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_mtu(tipc_bc_sndlink(net)) - INT_H_SIZE;
  88 }
  89
  90 void tipc_bcast_disable_rcast(struct net *net)
  91 {
  92         tipc_bc_base(net)->rcast_support = false;
  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  * @cong_link_cnt: returns number of encountered congested destination links
 309  * Returns 0 if success, otherwise errno
 310  */
 311 static int tipc_mcast_send_sync(struct net *net, struct sk_buff *skb,
 312                                 struct tipc_mc_method *method,
 313                                 struct tipc_nlist *dests,
 314                                 u16 *cong_link_cnt)
 315 {
 316         struct tipc_msg *hdr, *_hdr;
 317         struct sk_buff_head tmpq;
 318         struct sk_buff *_skb;
 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
 347         __skb_queue_head_init(&tmpq);
 348         __skb_queue_tail(&tmpq, _skb);
 349         if (method->rcast)
 350                 tipc_bcast_xmit(net, &tmpq, cong_link_cnt);
 351         else
 352                 tipc_rcast_xmit(net, &tmpq, dests, cong_link_cnt);
 353
 354         /* This queue should normally be empty by now */
 355         __skb_queue_purge(&tmpq);
 356
 357         return 0;
 358 }
 359
 360 /* tipc_mcast_xmit - deliver message to indicated destination nodes
 361  *                   and to identified node local sockets
 362  * @net: the applicable net namespace
 363  * @pkts: chain of buffers containing message
 364  * @method: send method to be used
 365  * @dests: destination nodes for message.
 366  * @cong_link_cnt: returns number of encountered congested destination links
 367  * Consumes buffer chain.
 368  * Returns 0 if success, otherwise errno
 369  */
 370 int tipc_mcast_xmit(struct net *net, struct sk_buff_head *pkts,
 371                     struct tipc_mc_method *method, struct tipc_nlist *dests,
 372                     u16 *cong_link_cnt)
 373 {
 374         struct sk_buff_head inputq, localq;
 375         bool rcast = method->rcast;
 376         struct tipc_msg *hdr;
 377         struct sk_buff *skb;
 378         int rc = 0;
 379
 380         skb_queue_head_init(&inputq);
 381         __skb_queue_head_init(&localq);
 382
 383         /* Clone packets before they are consumed by next call */
 384         if (dests->local && !tipc_msg_reassemble(pkts, &localq)) {
 385                 rc = -ENOMEM;
 386                 goto exit;
 387         }
 388         /* Send according to determined transmit method */
 389         if (dests->remote) {
 390                 tipc_bcast_select_xmit_method(net, dests->remote, method);
 391
 392                 skb = skb_peek(pkts);
 393                 hdr = buf_msg(skb);
 394                 if (msg_user(hdr) == MSG_FRAGMENTER)
 395                         hdr = msg_inner_hdr(hdr);
 396                 msg_set_is_rcast(hdr, method->rcast);
 397
 398                 /* Switch method ? */
 399                 if (rcast != method->rcast)
 400                         tipc_mcast_send_sync(net, skb, method,
 401                                              dests, cong_link_cnt);
 402
 403                 if (method->rcast)
 404                         rc = tipc_rcast_xmit(net, pkts, dests, cong_link_cnt);
 405                 else
 406                         rc = tipc_bcast_xmit(net, pkts, cong_link_cnt);
 407         }
 408
 409         if (dests->local) {
 410                 tipc_loopback_trace(net, &localq);
 411                 tipc_sk_mcast_rcv(net, &localq, &inputq);
 412         }
 413 exit:
 414         /* This queue should normally be empty by now */
 415         __skb_queue_purge(pkts);
 416         return rc;
 417 }
 418
 419 /* tipc_bcast_rcv - receive a broadcast packet, and deliver to rcv link
 420  *
 421  * RCU is locked, no other locks set
 422  */
 423 int tipc_bcast_rcv(struct net *net, struct tipc_link *l, struct sk_buff *skb)
 424 {
 425         struct tipc_msg *hdr = buf_msg(skb);
 426         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 427         struct sk_buff_head xmitq;
 428         int rc;
 429
 430         __skb_queue_head_init(&xmitq);
 431
 432         if (msg_mc_netid(hdr) != tipc_netid(net) || !tipc_link_is_up(l)) {
 433                 kfree_skb(skb);
 434                 return 0;
 435         }
 436
 437         tipc_bcast_lock(net);
 438         if (msg_user(hdr) == BCAST_PROTOCOL)
 439                 rc = tipc_link_bc_nack_rcv(l, skb, &xmitq);
 440         else
 441                 rc = tipc_link_rcv(l, skb, NULL);
 442         tipc_bcast_unlock(net);
 443
 444         tipc_bcbase_xmit(net, &xmitq);
 445
 446         /* Any socket wakeup messages ? */
 447         if (!skb_queue_empty(inputq))
 448                 tipc_sk_rcv(net, inputq);
 449
 450         return rc;
 451 }
 452
 453 /* tipc_bcast_ack_rcv - receive and handle a broadcast acknowledge
 454  *
 455  * RCU is locked, no other locks set
 456  */
 457 void tipc_bcast_ack_rcv(struct net *net, struct tipc_link *l,
 458                         struct tipc_msg *hdr)
 459 {
 460         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 461         u16 acked = msg_bcast_ack(hdr);
 462         struct sk_buff_head xmitq;
 463
 464         /* Ignore bc acks sent by peer before bcast synch point was received */
 465         if (msg_bc_ack_invalid(hdr))
 466                 return;
 467
 468         __skb_queue_head_init(&xmitq);
 469
 470         tipc_bcast_lock(net);
 471         tipc_link_bc_ack_rcv(l, acked, &xmitq);
 472         tipc_bcast_unlock(net);
 473
 474         tipc_bcbase_xmit(net, &xmitq);
 475
 476         /* Any socket wakeup messages ? */
 477         if (!skb_queue_empty(inputq))
 478                 tipc_sk_rcv(net, inputq);
 479 }
 480
 481 /* tipc_bcast_synch_rcv -  check and update rcv link with peer's send state
 482  *
 483  * RCU is locked, no other locks set
 484  */
 485 int tipc_bcast_sync_rcv(struct net *net, struct tipc_link *l,
 486                         struct tipc_msg *hdr)
 487 {
 488         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 489         struct sk_buff_head xmitq;
 490         int rc = 0;
 491
 492         __skb_queue_head_init(&xmitq);
 493
 494         tipc_bcast_lock(net);
 495         if (msg_type(hdr) != STATE_MSG) {
 496                 tipc_link_bc_init_rcv(l, hdr);
 497         } else if (!msg_bc_ack_invalid(hdr)) {
 498                 tipc_link_bc_ack_rcv(l, msg_bcast_ack(hdr), &xmitq);
 499                 rc = tipc_link_bc_sync_rcv(l, hdr, &xmitq);
 500         }
 501         tipc_bcast_unlock(net);
 502
 503         tipc_bcbase_xmit(net, &xmitq);
 504
 505         /* Any socket wakeup messages ? */
 506         if (!skb_queue_empty(inputq))
 507                 tipc_sk_rcv(net, inputq);
 508         return rc;
 509 }
 510
 511 /* tipc_bcast_add_peer - add a peer node to broadcast link and bearer
 512  *
 513  * RCU is locked, node lock is set
 514  */
 515 void tipc_bcast_add_peer(struct net *net, struct tipc_link *uc_l,
 516                          struct sk_buff_head *xmitq)
 517 {
 518         struct tipc_link *snd_l = tipc_bc_sndlink(net);
 519
 520         tipc_bcast_lock(net);
 521         tipc_link_add_bc_peer(snd_l, uc_l, xmitq);
 522         tipc_bcbase_select_primary(net);
 523         tipc_bcbase_calc_bc_threshold(net);
 524         tipc_bcast_unlock(net);
 525 }
 526
 527 /* tipc_bcast_remove_peer - remove a peer node from broadcast link and bearer
 528  *
 529  * RCU is locked, node lock is set
 530  */
 531 void tipc_bcast_remove_peer(struct net *net, struct tipc_link *rcv_l)
 532 {
 533         struct tipc_link *snd_l = tipc_bc_sndlink(net);
 534         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 535         struct sk_buff_head xmitq;
 536
 537         __skb_queue_head_init(&xmitq);
 538
 539         tipc_bcast_lock(net);
 540         tipc_link_remove_bc_peer(snd_l, rcv_l, &xmitq);
 541         tipc_bcbase_select_primary(net);
 542         tipc_bcbase_calc_bc_threshold(net);
 543         tipc_bcast_unlock(net);
 544
 545         tipc_bcbase_xmit(net, &xmitq);
 546
 547         /* Any socket wakeup messages ? */
 548         if (!skb_queue_empty(inputq))
 549                 tipc_sk_rcv(net, inputq);
 550 }
 551
 552 int tipc_bclink_reset_stats(struct net *net)
 553 {
 554         struct tipc_link *l = tipc_bc_sndlink(net);
 555
 556         if (!l)
 557                 return -ENOPROTOOPT;
 558
 559         tipc_bcast_lock(net);
 560         tipc_link_reset_stats(l);
 561         tipc_bcast_unlock(net);
 562         return 0;
 563 }
 564
 565 static int tipc_bc_link_set_queue_limits(struct net *net, u32 limit)
 566 {
 567         struct tipc_link *l = tipc_bc_sndlink(net);
 568
 569         if (!l)
 570                 return -ENOPROTOOPT;
 571         if (limit < BCLINK_WIN_MIN)
 572                 limit = BCLINK_WIN_MIN;
 573         if (limit > TIPC_MAX_LINK_WIN)
 574                 return -EINVAL;
 575         tipc_bcast_lock(net);
 576         tipc_link_set_queue_limits(l, limit);
 577         tipc_bcast_unlock(net);
 578         return 0;
 579 }
 580
 581 static int tipc_bc_link_set_broadcast_mode(struct net *net, u32 bc_mode)
 582 {
 583         struct tipc_bc_base *bb = tipc_bc_base(net);
 584
 585         switch (bc_mode) {
 586         case BCLINK_MODE_BCAST:
 587                 if (!bb->bcast_support)
 588                         return -ENOPROTOOPT;
 589
 590                 bb->force_bcast = true;
 591                 bb->force_rcast = false;
 592                 break;
 593         case BCLINK_MODE_RCAST:
 594                 if (!bb->rcast_support)
 595                         return -ENOPROTOOPT;
 596
 597                 bb->force_bcast = false;
 598                 bb->force_rcast = true;
 599                 break;
 600         case BCLINK_MODE_SEL:
 601                 if (!bb->bcast_support || !bb->rcast_support)
 602                         return -ENOPROTOOPT;
 603
 604                 bb->force_bcast = false;
 605                 bb->force_rcast = false;
 606                 break;
 607         default:
 608                 return -EINVAL;
 609         }
 610
 611         return 0;
 612 }
 613
 614 static int tipc_bc_link_set_broadcast_ratio(struct net *net, u32 bc_ratio)
 615 {
 616         struct tipc_bc_base *bb = tipc_bc_base(net);
 617
 618         if (!bb->bcast_support || !bb->rcast_support)
 619                 return -ENOPROTOOPT;
 620
 621         if (bc_ratio > 100 || bc_ratio <= 0)
 622                 return -EINVAL;
 623
 624         bb->rc_ratio = bc_ratio;
 625         tipc_bcast_lock(net);
 626         tipc_bcbase_calc_bc_threshold(net);
 627         tipc_bcast_unlock(net);
 628
 629         return 0;
 630 }
 631
 632 int tipc_nl_bc_link_set(struct net *net, struct nlattr *attrs[])
 633 {
 634         int err;
 635         u32 win;
 636         u32 bc_mode;
 637         u32 bc_ratio;
 638         struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
 639
 640         if (!attrs[TIPC_NLA_LINK_PROP])
 641                 return -EINVAL;
 642
 643         err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP], props);
 644         if (err)
 645                 return err;
 646
 647         if (!props[TIPC_NLA_PROP_WIN] &&
 648             !props[TIPC_NLA_PROP_BROADCAST] &&
 649             !props[TIPC_NLA_PROP_BROADCAST_RATIO]) {
 650                 return -EOPNOTSUPP;
 651         }
 652
 653         if (props[TIPC_NLA_PROP_BROADCAST]) {
 654                 bc_mode = nla_get_u32(props[TIPC_NLA_PROP_BROADCAST]);
 655                 err = tipc_bc_link_set_broadcast_mode(net, bc_mode);
 656         }
 657
 658         if (!err && props[TIPC_NLA_PROP_BROADCAST_RATIO]) {
 659                 bc_ratio = nla_get_u32(props[TIPC_NLA_PROP_BROADCAST_RATIO]);
 660                 err = tipc_bc_link_set_broadcast_ratio(net, bc_ratio);
 661         }
 662
 663         if (!err && props[TIPC_NLA_PROP_WIN]) {
 664                 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
 665                 err = tipc_bc_link_set_queue_limits(net, win);
 666         }
 667
 668         return err;
 669 }
 670
 671 int tipc_bcast_init(struct net *net)
 672 {
 673         struct tipc_net *tn = tipc_net(net);
 674         struct tipc_bc_base *bb = NULL;
 675         struct tipc_link *l = NULL;
 676
 677         bb = kzalloc(sizeof(*bb), GFP_KERNEL);
 678         if (!bb)
 679                 goto enomem;
 680         tn->bcbase = bb;
 681         spin_lock_init(&tipc_net(net)->bclock);
 682
 683         if (!tipc_link_bc_create(net, 0, 0,
 684                                  FB_MTU,
 685                                  BCLINK_WIN_DEFAULT,
 686                                  0,
 687                                  &bb->inputq,
 688                                  NULL,
 689                                  NULL,
 690                                  &l))
 691                 goto enomem;
 692         bb->link = l;
 693         tn->bcl = l;
 694         bb->rc_ratio = 10;
 695         bb->rcast_support = true;
 696         return 0;
 697 enomem:
 698         kfree(bb);
 699         kfree(l);
 700         return -ENOMEM;
 701 }
 702
 703 void tipc_bcast_stop(struct net *net)
 704 {
 705         struct tipc_net *tn = net_generic(net, tipc_net_id);
 706
 707         synchronize_net();
 708         kfree(tn->bcbase);
 709         kfree(tn->bcl);
 710 }
 711
 712 void tipc_nlist_init(struct tipc_nlist *nl, u32 self)
 713 {
 714         memset(nl, 0, sizeof(*nl));
 715         INIT_LIST_HEAD(&nl->list);
 716         nl->self = self;
 717 }
 718
 719 void tipc_nlist_add(struct tipc_nlist *nl, u32 node)
 720 {
 721         if (node == nl->self)
 722                 nl->local = true;
 723         else if (tipc_dest_push(&nl->list, node, 0))
 724                 nl->remote++;
 725 }
 726
 727 void tipc_nlist_del(struct tipc_nlist *nl, u32 node)
 728 {
 729         if (node == nl->self)
 730                 nl->local = false;
 731         else if (tipc_dest_del(&nl->list, node, 0))
 732                 nl->remote--;
 733 }
 734
 735 void tipc_nlist_purge(struct tipc_nlist *nl)
 736 {
 737         tipc_dest_list_purge(&nl->list);
 738         nl->remote = 0;
 739         nl->local = false;
 740 }
 741
 742 u32 tipc_bcast_get_broadcast_mode(struct net *net)
 743 {
 744         struct tipc_bc_base *bb = tipc_bc_base(net);
 745
 746         if (bb->force_bcast)
 747                 return BCLINK_MODE_BCAST;
 748
 749         if (bb->force_rcast)
 750                 return BCLINK_MODE_RCAST;
 751
 752         if (bb->bcast_support && bb->rcast_support)
 753                 return BCLINK_MODE_SEL;
 754
 755         return 0;
 756 }
 757
 758 u32 tipc_bcast_get_broadcast_ratio(struct net *net)
 759 {
 760         struct tipc_bc_base *bb = tipc_bc_base(net);
 761
 762         return bb->rc_ratio;
 763 }
 764
 765 void tipc_mcast_filter_msg(struct net *net, struct sk_buff_head *defq,
 766                            struct sk_buff_head *inputq)
 767 {
 768         struct sk_buff *skb, *_skb, *tmp;
 769         struct tipc_msg *hdr, *_hdr;
 770         bool match = false;
 771         u32 node, port;
 772
 773         skb = skb_peek(inputq);
 774         if (!skb)
 775                 return;
 776
 777         hdr = buf_msg(skb);
 778
 779         if (likely(!msg_is_syn(hdr) && skb_queue_empty(defq)))
 780                 return;
 781
 782         node = msg_orignode(hdr);
 783         if (node == tipc_own_addr(net))
 784                 return;
 785
 786         port = msg_origport(hdr);
 787
 788         /* Has the twin SYN message already arrived ? */
 789         skb_queue_walk(defq, _skb) {
 790                 _hdr = buf_msg(_skb);
 791                 if (msg_orignode(_hdr) != node)
 792                         continue;
 793                 if (msg_origport(_hdr) != port)
 794                         continue;
 795                 match = true;
 796                 break;
 797         }
 798
 799         if (!match) {
 800                 if (!msg_is_syn(hdr))
 801                         return;
 802                 __skb_dequeue(inputq);
 803                 __skb_queue_tail(defq, skb);
 804                 return;
 805         }
 806
 807         /* Deliver non-SYN message from other link, otherwise queue it */
 808         if (!msg_is_syn(hdr)) {
 809                 if (msg_is_rcast(hdr) != msg_is_rcast(_hdr))
 810                         return;
 811                 __skb_dequeue(inputq);
 812                 __skb_queue_tail(defq, skb);
 813                 return;
 814         }
 815
 816         /* Queue non-SYN/SYN message from same link */
 817         if (msg_is_rcast(hdr) == msg_is_rcast(_hdr)) {
 818                 __skb_dequeue(inputq);
 819                 __skb_queue_tail(defq, skb);
 820                 return;
 821         }
 822
 823         /* Matching SYN messages => return the one with data, if any */
 824         __skb_unlink(_skb, defq);
 825         if (msg_data_sz(hdr)) {
 826                 kfree_skb(_skb);
 827         } else {
 828                 __skb_dequeue(inputq);
 829                 kfree_skb(skb);
 830                 __skb_queue_tail(inputq, _skb);
 831         }
 832
 833         /* Deliver subsequent non-SYN messages from same peer */
 834         skb_queue_walk_safe(defq, _skb, tmp) {
 835                 _hdr = buf_msg(_skb);
 836                 if (msg_orignode(_hdr) != node)
 837                         continue;
 838                 if (msg_origport(_hdr) != port)
 839                         continue;
 840                 if (msg_is_syn(_hdr))
 841                         break;
 842                 __skb_unlink(_skb, defq);
 843                 __skb_queue_tail(inputq, _skb);
 844         }
 845 }