if (less(msg_seqno(hdr), l->drop_point))
1082                         goto drop;
1083                 if (tipc_data_input(l, skb, inputq))
1084                         return rc;
1085                 usr = msg_user(hdr);
1086                 reasm_skb = &l->failover_reasm_skb;
1087         }
1088
1089         if (usr == MSG_BUNDLER) {
1090                 skb_queue_head_init(&tmpq);
1091                 l->stats.recv_bundles++;
1092                 l->stats.recv_bundled += msg_msgcnt(hdr);
1093                 while (tipc_msg_extract(skb, &iskb, &pos))
1094                         tipc_data_input(l, iskb, &tmpq);
1095                 tipc_skb_queue_splice_tail(&tmpq, inputq);
1096                 return 0;
1097         } else if (usr == MSG_FRAGMENTER) {
1098                 l->stats.recv_fragments++;
1099                 if (tipc_buf_append(reasm_skb, &skb)) {
1100                         l->stats.recv_fragmented++;
1101                         tipc_data_input(l, skb, inputq);
1102                 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1103                         pr_warn_ratelimited("Unable to build fragment list\n");
1104                         return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1105                 }
1106                 return 0;
1107         } else if (usr == BCAST_PROTOCOL) {
1108                 tipc_bcast_lock(l->net);
1109                 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1110                 tipc_bcast_unlock(l->net);
1111         }
1112 drop:
1113         kfree_skb(skb);
1114         return 0;
1115 }
1116
1117 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1118 {
1119         bool released = false;
1120         struct sk_buff *skb, *tmp;
1121
1122         skb_queue_walk_safe(&l->transmq, skb, tmp) {
1123                 if (more(buf_seqno(skb), acked))
1124                         break;
1125                 __skb_unlink(skb, &l->transmq);
1126                 kfree_skb(skb);
1127                 released = true;
1128         }
1129         return released;
1130 }
1131
1132 /* tipc_link_build_ack_msg: prepare link acknowledge message for transmission
1133  *
1134  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1135  * risk of ack storms towards the sender
1136  */
1137 int tipc_link_build_ack_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1138 {
1139         if (!l)
1140                 return 0;
1141
1142         /* Broadcast ACK must be sent via a unicast link => defer to caller */
1143         if (link_is_bc_rcvlink(l)) {
1144                 if (((l->rcv_nxt ^ link_own_addr(l)) & 0xf) != 0xf)
1145                         return 0;
1146                 l->rcv_unacked = 0;
1147                 return TIPC_LINK_SND_BC_ACK;
1148         }
1149
1150         /* Unicast ACK */
1151         l->rcv_unacked = 0;
1152         l->stats.sent_acks++;
1153         tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1154         return 0;
1155 }
1156
1157 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1158  */
1159 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1160 {
1161         int mtyp = RESET_MSG;
1162
1163         if (l->state == LINK_ESTABLISHING)
1164                 mtyp = ACTIVATE_MSG;
1165
1166         tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1167 }
1168
1169 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1170  */
1171 static void tipc_link_build_nack_msg(struct tipc_link *l,
1172                                      struct sk_buff_head *xmitq)
1173 {
1174         u32 def_cnt = ++l->stats.deferred_recv;
1175
1176         if (link_is_bc_rcvlink(l))
1177                 return;
1178
1179         if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1180                 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1181 }
1182
1183 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1184  * @l: the link that should handle the message
1185  * @skb: TIPC packet
1186  * @xmitq: queue to place packets to be sent after this call
1187  */
1188 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1189                   struct sk_buff_head *xmitq)
1190 {
1191         struct sk_buff_head *defq = &l->deferdq;
1192         struct tipc_msg *hdr;
1193         u16 seqno, rcv_nxt, win_lim;
1194         int rc = 0;
1195
1196         do {
1197                 hdr = buf_msg(skb);
1198                 seqno = msg_seqno(hdr);
1199                 rcv_nxt = l->rcv_nxt;
1200                 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1201
1202                 /* Verify and update link state */
1203                 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1204                         return tipc_link_proto_rcv(l, skb, xmitq);
1205
1206                 if (unlikely(!link_is_up(l))) {
1207                         if (l->state == LINK_ESTABLISHING)
1208                                 rc = TIPC_LINK_UP_EVT;
1209                         goto drop;
1210                 }
1211
1212                 /* Don't send probe at next timeout expiration */
1213                 l->silent_intv_cnt = 0;
1214
1215                 /* Drop if outside receive window */
1216                 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1217                         l->stats.duplicates++;
1218                         goto drop;
1219                 }
1220
1221                 /* Forward queues and wake up waiting users */
1222                 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1223                         tipc_link_advance_backlog(l, xmitq);
1224                         if (unlikely(!skb_queue_empty(&l->wakeupq)))
1225                                 link_prepare_wakeup(l);
1226                 }
1227
1228                 /* Defer delivery if sequence gap */
1229                 if (unlikely(seqno != rcv_nxt)) {
1230                         __tipc_skb_queue_sorted(defq, seqno, skb);
1231                         tipc_link_build_nack_msg(l, xmitq);
1232                         break;
1233                 }
1234
1235                 /* Deliver packet */
1236                 l->rcv_nxt++;
1237                 l->stats.recv_info++;
1238                 if (!tipc_data_input(l, skb, l->inputq))
1239                         rc |= tipc_link_input(l, skb, l->inputq);
1240                 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1241                         rc |= tipc_link_build_ack_msg(l, xmitq);
1242                 if (unlikely(rc & ~TIPC_LINK_SND_BC_ACK))
1243                         break;
1244         } while ((skb = __skb_dequeue(defq)));
1245
1246         return rc;
1247 drop:
1248         kfree_skb(skb);
1249         return rc;
1250 }
1251
1252 /**
1253  * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1254  *
1255  * Returns increase in queue length (i.e. 0 or 1)
1256  */
1257 u32 tipc_link_defer_pkt(struct sk_buff_head *list, struct sk_buff *skb)
1258 {
1259         struct sk_buff *skb1;
1260         u16 seq_no = buf_seqno(skb);
1261
1262         /* Empty queue ? */
1263         if (skb_queue_empty(list)) {
1264                 __skb_queue_tail(list, skb);
1265                 return 1;
1266         }
1267
1268         /* Last ? */
1269         if (less(buf_seqno(skb_peek_tail(list)), seq_no)) {
1270                 __skb_queue_tail(list, skb);
1271                 return 1;
1272         }
1273
1274         /* Locate insertion point in queue, then insert; discard if duplicate */
1275         skb_queue_walk(list, skb1) {
1276                 u16 curr_seqno = buf_seqno(skb1);
1277
1278                 if (seq_no == curr_seqno) {
1279                         kfree_skb(skb);
1280                         return 0;
1281                 }
1282
1283                 if (less(seq_no, curr_seqno))
1284                         break;
1285         }
1286
1287         __skb_queue_before(list, skb1, skb);
1288         return 1;
1289 }
1290
1291 /*
1292  * Send protocol message to the other endpoint.
1293  */
1294 void tipc_link_proto_xmit(struct tipc_link *l, u32 msg_typ, int probe_msg,
1295                           u32 gap, u32 tolerance, u32 priority)
1296 {
1297         struct sk_buff *skb = NULL;
1298         struct sk_buff_head xmitq;
1299
1300         __skb_queue_head_init(&xmitq);
1301         tipc_link_build_proto_msg(l, msg_typ, probe_msg, gap,
1302                                   tolerance, priority, &xmitq);
1303         skb = __skb_dequeue(&xmitq);
1304         if (!skb)
1305                 return;
1306         tipc_bearer_xmit_skb(l->net, l->bearer_id, skb, l->media_addr);
1307         l->rcv_unacked = 0;
1308 }
1309
1310 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1311                                       u16 rcvgap, int tolerance, int priority,
1312                                       struct sk_buff_head *xmitq)
1313 {
1314         struct sk_buff *skb = NULL;
1315         struct tipc_msg *hdr = l->pmsg;
1316         bool node_up = link_is_up(l->bc_rcvlink);
1317
1318         /* Don't send protocol message during reset or link failover */
1319         if (tipc_link_is_blocked(l))
1320                 return;
1321
1322         msg_set_type(hdr, mtyp);
1323         msg_set_net_plane(hdr, l->net_plane);
1324         msg_set_next_sent(hdr, l->snd_nxt);
1325         msg_set_ack(hdr, l->rcv_nxt - 1);
1326         msg_set_bcast_ack(hdr, l->bc_rcvlink->rcv_nxt - 1);
1327         msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1328         msg_set_link_tolerance(hdr, tolerance);
1329         msg_set_linkprio(hdr, priority);
1330         msg_set_redundant_link(hdr, node_up);
1331         msg_set_seq_gap(hdr, 0);
1332
1333         /* Compatibility: created msg must not be in sequence with pkt flow */
1334         msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1335
1336         if (mtyp == STATE_MSG) {
1337                 if (!tipc_link_is_up(l))
1338                         return;
1339
1340                 /* Override rcvgap if there are packets in deferred queue */
1341                 if (!skb_queue_empty(&l->deferdq))
1342                         rcvgap = buf_seqno(skb_peek(&l->deferdq)) - l->rcv_nxt;
1343                 if (rcvgap) {
1344                         msg_set_seq_gap(hdr, rcvgap);
1345                         l->stats.sent_nacks++;
1346                 }
1347                 msg_set_probe(hdr, probe);
1348                 if (probe)
1349                         l->stats.sent_probes++;
1350                 l->stats.sent_states++;
1351                 l->rcv_unacked = 0;
1352         } else {
1353                 /* RESET_MSG or ACTIVATE_MSG */
1354                 msg_set_max_pkt(hdr, l->advertised_mtu);
1355                 msg_set_ack(hdr, l->rcv_nxt - 1);
1356                 msg_set_next_sent(hdr, 1);
1357         }
1358         skb = tipc_buf_acquire(msg_size(hdr));
1359         if (!skb)
1360                 return;
1361         skb_copy_to_linear_data(skb, hdr, msg_size(hdr));
1362         skb->priority = TC_PRIO_CONTROL;
1363         __skb_queue_tail(xmitq, skb);
1364 }
1365
1366 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1367  * with contents of the link's transmit and backlog queues.
1368  */
1369 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1370                            int mtyp, struct sk_buff_head *xmitq)
1371 {
1372         struct sk_buff *skb, *tnlskb;
1373         struct tipc_msg *hdr, tnlhdr;
1374         struct sk_buff_head *queue = &l->transmq;
1375         struct sk_buff_head tmpxq, tnlq;
1376         u16 pktlen, pktcnt, seqno = l->snd_nxt;
1377
1378         if (!tnl)
1379                 return;
1380
1381         skb_queue_head_init(&tnlq);
1382         skb_queue_head_init(&tmpxq);
1383
1384         /* At least one packet required for safe algorithm => add dummy */
1385         skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1386                               BASIC_H_SIZE, 0, l->addr, link_own_addr(l),
1387                               0, 0, TIPC_ERR_NO_PORT);
1388         if (!skb) {
1389                 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1390                 return;
1391         }
1392         skb_queue_tail(&tnlq, skb);
1393         tipc_link_xmit(l, &tnlq, &tmpxq);
1394         __skb_queue_purge(&tmpxq);
1395
1396         /* Initialize reusable tunnel packet header */
1397         tipc_msg_init(link_own_addr(l), &tnlhdr, TUNNEL_PROTOCOL,
1398                       mtyp, INT_H_SIZE, l->addr);
1399         pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1400         msg_set_msgcnt(&tnlhdr, pktcnt);
1401         msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1402 tnl:
1403         /* Wrap each packet into a tunnel packet */
1404         skb_queue_walk(queue, skb) {
1405                 hdr = buf_msg(skb);
1406                 if (queue == &l->backlogq)
1407                         msg_set_seqno(hdr, seqno++);
1408                 pktlen = msg_size(hdr);
1409                 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1410                 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE);
1411                 if (!tnlskb) {
1412                         pr_warn("%sunable to send packet\n", link_co_err);
1413                         return;
1414                 }
1415                 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1416                 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1417                 __skb_queue_tail(&tnlq, tnlskb);
1418         }
1419         if (queue != &l->backlogq) {
1420                 queue = &l->backlogq;
1421                 goto tnl;
1422         }
1423
1424         tipc_link_xmit(tnl, &tnlq, xmitq);
1425
1426         if (mtyp == FAILOVER_MSG) {
1427                 tnl->drop_point = l->rcv_nxt;
1428                 tnl->failover_reasm_skb = l->reasm_buf;
1429                 l->reasm_buf = NULL;
1430         }
1431 }
1432
1433 /* tipc_link_proto_rcv(): receive link level protocol message :
1434  * Note that network plane id propagates through the network, and may
1435  * change at any time. The node with lowest numerical id determines
1436  * network plane
1437  */
1438 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1439                                struct sk_buff_head *xmitq)
1440 {
1441         struct tipc_msg *hdr = buf_msg(skb);
1442         u16 rcvgap = 0;
1443         u16 ack = msg_ack(hdr);
1444         u16 gap = msg_seq_gap(hdr);
1445         u16 peers_snd_nxt =  msg_next_sent(hdr);
1446         u16 peers_tol = msg_link_tolerance(hdr);
1447         u16 peers_prio = msg_linkprio(hdr);
1448         u16 rcv_nxt = l->rcv_nxt;
1449         int mtyp = msg_type(hdr);
1450         char *if_name;
1451         int rc = 0;
1452
1453         if (tipc_link_is_blocked(l) || !xmitq)
1454                 goto exit;
1455
1456         if (link_own_addr(l) > msg_prevnode(hdr))
1457                 l->net_plane = msg_net_plane(hdr);
1458
1459         switch (mtyp) {
1460         case RESET_MSG:
1461
1462                 /* Ignore duplicate RESET with old session number */
1463                 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1464                     (l->peer_session != WILDCARD_SESSION))
1465                         break;
1466                 /* fall thru' */
1467
1468         case ACTIVATE_MSG:
1469
1470                 /* Complete own link name with peer's interface name */
1471                 if_name =  strrchr(l->name, ':') + 1;
1472                 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1473                         break;
1474                 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1475                         break;
1476                 strncpy(if_name, msg_data(hdr), TIPC_MAX_IF_NAME);
1477
1478                 /* Update own tolerance if peer indicates a non-zero value */
1479                 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1480                         l->tolerance = peers_tol;
1481
1482                 /* Update own priority if peer's priority is higher */
1483                 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1484                         l->priority = peers_prio;
1485
1486                 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1487                 if ((mtyp == RESET_MSG) || !link_is_up(l))
1488                         rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1489
1490                 /* ACTIVATE_MSG takes up link if it was already locally reset */
1491                 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1492                         rc = TIPC_LINK_UP_EVT;
1493
1494                 l->peer_session = msg_session(hdr);
1495                 l->peer_bearer_id = msg_bearer_id(hdr);
1496                 if (l->mtu > msg_max_pkt(hdr))
1497                         l->mtu = msg_max_pkt(hdr);
1498                 break;
1499
1500         case STATE_MSG:
1501
1502                 /* Update own tolerance if peer indicates a non-zero value */
1503                 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1504                         l->tolerance = peers_tol;
1505
1506                 l->silent_intv_cnt = 0;
1507                 l->stats.recv_states++;
1508                 if (msg_probe(hdr))
1509                         l->stats.recv_probes++;
1510
1511                 if (!link_is_up(l)) {
1512                         if (l->state == LINK_ESTABLISHING)
1513                                 rc = TIPC_LINK_UP_EVT;
1514                         break;
1515                 }
1516
1517                 /* Send NACK if peer has sent pkts we haven't received yet */
1518                 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1519                         rcvgap = peers_snd_nxt - l->rcv_nxt;
1520                 if (rcvgap || (msg_probe(hdr)))
1521                         tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1522                                                   0, 0, xmitq);
1523                 tipc_link_release_pkts(l, ack);
1524
1525                 /* If NACK, retransmit will now start at right position */
1526                 if (gap) {
1527                         rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
1528                         l->stats.recv_nacks++;
1529                 }
1530
1531                 tipc_link_advance_backlog(l, xmitq);
1532                 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1533                         link_prepare_wakeup(l);
1534         }
1535 exit:
1536         kfree_skb(skb);
1537         return rc;
1538 }
1539
1540 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1541  */
1542 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1543                                          u16 peers_snd_nxt,
1544                                          struct sk_buff_head *xmitq)
1545 {
1546         struct sk_buff *skb;
1547         struct tipc_msg *hdr;
1548         struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1549         u16 ack = l->rcv_nxt - 1;
1550         u16 gap_to = peers_snd_nxt - 1;
1551
1552         skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1553                               0, l->addr, link_own_addr(l), 0, 0, 0);
1554         if (!skb)
1555                 return false;
1556         hdr = buf_msg(skb);
1557         msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1558         msg_set_bcast_ack(hdr, ack);
1559         msg_set_bcgap_after(hdr, ack);
1560         if (dfrd_skb)
1561                 gap_to = buf_seqno(dfrd_skb) - 1;
1562         msg_set_bcgap_to(hdr, gap_to);
1563         msg_set_non_seq(hdr, bcast);
1564         __skb_queue_tail(xmitq, skb);
1565         return true;
1566 }
1567
1568 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1569  *
1570  * Give a newly added peer node the sequence number where it should
1571  * start receiving and acking broadcast packets.
1572  */
1573 void tipc_link_build_bc_init_msg(struct tipc_link *l,
1574                                  struct sk_buff_head *xmitq)
1575 {
1576         struct sk_buff_head list;
1577
1578         __skb_queue_head_init(&list);
1579         if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1580                 return;
1581         tipc_link_xmit(l, &list, xmitq);
1582 }
1583
1584 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1585  */
1586 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1587 {
1588         int mtyp = msg_type(hdr);
1589         u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1590
1591         if (link_is_up(l))
1592                 return;
1593
1594         if (msg_user(hdr) == BCAST_PROTOCOL) {
1595                 l->rcv_nxt = peers_snd_nxt;
1596                 l->state = LINK_ESTABLISHED;
1597                 return;
1598         }
1599
1600         if (l->peer_caps & TIPC_BCAST_SYNCH)
1601                 return;
1602
1603         if (msg_peer_node_is_up(hdr))
1604                 return;
1605
1606         /* Compatibility: accept older, less safe initial synch data */
1607         if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1608                 l->rcv_nxt = peers_snd_nxt;
1609 }
1610
1611 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1612  */
1613 void tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1614                            struct sk_buff_head *xmitq)
1615 {
1616         u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1617
1618         if (!link_is_up(l))
1619                 return;
1620
1621         if (!msg_peer_node_is_up(hdr))
1622                 return;
1623
1624         l->bc_peer_is_up = true;
1625
1626         /* Ignore if peers_snd_nxt goes beyond receive window */
1627         if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1628                 return;
1629
1630         if (!more(peers_snd_nxt, l->rcv_nxt)) {
1631                 l->nack_state = BC_NACK_SND_CONDITIONAL;
1632                 return;
1633         }
1634
1635         /* Don't NACK if one was recently sent or peeked */
1636         if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1637                 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1638                 return;
1639         }
1640
1641         /* Conditionally delay NACK sending until next synch rcv */
1642         if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1643                 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1644                 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1645                         return;
1646         }
1647
1648         /* Send NACK now but suppress next one */
1649         tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1650         l->nack_state = BC_NACK_SND_SUPPRESS;
1651 }
1652
1653 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1654                           struct sk_buff_head *xmitq)
1655 {
1656         struct sk_buff *skb, *tmp;
1657         struct tipc_link *snd_l = l->bc_sndlink;
1658
1659         if (!link_is_up(l) || !l->bc_peer_is_up)
1660                 return;
1661
1662         if (!more(acked, l->acked))
1663                 return;
1664
1665         /* Skip over packets peer has already acked */
1666         skb_queue_walk(&snd_l->transmq, skb) {
1667                 if (more(buf_seqno(skb), l->acked))
1668                         break;
1669         }
1670
1671         /* Update/release the packets peer is acking now */
1672         skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1673                 if (more(buf_seqno(skb), acked))
1674                         break;
1675                 if (!--TIPC_SKB_CB(skb)->ackers) {
1676                         __skb_unlink(skb, &snd_l->transmq);
1677                         kfree_skb(skb);
1678                 }
1679         }
1680         l->acked = acked;
1681         tipc_link_advance_backlog(snd_l, xmitq);
1682         if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1683                 link_prepare_wakeup(snd_l);
1684 }
1685
1686 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1687  */
1688 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1689                           struct sk_buff_head *xmitq)
1690 {
1691         struct tipc_msg *hdr = buf_msg(skb);
1692         u32 dnode = msg_destnode(hdr);
1693         int mtyp = msg_type(hdr);
1694         u16 acked = msg_bcast_ack(hdr);
1695         u16 from = acked + 1;
1696         u16 to = msg_bcgap_to(hdr);
1697         u16 peers_snd_nxt = to + 1;
1698         int rc = 0;
1699
1700         kfree_skb(skb);
1701
1702         if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1703                 return 0;
1704
1705         if (mtyp != STATE_MSG)
1706                 return 0;
1707
1708         if (dnode == link_own_addr(l)) {
1709                 tipc_link_bc_ack_rcv(l, acked, xmitq);
1710                 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1711                 l->stats.recv_nacks++;
1712                 return rc;
1713         }
1714
1715         /* Msg for other node => suppress own NACK at next sync if applicable */
1716         if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1717                 l->nack_state = BC_NACK_SND_SUPPRESS;
1718
1719         return 0;
1720 }
1721
1722 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1723 {
1724         int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1725
1726         l->window = win;
1727         l->backlog[TIPC_LOW_IMPORTANCE].limit      = win / 2;
1728         l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = win;
1729         l->backlog[TIPC_HIGH_IMPORTANCE].limit     = win / 2 * 3;
1730         l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = win * 2;
1731         l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
1732 }
1733
1734 /* tipc_link_find_owner - locate owner node of link by link's name
1735  * @net: the applicable net namespace
1736  * @name: pointer to link name string
1737  * @bearer_id: pointer to index in 'node->links' array where the link was found.
1738  *
1739  * Returns pointer to node owning the link, or 0 if no matching link is found.
1740  */
1741 static struct tipc_node *tipc_link_find_owner(struct net *net,
1742                                               const char *link_name,
1743                                               unsigned int *bearer_id)
1744 {
1745         struct tipc_net *tn = net_generic(net, tipc_net_id);
1746         struct tipc_link *l_ptr;
1747         struct tipc_node *n_ptr;
1748         struct tipc_node *found_node = NULL;
1749         int i;
1750
1751         *bearer_id = 0;
1752         rcu_read_lock();
1753         list_for_each_entry_rcu(n_ptr, &tn->node_list, list) {
1754                 tipc_node_lock(n_ptr);
1755                 for (i = 0; i < MAX_BEARERS; i++) {
1756                         l_ptr = n_ptr->links[i].link;
1757                         if (l_ptr && !strcmp(l_ptr->name, link_name)) {
1758                                 *bearer_id = i;
1759                                 found_node = n_ptr;
1760                                 break;
1761                         }
1762                 }
1763                 tipc_node_unlock(n_ptr);
1764                 if (found_node)
1765                         break;
1766         }
1767         rcu_read_unlock();
1768
1769         return found_node;
1770 }
1771
1772 /**
1773  * link_reset_statistics - reset link statistics
1774  * @l_ptr: pointer to link
1775  */
1776 static void link_reset_statistics(struct tipc_link *l_ptr)
1777 {
1778         memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
1779         l_ptr->stats.sent_info = l_ptr->snd_nxt;
1780         l_ptr->stats.recv_info = l_ptr->rcv_nxt;
1781 }
1782
1783 static void link_print(struct tipc_link *l, const char *str)
1784 {
1785         struct sk_buff *hskb = skb_peek(&l->transmq);
1786         u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1787         u16 tail = l->snd_nxt - 1;
1788
1789         pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1790         pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1791                 skb_queue_len(&l->transmq), head, tail,
1792                 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1793 }
1794
1795 /* Parse and validate nested (link) properties valid for media, bearer and link
1796  */
1797 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1798 {
1799         int err;
1800
1801         err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1802                                tipc_nl_prop_policy);
1803         if (err)
1804                 return err;
1805
1806         if (props[TIPC_NLA_PROP_PRIO]) {
1807                 u32 prio;
1808
1809                 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1810                 if (prio > TIPC_MAX_LINK_PRI)
1811                         return -EINVAL;
1812         }
1813
1814         if (props[TIPC_NLA_PROP_TOL]) {
1815                 u32 tol;
1816
1817                 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1818                 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1819                         return -EINVAL;
1820         }
1821
1822         if (props[TIPC_NLA_PROP_WIN]) {
1823                 u32 win;
1824
1825                 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1826                 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1827                         return -EINVAL;
1828         }
1829
1830         return 0;
1831 }
1832
1833 int tipc_nl_link_set(struct sk_buff *skb, struct genl_info *info)
1834 {
1835         int err;
1836         int res = 0;
1837         int bearer_id;
1838         char *name;
1839         struct tipc_link *link;
1840         struct tipc_node *node;
1841         struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
1842         struct net *net = sock_net(skb->sk);
1843
1844         if (!info->attrs[TIPC_NLA_LINK])
1845                 return -EINVAL;
1846
1847         err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
1848                                info->attrs[TIPC_NLA_LINK],
1849                                tipc_nl_link_policy);
1850         if (err)
1851                 return err;
1852
1853         if (!attrs[TIPC_NLA_LINK_NAME])
1854                 return -EINVAL;
1855
1856         name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
1857
1858         if (strcmp(name, tipc_bclink_name) == 0)
1859                 return tipc_nl_bc_link_set(net, attrs);
1860
1861         node = tipc_link_find_owner(net, name, &bearer_id);
1862         if (!node)
1863                 return -EINVAL;
1864
1865         tipc_node_lock(node);
1866
1867         link = node->links[bearer_id].link;
1868         if (!link) {
1869                 res = -EINVAL;
1870                 goto out;
1871         }
1872
1873         if (attrs[TIPC_NLA_LINK_PROP]) {
1874                 struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
1875
1876                 err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP],
1877                                               props);
1878                 if (err) {
1879                         res = err;
1880                         goto out;
1881                 }
1882
1883                 if (props[TIPC_NLA_PROP_TOL]) {
1884                         u32 tol;
1885
1886                         tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1887                         link->tolerance = tol;
1888                         tipc_link_proto_xmit(link, STATE_MSG, 0, 0, tol, 0);
1889                 }
1890                 if (props[TIPC_NLA_PROP_PRIO]) {
1891                         u32 prio;
1892
1893                         prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1894                         link->priority = prio;
1895                         tipc_link_proto_xmit(link, STATE_MSG, 0, 0, 0, prio);
1896                 }
1897                 if (props[TIPC_NLA_PROP_WIN]) {
1898                         u32 win;
1899
1900                         win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1901                         tipc_link_set_queue_limits(link, win);
1902                 }
1903         }
1904
1905 out:
1906         tipc_node_unlock(node);
1907
1908         return res;
1909 }
1910
1911 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1912 {
1913         int i;
1914         struct nlattr *stats;
1915
1916         struct nla_map {
1917                 u32 key;
1918                 u32 val;
1919         };
1920
1921         struct nla_map map[] = {
1922                 {TIPC_NLA_STATS_RX_INFO, s->recv_info},
1923                 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1924                 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1925                 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1926                 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1927                 {TIPC_NLA_STATS_TX_INFO, s->sent_info},
1928                 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1929                 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1930                 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1931                 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1932                 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1933                         s->msg_length_counts : 1},
1934                 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1935                 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1936                 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1937                 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1938                 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1939                 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1940                 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1941                 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1942                 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1943                 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1944                 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1945                 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1946                 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1947                 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1948                 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1949                 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1950                 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1951                 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1952                 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1953                 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1954                 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1955                 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1956                         (s->accu_queue_sz / s->queue_sz_counts) : 0}
1957         };
1958
1959         stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1960         if (!stats)
1961                 return -EMSGSIZE;
1962
1963         for (i = 0; i <  ARRAY_SIZE(map); i++)
1964                 if (nla_put_u32(skb, map[i].key, map[i].val))
1965                         goto msg_full;
1966
1967         nla_nest_end(skb, stats);
1968
1969         return 0;
1970 msg_full:
1971         nla_nest_cancel(skb, stats);
1972
1973         return -EMSGSIZE;
1974 }
1975
1976 /* Caller should hold appropriate locks to protect the link */
1977 static int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1978                               struct tipc_link *link, int nlflags)
1979 {
1980         int err;
1981         void *hdr;
1982         struct nlattr *attrs;
1983         struct nlattr *prop;
1984         struct tipc_net *tn = net_generic(net, tipc_net_id);
1985
1986         hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1987                           nlflags, TIPC_NL_LINK_GET);
1988         if (!hdr)
1989                 return -EMSGSIZE;
1990
1991         attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1992         if (!attrs)
1993                 goto msg_full;
1994
1995         if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1996                 goto attr_msg_full;
1997         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1998                         tipc_cluster_mask(tn->own_addr)))
1999                 goto attr_msg_full;
2000         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2001                 goto attr_msg_full;
2002         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->rcv_nxt))
2003                 goto attr_msg_full;
2004         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->snd_nxt))
2005                 goto attr_msg_full;
2006
2007         if (tipc_link_is_up(link))
2008                 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2009                         goto attr_msg_full;
2010         if (link->active)
2011                 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2012                         goto attr_msg_full;
2013
2014         prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
2015         if (!prop)
2016                 goto attr_msg_full;
2017         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2018                 goto prop_msg_full;
2019         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2020                 goto prop_msg_full;
2021         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2022                         link->window))
2023                 goto prop_msg_full;
2024         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2025                 goto prop_msg_full;
2026         nla_nest_end(msg->skb, prop);
2027
2028         err = __tipc_nl_add_stats(msg->skb, &link->stats);
2029         if (err)
2030                 goto attr_msg_full;
2031
2032         nla_nest_end(msg->skb, attrs);
2033         genlmsg_end(msg->skb, hdr);
2034
2035         return 0;
2036
2037 prop_msg_full:
2038         nla_nest_cancel(msg->skb, prop);
2039 attr_msg_full:
2040         nla_nest_cancel(msg->skb, attrs);
2041 msg_full:
2042         genlmsg_cancel(msg->skb, hdr);
2043
2044         return -EMSGSIZE;
2045 }
2046
2047 /* Caller should hold node lock  */
2048 static int __tipc_nl_add_node_links(struct net *net, struct tipc_nl_msg *msg,
2049                                     struct tipc_node *node, u32 *prev_link)
2050 {
2051         u32 i;
2052         int err;
2053
2054         for (i = *prev_link; i < MAX_BEARERS; i++) {
2055                 *prev_link = i;
2056
2057                 if (!node->links[i].link)
2058                         continue;
2059
2060                 err = __tipc_nl_add_link(net, msg,
2061                                          node->links[i].link, NLM_F_MULTI);
2062                 if (err)
2063                         return err;
2064         }
2065         *prev_link = 0;
2066
2067         return 0;
2068 }
2069
2070 int tipc_nl_link_dump(struct sk_buff *skb, struct netlink_callback *cb)
2071 {
2072         struct net *net = sock_net(skb->sk);
2073         struct tipc_net *tn = net_generic(net, tipc_net_id);
2074         struct tipc_node *node;
2075         struct tipc_nl_msg msg;
2076         u32 prev_node = cb->args[0];
2077         u32 prev_link = cb->args[1];
2078         int done = cb->args[2];
2079         int err;
2080
2081         if (done)
2082                 return 0;
2083
2084         msg.skb = skb;
2085         msg.portid = NETLINK_CB(cb->skb).portid;
2086         msg.seq = cb->nlh->nlmsg_seq;
2087
2088         rcu_read_lock();
2089         if (prev_node) {
2090                 node = tipc_node_find(net, prev_node);
2091                 if (!node) {
2092                         /* We never set seq or call nl_dump_check_consistent()
2093                          * this means that setting prev_seq here will cause the
2094                          * consistence check to fail in the netlink callback
2095                          * handler. Resulting in the last NLMSG_DONE message
2096                          * having the NLM_F_DUMP_INTR flag set.
2097                          */
2098                         cb->prev_seq = 1;
2099                         goto out;
2100                 }
2101                 tipc_node_put(node);
2102
2103                 list_for_each_entry_continue_rcu(node, &tn->node_list,
2104                                                  list) {
2105                         tipc_node_lock(node);
2106                         err = __tipc_nl_add_node_links(net, &msg, node,
2107                                                        &prev_link);
2108                         tipc_node_unlock(node);
2109                         if (err)
2110                                 goto out;
2111
2112                         prev_node = node->addr;
2113                 }
2114         } else {
2115                 err = tipc_nl_add_bc_link(net, &msg);
2116                 if (err)
2117                         goto out;
2118
2119                 list_for_each_entry_rcu(node, &tn->node_list, list) {
2120                         tipc_node_lock(node);
2121                         err = __tipc_nl_add_node_links(net, &msg, node,
2122                                                        &prev_link);
2123                         tipc_node_unlock(node);
2124                         if (err)
2125                                 goto out;
2126
2127                         prev_node = node->addr;
2128                 }
2129         }
2130         done = 1;
2131 out:
2132         rcu_read_unlock();
2133
2134         cb->args[0] = prev_node;
2135         cb->args[1] = prev_link;
2136         cb->args[2] = done;
2137
2138         return skb->len;
2139 }
2140
2141 int tipc_nl_link_get(struct sk_buff *skb, struct genl_info *info)
2142 {
2143         struct net *net = genl_info_net(info);
2144         struct tipc_nl_msg msg;
2145         char *name;
2146         int err;
2147
2148         msg.portid = info->snd_portid;
2149         msg.seq = info->snd_seq;
2150
2151         if (!info->attrs[TIPC_NLA_LINK_NAME])
2152                 return -EINVAL;
2153         name = nla_data(info->attrs[TIPC_NLA_LINK_NAME]);
2154
2155         msg.skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
2156         if (!msg.skb)
2157                 return -ENOMEM;
2158
2159         if (strcmp(name, tipc_bclink_name) == 0) {
2160                 err = tipc_nl_add_bc_link(net, &msg);
2161                 if (err) {
2162                         nlmsg_free(msg.skb);
2163                         return err;
2164                 }
2165         } else {
2166                 int bearer_id;
2167                 struct tipc_node *node;
2168                 struct tipc_link *link;
2169
2170                 node = tipc_link_find_owner(net, name, &bearer_id);
2171                 if (!node)
2172                         return -EINVAL;
2173
2174                 tipc_node_lock(node);
2175                 link = node->links[bearer_id].link;
2176                 if (!link) {
2177                         tipc_node_unlock(node);
2178                         nlmsg_free(msg.skb);
2179                         return -EINVAL;
2180                 }
2181
2182                 err = __tipc_nl_add_link(net, &msg, link, 0);
2183                 tipc_node_unlock(node);
2184                 if (err) {
2185                         nlmsg_free(msg.skb);
2186                         return err;
2187                 }
2188         }
2189
2190         return genlmsg_reply(msg.skb, info);
2191 }
2192
2193 int tipc_nl_link_reset_stats(struct sk_buff *skb, struct genl_info *info)
2194 {
2195         int err;
2196         char *link_name;
2197         unsigned int bearer_id;
2198         struct tipc_link *link;
2199         struct tipc_node *node;
2200         struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
2201         struct net *net = sock_net(skb->sk);
2202
2203         if (!info->attrs[TIPC_NLA_LINK])
2204                 return -EINVAL;
2205
2206         err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
2207                                info->attrs[TIPC_NLA_LINK],
2208                                tipc_nl_link_policy);
2209         if (err)
2210                 return err;
2211
2212         if (!attrs[TIPC_NLA_LINK_NAME])
2213                 return -EINVAL;
2214
2215         link_name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
2216
2217         if (strcmp(link_name, tipc_bclink_name) == 0) {
2218                 err = tipc_bclink_reset_stats(net);
2219                 if (err)
2220                         return err;
2221                 return 0;
2222         }
2223
2224         node = tipc_link_find_owner(net, link_name, &bearer_id);
2225         if (!node)
2226                 return -EINVAL;
2227
2228         tipc_node_lock(node);
2229
2230         link = node->links[bearer_id].link;
2231         if (!link) {
2232                 tipc_node_unlock(node);
2233                 return -EINVAL;
2234         }
2235
2236         link_reset_statistics(link);
2237
2238         tipc_node_unlock(node);
2239
2240         return 0;
2241 }