net/tipc/bcast.c

   1 /*
   2  * net/tipc/bcast.c: TIPC broadcast code
   3  *
   4  * Copyright (c) 2004-2006, 2014-2017, Ericsson AB
   5  * Copyright (c) 2004, Intel Corporation.
   6  * Copyright (c) 2005, 2010-2011, Wind River Systems
   7  * All rights reserved.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions are met:
  11  *
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. Neither the names of the copyright holders nor the names of its
  18  *    contributors may be used to endorse or promote products derived from
  19  *    this software without specific prior written permission.
  20  *
  21  * Alternatively, this software may be distributed under the terms of the
  22  * GNU General Public License ("GPL") version 2 as published by the Free
  23  * Software Foundation.
  24  *
  25  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  26  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  28  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  29  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  35  * POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/tipc_config.h>
  39 #include "socket.h"
  40 #include "msg.h"
  41 #include "bcast.h"
  42 #include "link.h"
  43 #include "name_table.h"
  44
  45 #define BCLINK_WIN_DEFAULT  50  /* bcast link window size (default) */
  46 #define BCLINK_WIN_MIN      32  /* bcast minimum link window size */
  47
  48 const char tipc_bclink_name[] = "broadcast-link";
  49
  50 /**
  51  * struct tipc_bc_base - base structure for keeping broadcast send state
  52  * @link: broadcast send link structure
  53  * @inputq: data input queue; will only carry SOCK_WAKEUP messages
  54  * @dests: array keeping number of reachable destinations per bearer
  55  * @primary_bearer: a bearer having links to all broadcast destinations, if any
  56  * @bcast_support: indicates if primary bearer, if any, supports broadcast
  57  * @force_bcast: forces broadcast for multicast traffic
  58  * @rcast_support: indicates if all peer nodes support replicast
  59  * @force_rcast: forces replicast for multicast traffic
  60  * @rc_ratio: dest count as percentage of cluster size where send method changes
  61  * @bc_threshold: calculated from rc_ratio; if dests > threshold use broadcast
  62  */
  63 struct tipc_bc_base {
  64         struct tipc_link *link;
  65         struct sk_buff_head inputq;
  66         int dests[MAX_BEARERS];
  67         int primary_bearer;
  68         bool bcast_support;
  69         bool force_bcast;
  70         bool rcast_support;
  71         bool force_rcast;
  72         int rc_ratio;
  73         int bc_threshold;
  74 };
  75
  76 static struct tipc_bc_base *tipc_bc_base(struct net *net)
  77 {
  78         return tipc_net(net)->bcbase;
  79 }
  80
  81 /* tipc_bcast_get_mtu(): -get the MTU currently used by broadcast link
  82  * Note: the MTU is decremented to give room for a tunnel header, in
  83  * case the message needs to be sent as replicast
  84  */
  85 int tipc_bcast_get_mtu(struct net *net)
  86 {
  87         return tipc_link_mss(tipc_bc_sndlink(net));
  88 }
  89
  90 void tipc_bcast_toggle_rcast(struct net *net, bool supp)
  91 {
  92         tipc_bc_base(net)->rcast_support = supp;
  93 }
  94
  95 static void tipc_bcbase_calc_bc_threshold(struct net *net)
  96 {
  97         struct tipc_bc_base *bb = tipc_bc_base(net);
  98         int cluster_size = tipc_link_bc_peers(tipc_bc_sndlink(net));
  99
 100         bb->bc_threshold = 1 + (cluster_size * bb->rc_ratio / 100);
 101 }
 102
 103 /* tipc_bcbase_select_primary(): find a bearer with links to all destinations,
 104  *                               if any, and make it primary bearer
 105  */
 106 static void tipc_bcbase_select_primary(struct net *net)
 107 {
 108         struct tipc_bc_base *bb = tipc_bc_base(net);
 109         int all_dests =  tipc_link_bc_peers(bb->link);
 110         int i, mtu, prim;
 111
 112         bb->primary_bearer = INVALID_BEARER_ID;
 113         bb->bcast_support = true;
 114
 115         if (!all_dests)
 116                 return;
 117
 118         for (i = 0; i < MAX_BEARERS; i++) {
 119                 if (!bb->dests[i])
 120                         continue;
 121
 122                 mtu = tipc_bearer_mtu(net, i);
 123                 if (mtu < tipc_link_mtu(bb->link))
 124                         tipc_link_set_mtu(bb->link, mtu);
 125                 bb->bcast_support &= tipc_bearer_bcast_support(net, i);
 126                 if (bb->dests[i] < all_dests)
 127                         continue;
 128
 129                 bb->primary_bearer = i;
 130
 131                 /* Reduce risk that all nodes select same primary */
 132                 if ((i ^ tipc_own_addr(net)) & 1)
 133                         break;
 134         }
 135         prim = bb->primary_bearer;
 136         if (prim != INVALID_BEARER_ID)
 137                 bb->bcast_support = tipc_bearer_bcast_support(net, prim);
 138 }
 139
 140 void tipc_bcast_inc_bearer_dst_cnt(struct net *net, int bearer_id)
 141 {
 142         struct tipc_bc_base *bb = tipc_bc_base(net);
 143
 144         tipc_bcast_lock(net);
 145         bb->dests[bearer_id]++;
 146         tipc_bcbase_select_primary(net);
 147         tipc_bcast_unlock(net);
 148 }
 149
 150 void tipc_bcast_dec_bearer_dst_cnt(struct net *net, int bearer_id)
 151 {
 152         struct tipc_bc_base *bb = tipc_bc_base(net);
 153
 154         tipc_bcast_lock(net);
 155         bb->dests[bearer_id]--;
 156         tipc_bcbase_select_primary(net);
 157         tipc_bcast_unlock(net);
 158 }
 159
 160 /* tipc_bcbase_xmit - broadcast a packet queue across one or more bearers
 161  *
 162  * Note that number of reachable destinations, as indicated in the dests[]
 163  * array, may transitionally differ from the number of destinations indicated
 164  * in each sent buffer. We can sustain this. Excess destination nodes will
 165  * drop and never acknowledge the unexpected packets, and missing destinations
 166  * will either require retransmission (if they are just about to be added to
 167  * the bearer), or be removed from the buffer's 'ackers' counter (if they
 168  * just went down)
 169  */
 170 static void tipc_bcbase_xmit(struct net *net, struct sk_buff_head *xmitq)
 171 {
 172         int bearer_id;
 173         struct tipc_bc_base *bb = tipc_bc_base(net);
 174         struct sk_buff *skb, *_skb;
 175         struct sk_buff_head _xmitq;
 176
 177         if (skb_queue_empty(xmitq))
 178                 return;
 179
 180         /* The typical case: at least one bearer has links to all nodes */
 181         bearer_id = bb->primary_bearer;
 182         if (bearer_id >= 0) {
 183                 tipc_bearer_bc_xmit(net, bearer_id, xmitq);
 184                 return;
 185         }
 186
 187         /* We have to transmit across all bearers */
 188         __skb_queue_head_init(&_xmitq);
 189         for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
 190                 if (!bb->dests[bearer_id])
 191                         continue;
 192
 193                 skb_queue_walk(xmitq, skb) {
 194                         _skb = pskb_copy_for_clone(skb, GFP_ATOMIC);
 195                         if (!_skb)
 196                                 break;
 197                         __skb_queue_tail(&_xmitq, _skb);
 198                 }
 199                 tipc_bearer_bc_xmit(net, bearer_id, &_xmitq);
 200         }
 201         __skb_queue_purge(xmitq);
 202         __skb_queue_purge(&_xmitq);
 203 }
 204
 205 static void tipc_bcast_select_xmit_method(struct net *net, int dests,
 206                                           struct tipc_mc_method *method)
 207 {
 208         struct tipc_bc_base *bb = tipc_bc_base(net);
 209         unsigned long exp = method->expires;
 210
 211         /* Broadcast supported by used bearer/bearers? */
 212         if (!bb->bcast_support) {
 213                 method->rcast = true;
 214                 return;
 215         }
 216         /* Any destinations which don't support replicast ? */
 217         if (!bb->rcast_support) {
 218                 method->rcast = false;
 219                 return;
 220         }
 221         /* Can current method be changed ? */
 222         method->expires = jiffies + TIPC_METHOD_EXPIRE;
 223         if (method->mandatory)
 224                 return;
 225
 226         if (!(tipc_net(net)->capabilities & TIPC_MCAST_RBCTL) &&
 227             time_before(jiffies, exp))
 228                 return;
 229
 230         /* Configuration as force 'broadcast' method */
 231         if (bb->force_bcast) {
 232                 method->rcast = false;
 233                 return;
 234         }
 235         /* Configuration as force 'replicast' method */
 236         if (bb->force_rcast) {
 237                 method->rcast = true;
 238                 return;
 239         }
 240         /* Configuration as 'autoselect' or default method */
 241         /* Determine method to use now */
 242         method->rcast = dests <= bb->bc_threshold;
 243 }
 244
 245 /* tipc_bcast_xmit - broadcast the buffer chain to all external nodes
 246  * @net: the applicable net namespace
 247  * @pkts: chain of buffers containing message
 248  * @cong_link_cnt: set to 1 if broadcast link is congested, otherwise 0
 249  * Consumes the buffer chain.
 250  * Returns 0 if success, otherwise errno: -EHOSTUNREACH,-EMSGSIZE
 251  */
 252 static int tipc_bcast_xmit(struct net *net, struct sk_buff_head *pkts,
 253                            u16 *cong_link_cnt)
 254 {
 255         struct tipc_link *l = tipc_bc_sndlink(net);
 256         struct sk_buff_head xmitq;
 257         int rc = 0;
 258
 259         __skb_queue_head_init(&xmitq);
 260         tipc_bcast_lock(net);
 261         if (tipc_link_bc_peers(l))
 262                 rc = tipc_link_xmit(l, pkts, &xmitq);
 263         tipc_bcast_unlock(net);
 264         tipc_bcbase_xmit(net, &xmitq);
 265         __skb_queue_purge(pkts);
 266         if (rc == -ELINKCONG) {
 267                 *cong_link_cnt = 1;
 268                 rc = 0;
 269         }
 270         return rc;
 271 }
 272
 273 /* tipc_rcast_xmit - replicate and send a message to given destination nodes
 274  * @net: the applicable net namespace
 275  * @pkts: chain of buffers containing message
 276  * @dests: list of destination nodes
 277  * @cong_link_cnt: returns number of congested links
 278  * @cong_links: returns identities of congested links
 279  * Returns 0 if success, otherwise errno
 280  */
 281 static int tipc_rcast_xmit(struct net *net, struct sk_buff_head *pkts,
 282                            struct tipc_nlist *dests, u16 *cong_link_cnt)
 283 {
 284         struct tipc_dest *dst, *tmp;
 285         struct sk_buff_head _pkts;
 286         u32 dnode, selector;
 287
 288         selector = msg_link_selector(buf_msg(skb_peek(pkts)));
 289         __skb_queue_head_init(&_pkts);
 290
 291         list_for_each_entry_safe(dst, tmp, &dests->list, list) {
 292                 dnode = dst->node;
 293                 if (!tipc_msg_pskb_copy(dnode, pkts, &_pkts))
 294                         return -ENOMEM;
 295
 296                 /* Any other return value than -ELINKCONG is ignored */
 297                 if (tipc_node_xmit(net, &_pkts, dnode, selector) == -ELINKCONG)
 298                         (*cong_link_cnt)++;
 299         }
 300         return 0;
 301 }
 302
 303 /* tipc_mcast_send_sync - deliver a dummy message with SYN bit
 304  * @net: the applicable net namespace
 305  * @skb: socket buffer to copy
 306  * @method: send method to be used
 307  * @dests: destination nodes for message.
 308  * Returns 0 if success, otherwise errno
 309  */
 310 static int tipc_mcast_send_sync(struct net *net, struct sk_buff *skb,
 311                                 struct tipc_mc_method *method,
 312                                 struct tipc_nlist *dests)
 313 {
 314         struct tipc_msg *hdr, *_hdr;
 315         struct sk_buff_head tmpq;
 316         struct sk_buff *_skb;
 317         u16 cong_link_cnt;
 318         int rc = 0;
 319
 320         /* Is a cluster supporting with new capabilities ? */
 321         if (!(tipc_net(net)->capabilities & TIPC_MCAST_RBCTL))
 322                 return 0;
 323
 324         hdr = buf_msg(skb);
 325         if (msg_user(hdr) == MSG_FRAGMENTER)
 326                 hdr = msg_inner_hdr(hdr);
 327         if (msg_type(hdr) != TIPC_MCAST_MSG)
 328                 return 0;
 329
 330         /* Allocate dummy message */
 331         _skb = tipc_buf_acquire(MCAST_H_SIZE, GFP_KERNEL);
 332         if (!_skb)
 333                 return -ENOMEM;
 334
 335         /* Preparing for 'synching' header */
 336         msg_set_syn(hdr, 1);
 337
 338         /* Copy skb's header into a dummy header */
 339         skb_copy_to_linear_data(_skb, hdr, MCAST_H_SIZE);
 340         skb_orphan(_skb);
 341
 342         /* Reverse method for dummy message */
 343         _hdr = buf_msg(_skb);
 344         msg_set_size(_hdr, MCAST_H_SIZE);
 345         msg_set_is_rcast(_hdr, !msg_is_rcast(hdr));
 346         msg_set_errcode(_hdr, TIPC_ERR_NO_PORT);
 347
 348         __skb_queue_head_init(&tmpq);
 349         __skb_queue_tail(&tmpq, _skb);
 350         if (method->rcast)
 351                 rc = tipc_bcast_xmit(net, &tmpq, &cong_link_cnt);
 352         else
 353                 rc = tipc_rcast_xmit(net, &tmpq, dests, &cong_link_cnt);
 354
 355         /* This queue should normally be empty by now */
 356         __skb_queue_purge(&tmpq);
 357
 358         return rc;
 359 }
 360
 361 /* tipc_mcast_xmit - deliver message to indicated destination nodes
 362  *                   and to identified node local sockets
 363  * @net: the applicable net namespace
 364  * @pkts: chain of buffers containing message
 365  * @method: send method to be used
 366  * @dests: destination nodes for message.
 367  * @cong_link_cnt: returns number of encountered congested destination links
 368  * Consumes buffer chain.
 369  * Returns 0 if success, otherwise errno
 370  */
 371 int tipc_mcast_xmit(struct net *net, struct sk_buff_head *pkts,
 372                     struct tipc_mc_method *method, struct tipc_nlist *dests,
 373                     u16 *cong_link_cnt)
 374 {
 375         struct sk_buff_head inputq, localq;
 376         bool rcast = method->rcast;
 377         struct tipc_msg *hdr;
 378         struct sk_buff *skb;
 379         int rc = 0;
 380
 381         skb_queue_head_init(&inputq);
 382         __skb_queue_head_init(&localq);
 383
 384         /* Clone packets before they are consumed by next call */
 385         if (dests->local && !tipc_msg_reassemble(pkts, &localq)) {
 386                 rc = -ENOMEM;
 387                 goto exit;
 388         }
 389         /* Send according to determined transmit method */
 390         if (dests->remote) {
 391                 tipc_bcast_select_xmit_method(net, dests->remote, method);
 392
 393                 skb = skb_peek(pkts);
 394                 hdr = buf_msg(skb);
 395                 if (msg_user(hdr) == MSG_FRAGMENTER)
 396                         hdr = msg_inner_hdr(hdr);
 397                 msg_set_is_rcast(hdr, method->rcast);
 398
 399                 /* Switch method ? */
 400                 if (rcast != method->rcast) {
 401                         rc = tipc_mcast_send_sync(net, skb, method, dests);
 402                         if (unlikely(rc)) {
 403                                 pr_err("Unable to send SYN: method %d, rc %d\n",
 404                                        rcast, rc);
 405                                 goto exit;
 406                         }
 407                 }
 408
 409                 if (method->rcast)
 410                         rc = tipc_rcast_xmit(net, pkts, dests, cong_link_cnt);
 411                 else
 412                         rc = tipc_bcast_xmit(net, pkts, cong_link_cnt);
 413         }
 414
 415         if (dests->local) {
 416                 tipc_loopback_trace(net, &localq);
 417                 tipc_sk_mcast_rcv(net, &localq, &inputq);
 418         }
 419 exit:
 420         /* This queue should normally be empty by now */
 421         __skb_queue_purge(pkts);
 422         return rc;
 423 }
 424
 425 /* tipc_bcast_rcv - receive a broadcast packet, and deliver to rcv link
 426  *
 427  * RCU is locked, no other locks set
 428  */
 429 int tipc_bcast_rcv(struct net *net, struct tipc_link *l, struct sk_buff *skb)
 430 {
 431         struct tipc_msg *hdr = buf_msg(skb);
 432         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 433         struct sk_buff_head xmitq;
 434         int rc;
 435
 436         __skb_queue_head_init(&xmitq);
 437
 438         if (msg_mc_netid(hdr) != tipc_netid(net) || !tipc_link_is_up(l)) {
 439                 kfree_skb(skb);
 440                 return 0;
 441         }
 442
 443         tipc_bcast_lock(net);
 444         if (msg_user(hdr) == BCAST_PROTOCOL)
 445                 rc = tipc_link_bc_nack_rcv(l, skb, &xmitq);
 446         else
 447                 rc = tipc_link_rcv(l, skb, NULL);
 448         tipc_bcast_unlock(net);
 449
 450         tipc_bcbase_xmit(net, &xmitq);
 451
 452         /* Any socket wakeup messages ? */
 453         if (!skb_queue_empty(inputq))
 454                 tipc_sk_rcv(net, inputq);
 455
 456         return rc;
 457 }
 458
 459 /* tipc_bcast_ack_rcv - receive and handle a broadcast acknowledge
 460  *
 461  * RCU is locked, no other locks set
 462  */
 463 void tipc_bcast_ack_rcv(struct net *net, struct tipc_link *l,
 464                         struct tipc_msg *hdr)
 465 {
 466         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 467         u16 acked = msg_bcast_ack(hdr);
 468         struct sk_buff_head xmitq;
 469
 470         /* Ignore bc acks sent by peer before bcast synch point was received */
 471         if (msg_bc_ack_invalid(hdr))
 472                 return;
 473
 474         __skb_queue_head_init(&xmitq);
 475
 476         tipc_bcast_lock(net);
 477         tipc_link_bc_ack_rcv(l, acked, &xmitq);
 478         tipc_bcast_unlock(net);
 479
 480         tipc_bcbase_xmit(net, &xmitq);
 481
 482         /* Any socket wakeup messages ? */
 483         if (!skb_queue_empty(inputq))
 484                 tipc_sk_rcv(net, inputq);
 485 }
 486
 487 /* tipc_bcast_synch_rcv -  check and update rcv link with peer's send state
 488  *
 489  * RCU is locked, no other locks set
 490  */
 491 int tipc_bcast_sync_rcv(struct net *net, struct tipc_link *l,
 492                         struct tipc_msg *hdr)
 493 {
 494         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 495         struct sk_buff_head xmitq;
 496         int rc = 0;
 497
 498         __skb_queue_head_init(&xmitq);
 499
 500         tipc_bcast_lock(net);
 501         if (msg_type(hdr) != STATE_MSG) {
 502                 tipc_link_bc_init_rcv(l, hdr);
 503         } else if (!msg_bc_ack_invalid(hdr)) {
 504                 tipc_link_bc_ack_rcv(l, msg_bcast_ack(hdr), &xmitq);
 505                 rc = tipc_link_bc_sync_rcv(l, hdr, &xmitq);
 506         }
 507         tipc_bcast_unlock(net);
 508
 509         tipc_bcbase_xmit(net, &xmitq);
 510
 511         /* Any socket wakeup messages ? */
 512         if (!skb_queue_empty(inputq))
 513                 tipc_sk_rcv(net, inputq);
 514         return rc;
 515 }
 516
 517 /* tipc_bcast_add_peer - add a peer node to broadcast link and bearer
 518  *
 519  * RCU is locked, node lock is set
 520  */
 521 void tipc_bcast_add_peer(struct net *net, struct tipc_link *uc_l,
 522                          struct sk_buff_head *xmitq)
 523 {
 524         struct tipc_link *snd_l = tipc_bc_sndlink(net);
 525
 526         tipc_bcast_lock(net);
 527         tipc_link_add_bc_peer(snd_l, uc_l, xmitq);
 528         tipc_bcbase_select_primary(net);
 529         tipc_bcbase_calc_bc_threshold(net);
 530         tipc_bcast_unlock(net);
 531 }
 532
 533 /* tipc_bcast_remove_peer - remove a peer node from broadcast link and bearer
 534  *
 535  * RCU is locked, node lock is set
 536  */
 537 void tipc_bcast_remove_peer(struct net *net, struct tipc_link *rcv_l)
 538 {
 539         struct tipc_link *snd_l = tipc_bc_sndlink(net);
 540         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 541         struct sk_buff_head xmitq;
 542
 543         __skb_queue_head_init(&xmitq);
 544
 545         tipc_bcast_lock(net);
 546         tipc_link_remove_bc_peer(snd_l, rcv_l, &xmitq);
 547         tipc_bcbase_select_primary(net);
 548         tipc_bcbase_calc_bc_threshold(net);
 549         tipc_bcast_unlock(net);
 550
 551         tipc_bcbase_xmit(net, &xmitq);
 552
 553         /* Any socket wakeup messages ? */
 554         if (!skb_queue_empty(inputq))
 555                 tipc_sk_rcv(net, inputq);
 556 }
 557
 558 int tipc_bclink_reset_stats(struct net *net)
 559 {
 560         struct tipc_link *l = tipc_bc_sndlink(net);
 561
 562         if (!l)
 563                 return -ENOPROTOOPT;
 564
 565         tipc_bcast_lock(net);
 566         tipc_link_reset_stats(l);
 567         tipc_bcast_unlock(net);
 568         return 0;
 569 }
 570
 571 static int tipc_bc_link_set_queue_limits(struct net *net, u32 max_win)
 572 {
 573         struct tipc_link *l = tipc_bc_sndlink(net);
 574
 575         if (!l)
 576                 return -ENOPROTOOPT;
 577         if (max_win < BCLINK_WIN_MIN)
 578                 max_win = BCLINK_WIN_MIN;
 579         if (max_win > TIPC_MAX_LINK_WIN)
 580                 return -EINVAL;
 581         tipc_bcast_lock(net);
 582         tipc_link_set_queue_limits(l, BCLINK_WIN_MIN, max_win);
 583         tipc_bcast_unlock(net);
 584         return 0;
 585 }
 586
 587 static int tipc_bc_link_set_broadcast_mode(struct net *net, u32 bc_mode)
 588 {
 589         struct tipc_bc_base *bb = tipc_bc_base(net);
 590
 591         switch (bc_mode) {
 592         case BCLINK_MODE_BCAST:
 593                 if (!bb->bcast_support)
 594                         return -ENOPROTOOPT;
 595
 596                 bb->force_bcast = true;
 597                 bb->force_rcast = false;
 598                 break;
 599         case BCLINK_MODE_RCAST:
 600                 if (!bb->rcast_support)
 601                         return -ENOPROTOOPT;
 602
 603                 bb->force_bcast = false;
 604                 bb->force_rcast = true;
 605                 break;
 606         case BCLINK_MODE_SEL:
 607                 if (!bb->bcast_support || !bb->rcast_support)
 608                         return -ENOPROTOOPT;
 609
 610                 bb->force_bcast = false;
 611                 bb->force_rcast = false;
 612                 break;
 613         default:
 614                 return -EINVAL;
 615         }
 616
 617         return 0;
 618 }
 619
 620 static int tipc_bc_link_set_broadcast_ratio(struct net *net, u32 bc_ratio)
 621 {
 622         struct tipc_bc_base *bb = tipc_bc_base(net);
 623
 624         if (!bb->bcast_support || !bb->rcast_support)
 625                 return -ENOPROTOOPT;
 626
 627         if (bc_ratio > 100 || bc_ratio <= 0)
 628                 return -EINVAL;
 629
 630         bb->rc_ratio = bc_ratio;
 631         tipc_bcast_lock(net);
 632         tipc_bcbase_calc_bc_threshold(net);
 633         tipc_bcast_unlock(net);
 634
 635         return 0;
 636 }
 637
 638 int tipc_nl_bc_link_set(struct net *net, struct nlattr *attrs[])
 639 {
 640         int err;
 641         u32 win;
 642         u32 bc_mode;
 643         u32 bc_ratio;
 644         struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
 645
 646         if (!attrs[TIPC_NLA_LINK_PROP])
 647                 return -EINVAL;
 648
 649         err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP], props);
 650         if (err)
 651                 return err;
 652
 653         if (!props[TIPC_NLA_PROP_WIN] &&
 654             !props[TIPC_NLA_PROP_BROADCAST] &&
 655             !props[TIPC_NLA_PROP_BROADCAST_RATIO]) {
 656                 return -EOPNOTSUPP;
 657         }
 658
 659         if (props[TIPC_NLA_PROP_BROADCAST]) {
 660                 bc_mode = nla_get_u32(props[TIPC_NLA_PROP_BROADCAST]);
 661                 err = tipc_bc_link_set_broadcast_mode(net, bc_mode);
 662         }
 663
 664         if (!err && props[TIPC_NLA_PROP_BROADCAST_RATIO]) {
 665                 bc_ratio = nla_get_u32(props[TIPC_NLA_PROP_BROADCAST_RATIO]);
 666                 err = tipc_bc_link_set_broadcast_ratio(net, bc_ratio);
 667         }
 668
 669         if (!err && props[TIPC_NLA_PROP_WIN]) {
 670                 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
 671                 err = tipc_bc_link_set_queue_limits(net, win);
 672         }
 673
 674         return err;
 675 }
 676
 677 int tipc_bcast_init(struct net *net)
 678 {
 679         struct tipc_net *tn = tipc_net(net);
 680         struct tipc_bc_base *bb = NULL;
 681         struct tipc_link *l = NULL;
 682
 683         bb = kzalloc(sizeof(*bb), GFP_KERNEL);
 684         if (!bb)
 685                 goto enomem;
 686         tn->bcbase = bb;
 687         spin_lock_init(&tipc_net(net)->bclock);
 688
 689         if (!tipc_link_bc_create(net, 0, 0,
 690                                  FB_MTU,
 691                                  BCLINK_WIN_DEFAULT,
 692                                  BCLINK_WIN_DEFAULT,
 693                                  0,
 694                                  &bb->inputq,
 695                                  NULL,
 696                                  NULL,
 697                                  &l))
 698                 goto enomem;
 699         bb->link = l;
 700         tn->bcl = l;
 701         bb->rc_ratio = 10;
 702         bb->rcast_support = true;
 703         return 0;
 704 enomem:
 705         kfree(bb);
 706         kfree(l);
 707         return -ENOMEM;
 708 }
 709
 710 void tipc_bcast_stop(struct net *net)
 711 {
 712         struct tipc_net *tn = net_generic(net, tipc_net_id);
 713
 714         synchronize_net();
 715         kfree(tn->bcbase);
 716         kfree(tn->bcl);
 717 }
 718
 719 void tipc_nlist_init(struct tipc_nlist *nl, u32 self)
 720 {
 721         memset(nl, 0, sizeof(*nl));
 722         INIT_LIST_HEAD(&nl->list);
 723         nl->self = self;
 724 }
 725
 726 void tipc_nlist_add(struct tipc_nlist *nl, u32 node)
 727 {
 728         if (node == nl->self)
 729                 nl->local = true;
 730         else if (tipc_dest_push(&nl->list, node, 0))
 731                 nl->remote++;
 732 }
 733
 734 void tipc_nlist_del(struct tipc_nlist *nl, u32 node)
 735 {
 736         if (node == nl->self)
 737                 nl->local = false;
 738         else if (tipc_dest_del(&nl->list, node, 0))
 739                 nl->remote--;
 740 }
 741
 742 void tipc_nlist_purge(struct tipc_nlist *nl)
 743 {
 744         tipc_dest_list_purge(&nl->list);
 745         nl->remote = 0;
 746         nl->local = false;
 747 }
 748
 749 u32 tipc_bcast_get_broadcast_mode(struct net *net)
 750 {
 751         struct tipc_bc_base *bb = tipc_bc_base(net);
 752
 753         if (bb->force_bcast)
 754                 return BCLINK_MODE_BCAST;
 755
 756         if (bb->force_rcast)
 757                 return BCLINK_MODE_RCAST;
 758
 759         if (bb->bcast_support && bb->rcast_support)
 760                 return BCLINK_MODE_SEL;
 761
 762         return 0;
 763 }
 764
 765 u32 tipc_bcast_get_broadcast_ratio(struct net *net)
 766 {
 767         struct tipc_bc_base *bb = tipc_bc_base(net);
 768
 769         return bb->rc_ratio;
 770 }
 771
 772 void tipc_mcast_filter_msg(struct net *net, struct sk_buff_head *defq,
 773                            struct sk_buff_head *inputq)
 774 {
 775         struct sk_buff *skb, *_skb, *tmp;
 776         struct tipc_msg *hdr, *_hdr;
 777         bool match = false;
 778         u32 node, port;
 779
 780         skb = skb_peek(inputq);
 781         if (!skb)
 782                 return;
 783
 784         hdr = buf_msg(skb);
 785
 786         if (likely(!msg_is_syn(hdr) && skb_queue_empty(defq)))
 787                 return;
 788
 789         node = msg_orignode(hdr);
 790         if (node == tipc_own_addr(net))
 791                 return;
 792
 793         port = msg_origport(hdr);
 794
 795         /* Has the twin SYN message already arrived ? */
 796         skb_queue_walk(defq, _skb) {
 797                 _hdr = buf_msg(_skb);
 798                 if (msg_orignode(_hdr) != node)
 799                         continue;
 800                 if (msg_origport(_hdr) != port)
 801                         continue;
 802                 match = true;
 803                 break;
 804         }
 805
 806         if (!match) {
 807                 if (!msg_is_syn(hdr))
 808                         return;
 809                 __skb_dequeue(inputq);
 810                 __skb_queue_tail(defq, skb);
 811                 return;
 812         }
 813
 814         /* Deliver non-SYN message from other link, otherwise queue it */
 815         if (!msg_is_syn(hdr)) {
 816                 if (msg_is_rcast(hdr) != msg_is_rcast(_hdr))
 817                         return;
 818                 __skb_dequeue(inputq);
 819                 __skb_queue_tail(defq, skb);
 820                 return;
 821         }
 822
 823         /* Queue non-SYN/SYN message from same link */
 824         if (msg_is_rcast(hdr) == msg_is_rcast(_hdr)) {
 825                 __skb_dequeue(inputq);
 826                 __skb_queue_tail(defq, skb);
 827                 return;
 828         }
 829
 830         /* Matching SYN messages => return the one with data, if any */
 831         __skb_unlink(_skb, defq);
 832         if (msg_data_sz(hdr)) {
 833                 kfree_skb(_skb);
 834         } else {
 835                 __skb_dequeue(inputq);
 836                 kfree_skb(skb);
 837                 __skb_queue_tail(inputq, _skb);
 838         }
 839
 840         /* Deliver subsequent non-SYN messages from same peer */
 841         skb_queue_walk_safe(defq, _skb, tmp) {
 842                 _hdr = buf_msg(_skb);
 843                 if (msg_orignode(_hdr) != node)
 844                         continue;
 845                 if (msg_origport(_hdr) != port)
 846                         continue;
 847                 if (msg_is_syn(_hdr))
 848                         break;
 849                 __skb_unlink(_skb, defq);
 850                 __skb_queue_tail(inputq, _skb);
 851         }
 852 }