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