net/tipc/bcast.c

   1 /*
   2  * net/tipc/bcast.c: TIPC broadcast code
   3  *
   4  * Copyright (c) 2004-2006, 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 "core.h"
  39 #include "link.h"
  40 #include "port.h"
  41 #include "bcast.h"
  42 #include "name_distr.h"
  43
  44 #define MAX_PKT_DEFAULT_MCAST 1500      /* bcast link max packet size (fixed) */
  45
  46 #define BCLINK_WIN_DEFAULT 20           /* bcast link window size (default) */
  47
  48 /**
  49  * struct tipc_bcbearer_pair - a pair of bearers used by broadcast link
  50  * @primary: pointer to primary bearer
  51  * @secondary: pointer to secondary bearer
  52  *
  53  * Bearers must have same priority and same set of reachable destinations
  54  * to be paired.
  55  */
  56
  57 struct tipc_bcbearer_pair {
  58         struct tipc_bearer *primary;
  59         struct tipc_bearer *secondary;
  60 };
  61
  62 /**
  63  * struct tipc_bcbearer - bearer used by broadcast link
  64  * @bearer: (non-standard) broadcast bearer structure
  65  * @media: (non-standard) broadcast media structure
  66  * @bpairs: array of bearer pairs
  67  * @bpairs_temp: temporary array of bearer pairs used by tipc_bcbearer_sort()
  68  * @remains: temporary node map used by tipc_bcbearer_send()
  69  * @remains_new: temporary node map used tipc_bcbearer_send()
  70  *
  71  * Note: The fields labelled "temporary" are incorporated into the bearer
  72  * to avoid consuming potentially limited stack space through the use of
  73  * large local variables within multicast routines.  Concurrent access is
  74  * prevented through use of the spinlock "bc_lock".
  75  */
  76
  77 struct tipc_bcbearer {
  78         struct tipc_bearer bearer;
  79         struct tipc_media media;
  80         struct tipc_bcbearer_pair bpairs[MAX_BEARERS];
  81         struct tipc_bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1];
  82         struct tipc_node_map remains;
  83         struct tipc_node_map remains_new;
  84 };
  85
  86 /**
  87  * struct tipc_bclink - link used for broadcast messages
  88  * @link: (non-standard) broadcast link structure
  89  * @node: (non-standard) node structure representing b'cast link's peer node
  90  * @bcast_nodes: map of broadcast-capable nodes
  91  * @retransmit_to: node that most recently requested a retransmit
  92  *
  93  * Handles sequence numbering, fragmentation, bundling, etc.
  94  */
  95
  96 struct tipc_bclink {
  97         struct tipc_link link;
  98         struct tipc_node node;
  99         struct tipc_node_map bcast_nodes;
 100         struct tipc_node *retransmit_to;
 101 };
 102
 103 static struct tipc_bcbearer bcast_bearer;
 104 static struct tipc_bclink bcast_link;
 105
 106 static struct tipc_bcbearer *bcbearer = &bcast_bearer;
 107 static struct tipc_bclink *bclink = &bcast_link;
 108 static struct tipc_link *bcl = &bcast_link.link;
 109
 110 static DEFINE_SPINLOCK(bc_lock);
 111
 112 const char tipc_bclink_name[] = "broadcast-link";
 113
 114 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
 115                            struct tipc_node_map *nm_b,
 116                            struct tipc_node_map *nm_diff);
 117
 118 static u32 bcbuf_acks(struct sk_buff *buf)
 119 {
 120         return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
 121 }
 122
 123 static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
 124 {
 125         TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
 126 }
 127
 128 static void bcbuf_decr_acks(struct sk_buff *buf)
 129 {
 130         bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
 131 }
 132
 133 void tipc_bclink_add_node(u32 addr)
 134 {
 135         spin_lock_bh(&bc_lock);
 136         tipc_nmap_add(&bclink->bcast_nodes, addr);
 137         spin_unlock_bh(&bc_lock);
 138 }
 139
 140 void tipc_bclink_remove_node(u32 addr)
 141 {
 142         spin_lock_bh(&bc_lock);
 143         tipc_nmap_remove(&bclink->bcast_nodes, addr);
 144         spin_unlock_bh(&bc_lock);
 145 }
 146
 147 static void bclink_set_last_sent(void)
 148 {
 149         if (bcl->next_out)
 150                 bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
 151         else
 152                 bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
 153 }
 154
 155 u32 tipc_bclink_get_last_sent(void)
 156 {
 157         return bcl->fsm_msg_cnt;
 158 }
 159
 160 /**
 161  * bclink_set_gap - set gap according to contents of current deferred pkt queue
 162  *
 163  * Called with 'node' locked, bc_lock unlocked
 164  */
 165
 166 static void bclink_set_gap(struct tipc_node *n_ptr)
 167 {
 168         struct sk_buff *buf = n_ptr->bclink.deferred_head;
 169
 170         n_ptr->bclink.gap_after = n_ptr->bclink.gap_to =
 171                 mod(n_ptr->bclink.last_in);
 172         if (unlikely(buf != NULL))
 173                 n_ptr->bclink.gap_to = mod(buf_seqno(buf) - 1);
 174 }
 175
 176 /**
 177  * bclink_ack_allowed - test if ACK or NACK message can be sent at this moment
 178  *
 179  * This mechanism endeavours to prevent all nodes in network from trying
 180  * to ACK or NACK at the same time.
 181  *
 182  * Note: TIPC uses a different trigger to distribute ACKs than it does to
 183  *       distribute NACKs, but tries to use the same spacing (divide by 16).
 184  */
 185
 186 static int bclink_ack_allowed(u32 n)
 187 {
 188         return (n % TIPC_MIN_LINK_WIN) == tipc_own_tag;
 189 }
 190
 191
 192 /**
 193  * tipc_bclink_retransmit_to - get most recent node to request retransmission
 194  *
 195  * Called with bc_lock locked
 196  */
 197
 198 struct tipc_node *tipc_bclink_retransmit_to(void)
 199 {
 200         return bclink->retransmit_to;
 201 }
 202
 203 /**
 204  * bclink_retransmit_pkt - retransmit broadcast packets
 205  * @after: sequence number of last packet to *not* retransmit
 206  * @to: sequence number of last packet to retransmit
 207  *
 208  * Called with bc_lock locked
 209  */
 210
 211 static void bclink_retransmit_pkt(u32 after, u32 to)
 212 {
 213         struct sk_buff *buf;
 214
 215         buf = bcl->first_out;
 216         while (buf && less_eq(buf_seqno(buf), after))
 217                 buf = buf->next;
 218         tipc_link_retransmit(bcl, buf, mod(to - after));
 219 }
 220
 221 /**
 222  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
 223  * @n_ptr: node that sent acknowledgement info
 224  * @acked: broadcast sequence # that has been acknowledged
 225  *
 226  * Node is locked, bc_lock unlocked.
 227  */
 228
 229 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
 230 {
 231         struct sk_buff *crs;
 232         struct sk_buff *next;
 233         unsigned int released = 0;
 234
 235         spin_lock_bh(&bc_lock);
 236
 237         /* Bail out if tx queue is empty (no clean up is required) */
 238         crs = bcl->first_out;
 239         if (!crs)
 240                 goto exit;
 241
 242         /* Determine which messages need to be acknowledged */
 243         if (acked == INVALID_LINK_SEQ) {
 244                 /*
 245                  * Contact with specified node has been lost, so need to
 246                  * acknowledge sent messages only (if other nodes still exist)
 247                  * or both sent and unsent messages (otherwise)
 248                  */
 249                 if (bclink->bcast_nodes.count)
 250                         acked = bcl->fsm_msg_cnt;
 251                 else
 252                         acked = bcl->next_out_no;
 253         } else {
 254                 /*
 255                  * Bail out if specified sequence number does not correspond
 256                  * to a message that has been sent and not yet acknowledged
 257                  */
 258                 if (less(acked, buf_seqno(crs)) ||
 259                     less(bcl->fsm_msg_cnt, acked) ||
 260                     less_eq(acked, n_ptr->bclink.acked))
 261                         goto exit;
 262         }
 263
 264         /* Skip over packets that node has previously acknowledged */
 265         while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked))
 266                 crs = crs->next;
 267
 268         /* Update packets that node is now acknowledging */
 269
 270         while (crs && less_eq(buf_seqno(crs), acked)) {
 271                 next = crs->next;
 272
 273                 if (crs != bcl->next_out)
 274                         bcbuf_decr_acks(crs);
 275                 else {
 276                         bcbuf_set_acks(crs, 0);
 277                         bcl->next_out = next;
 278                         bclink_set_last_sent();
 279                 }
 280
 281                 if (bcbuf_acks(crs) == 0) {
 282                         bcl->first_out = next;
 283                         bcl->out_queue_size--;
 284                         buf_discard(crs);
 285                         released = 1;
 286                 }
 287                 crs = next;
 288         }
 289         n_ptr->bclink.acked = acked;
 290
 291         /* Try resolving broadcast link congestion, if necessary */
 292
 293         if (unlikely(bcl->next_out)) {
 294                 tipc_link_push_queue(bcl);
 295                 bclink_set_last_sent();
 296         }
 297         if (unlikely(released && !list_empty(&bcl->waiting_ports)))
 298                 tipc_link_wakeup_ports(bcl, 0);
 299 exit:
 300         spin_unlock_bh(&bc_lock);
 301 }
 302
 303 /**
 304  * bclink_send_ack - unicast an ACK msg
 305  *
 306  * tipc_net_lock and node lock set
 307  */
 308
 309 static void bclink_send_ack(struct tipc_node *n_ptr)
 310 {
 311         struct tipc_link *l_ptr = n_ptr->active_links[n_ptr->addr & 1];
 312
 313         if (l_ptr != NULL)
 314                 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
 315 }
 316
 317 /**
 318  * bclink_send_nack- broadcast a NACK msg
 319  *
 320  * tipc_net_lock and node lock set
 321  */
 322
 323 static void bclink_send_nack(struct tipc_node *n_ptr)
 324 {
 325         struct sk_buff *buf;
 326         struct tipc_msg *msg;
 327
 328         if (!less(n_ptr->bclink.gap_after, n_ptr->bclink.gap_to))
 329                 return;
 330
 331         buf = tipc_buf_acquire(INT_H_SIZE);
 332         if (buf) {
 333                 msg = buf_msg(buf);
 334                 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
 335                          INT_H_SIZE, n_ptr->addr);
 336                 msg_set_non_seq(msg, 1);
 337                 msg_set_mc_netid(msg, tipc_net_id);
 338                 msg_set_bcast_ack(msg, mod(n_ptr->bclink.last_in));
 339                 msg_set_bcgap_after(msg, n_ptr->bclink.gap_after);
 340                 msg_set_bcgap_to(msg, n_ptr->bclink.gap_to);
 341                 msg_set_bcast_tag(msg, tipc_own_tag);
 342
 343                 tipc_bearer_send(&bcbearer->bearer, buf, NULL);
 344                 bcl->stats.sent_nacks++;
 345                 buf_discard(buf);
 346
 347                 /*
 348                  * Ensure we doesn't send another NACK msg to the node
 349                  * until 16 more deferred messages arrive from it
 350                  * (i.e. helps prevent all nodes from NACK'ing at same time)
 351                  */
 352
 353                 n_ptr->bclink.nack_sync = tipc_own_tag;
 354         }
 355 }
 356
 357 /**
 358  * tipc_bclink_check_gap - send a NACK if a sequence gap exists
 359  *
 360  * tipc_net_lock and node lock set
 361  */
 362
 363 void tipc_bclink_check_gap(struct tipc_node *n_ptr, u32 last_sent)
 364 {
 365         if (!n_ptr->bclink.supported ||
 366             less_eq(last_sent, mod(n_ptr->bclink.last_in)))
 367                 return;
 368
 369         bclink_set_gap(n_ptr);
 370         if (n_ptr->bclink.gap_after == n_ptr->bclink.gap_to)
 371                 n_ptr->bclink.gap_to = last_sent;
 372         bclink_send_nack(n_ptr);
 373 }
 374
 375 /**
 376  * tipc_bclink_peek_nack - process a NACK msg meant for another node
 377  *
 378  * Only tipc_net_lock set.
 379  */
 380
 381 static void tipc_bclink_peek_nack(u32 dest, u32 sender_tag, u32 gap_after, u32 gap_to)
 382 {
 383         struct tipc_node *n_ptr = tipc_node_find(dest);
 384         u32 my_after, my_to;
 385
 386         if (unlikely(!n_ptr || !tipc_node_is_up(n_ptr)))
 387                 return;
 388         tipc_node_lock(n_ptr);
 389         /*
 390          * Modify gap to suppress unnecessary NACKs from this node
 391          */
 392         my_after = n_ptr->bclink.gap_after;
 393         my_to = n_ptr->bclink.gap_to;
 394
 395         if (less_eq(gap_after, my_after)) {
 396                 if (less(my_after, gap_to) && less(gap_to, my_to))
 397                         n_ptr->bclink.gap_after = gap_to;
 398                 else if (less_eq(my_to, gap_to))
 399                         n_ptr->bclink.gap_to = n_ptr->bclink.gap_after;
 400         } else if (less_eq(gap_after, my_to)) {
 401                 if (less_eq(my_to, gap_to))
 402                         n_ptr->bclink.gap_to = gap_after;
 403         } else {
 404                 /*
 405                  * Expand gap if missing bufs not in deferred queue:
 406                  */
 407                 struct sk_buff *buf = n_ptr->bclink.deferred_head;
 408                 u32 prev = n_ptr->bclink.gap_to;
 409
 410                 for (; buf; buf = buf->next) {
 411                         u32 seqno = buf_seqno(buf);
 412
 413                         if (mod(seqno - prev) != 1) {
 414                                 buf = NULL;
 415                                 break;
 416                         }
 417                         if (seqno == gap_after)
 418                                 break;
 419                         prev = seqno;
 420                 }
 421                 if (buf == NULL)
 422                         n_ptr->bclink.gap_to = gap_after;
 423         }
 424         /*
 425          * Some nodes may send a complementary NACK now:
 426          */
 427         if (bclink_ack_allowed(sender_tag + 1)) {
 428                 if (n_ptr->bclink.gap_to != n_ptr->bclink.gap_after) {
 429                         bclink_send_nack(n_ptr);
 430                         bclink_set_gap(n_ptr);
 431                 }
 432         }
 433         tipc_node_unlock(n_ptr);
 434 }
 435
 436 /**
 437  * tipc_bclink_send_msg - broadcast a packet to all nodes in cluster
 438  */
 439
 440 int tipc_bclink_send_msg(struct sk_buff *buf)
 441 {
 442         int res;
 443
 444         spin_lock_bh(&bc_lock);
 445
 446         if (!bclink->bcast_nodes.count) {
 447                 res = msg_data_sz(buf_msg(buf));
 448                 buf_discard(buf);
 449                 goto exit;
 450         }
 451
 452         res = tipc_link_send_buf(bcl, buf);
 453         if (likely(res >= 0)) {
 454                 bclink_set_last_sent();
 455                 bcl->stats.queue_sz_counts++;
 456                 bcl->stats.accu_queue_sz += bcl->out_queue_size;
 457         }
 458 exit:
 459         spin_unlock_bh(&bc_lock);
 460         return res;
 461 }
 462
 463 /**
 464  * tipc_bclink_recv_pkt - receive a broadcast packet, and deliver upwards
 465  *
 466  * tipc_net_lock is read_locked, no other locks set
 467  */
 468
 469 void tipc_bclink_recv_pkt(struct sk_buff *buf)
 470 {
 471         struct tipc_msg *msg = buf_msg(buf);
 472         struct tipc_node *node;
 473         u32 next_in;
 474         u32 seqno;
 475         struct sk_buff *deferred;
 476
 477         /* Screen out unwanted broadcast messages */
 478
 479         if (msg_mc_netid(msg) != tipc_net_id)
 480                 goto exit;
 481
 482         node = tipc_node_find(msg_prevnode(msg));
 483         if (unlikely(!node))
 484                 goto exit;
 485
 486         tipc_node_lock(node);
 487         if (unlikely(!node->bclink.supported))
 488                 goto unlock;
 489
 490         if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
 491                 if (msg_type(msg) != STATE_MSG)
 492                         goto unlock;
 493                 if (msg_destnode(msg) == tipc_own_addr) {
 494                         tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
 495                         tipc_node_unlock(node);
 496                         spin_lock_bh(&bc_lock);
 497                         bcl->stats.recv_nacks++;
 498                         bclink->retransmit_to = node;
 499                         bclink_retransmit_pkt(msg_bcgap_after(msg),
 500                                               msg_bcgap_to(msg));
 501                         spin_unlock_bh(&bc_lock);
 502                 } else {
 503                         tipc_node_unlock(node);
 504                         tipc_bclink_peek_nack(msg_destnode(msg),
 505                                               msg_bcast_tag(msg),
 506                                               msg_bcgap_after(msg),
 507                                               msg_bcgap_to(msg));
 508                 }
 509                 goto exit;
 510         }
 511
 512         /* Handle in-sequence broadcast message */
 513
 514 receive:
 515         next_in = mod(node->bclink.last_in + 1);
 516         seqno = msg_seqno(msg);
 517
 518         if (likely(seqno == next_in)) {
 519                 bcl->stats.recv_info++;
 520                 node->bclink.last_in++;
 521                 bclink_set_gap(node);
 522                 if (unlikely(bclink_ack_allowed(seqno))) {
 523                         bclink_send_ack(node);
 524                         bcl->stats.sent_acks++;
 525                 }
 526                 if (likely(msg_isdata(msg))) {
 527                         tipc_node_unlock(node);
 528                         if (likely(msg_mcast(msg)))
 529                                 tipc_port_recv_mcast(buf, NULL);
 530                         else
 531                                 buf_discard(buf);
 532                 } else if (msg_user(msg) == MSG_BUNDLER) {
 533                         bcl->stats.recv_bundles++;
 534                         bcl->stats.recv_bundled += msg_msgcnt(msg);
 535                         tipc_node_unlock(node);
 536                         tipc_link_recv_bundle(buf);
 537                 } else if (msg_user(msg) == MSG_FRAGMENTER) {
 538                         bcl->stats.recv_fragments++;
 539                         if (tipc_link_recv_fragment(&node->bclink.defragm,
 540                                                     &buf, &msg))
 541                                 bcl->stats.recv_fragmented++;
 542                         tipc_node_unlock(node);
 543                         tipc_net_route_msg(buf);
 544                 } else if (msg_user(msg) == NAME_DISTRIBUTOR) {
 545                         tipc_node_unlock(node);
 546                         tipc_named_recv(buf);
 547                 } else {
 548                         tipc_node_unlock(node);
 549                         buf_discard(buf);
 550                 }
 551                 buf = NULL;
 552                 tipc_node_lock(node);
 553                 deferred = node->bclink.deferred_head;
 554                 if (deferred && (buf_seqno(deferred) == mod(next_in + 1))) {
 555                         buf = deferred;
 556                         msg = buf_msg(buf);
 557                         node->bclink.deferred_head = deferred->next;
 558                         goto receive;
 559                 }
 560         } else if (less(next_in, seqno)) {
 561                 u32 gap_after = node->bclink.gap_after;
 562                 u32 gap_to = node->bclink.gap_to;
 563
 564                 if (tipc_link_defer_pkt(&node->bclink.deferred_head,
 565                                         &node->bclink.deferred_tail,
 566                                         buf)) {
 567                         node->bclink.nack_sync++;
 568                         bcl->stats.deferred_recv++;
 569                         if (seqno == mod(gap_after + 1))
 570                                 node->bclink.gap_after = seqno;
 571                         else if (less(gap_after, seqno) && less(seqno, gap_to))
 572                                 node->bclink.gap_to = seqno;
 573                 }
 574                 buf = NULL;
 575                 if (bclink_ack_allowed(node->bclink.nack_sync)) {
 576                         if (gap_to != gap_after)
 577                                 bclink_send_nack(node);
 578                         bclink_set_gap(node);
 579                 }
 580         } else {
 581                 bcl->stats.duplicates++;
 582         }
 583 unlock:
 584         tipc_node_unlock(node);
 585 exit:
 586         buf_discard(buf);
 587 }
 588
 589 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
 590 {
 591         return (n_ptr->bclink.supported &&
 592                 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
 593 }
 594
 595
 596 /**
 597  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
 598  *
 599  * Send packet over as many bearers as necessary to reach all nodes
 600  * that have joined the broadcast link.
 601  *
 602  * Returns 0 (packet sent successfully) under all circumstances,
 603  * since the broadcast link's pseudo-bearer never blocks
 604  */
 605
 606 static int tipc_bcbearer_send(struct sk_buff *buf,
 607                               struct tipc_bearer *unused1,
 608                               struct tipc_media_addr *unused2)
 609 {
 610         int bp_index;
 611
 612         /*
 613          * Prepare broadcast link message for reliable transmission,
 614          * if first time trying to send it;
 615          * preparation is skipped for broadcast link protocol messages
 616          * since they are sent in an unreliable manner and don't need it
 617          */
 618
 619         if (likely(!msg_non_seq(buf_msg(buf)))) {
 620                 struct tipc_msg *msg;
 621
 622                 bcbuf_set_acks(buf, bclink->bcast_nodes.count);
 623                 msg = buf_msg(buf);
 624                 msg_set_non_seq(msg, 1);
 625                 msg_set_mc_netid(msg, tipc_net_id);
 626                 bcl->stats.sent_info++;
 627
 628                 if (WARN_ON(!bclink->bcast_nodes.count)) {
 629                         dump_stack();
 630                         return 0;
 631                 }
 632         }
 633
 634         /* Send buffer over bearers until all targets reached */
 635
 636         bcbearer->remains = bclink->bcast_nodes;
 637
 638         for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
 639                 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
 640                 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
 641
 642                 if (!p)
 643                         break;  /* no more bearers to try */
 644
 645                 tipc_nmap_diff(&bcbearer->remains, &p->nodes, &bcbearer->remains_new);
 646                 if (bcbearer->remains_new.count == bcbearer->remains.count)
 647                         continue;       /* bearer pair doesn't add anything */
 648
 649                 if (p->blocked ||
 650                     p->media->send_msg(buf, p, &p->media->bcast_addr)) {
 651                         /* unable to send on primary bearer */
 652                         if (!s || s->blocked ||
 653                             s->media->send_msg(buf, s,
 654                                                &s->media->bcast_addr)) {
 655                                 /* unable to send on either bearer */
 656                                 continue;
 657                         }
 658                 }
 659
 660                 if (s) {
 661                         bcbearer->bpairs[bp_index].primary = s;
 662                         bcbearer->bpairs[bp_index].secondary = p;
 663                 }
 664
 665                 if (bcbearer->remains_new.count == 0)
 666                         break;  /* all targets reached */
 667
 668                 bcbearer->remains = bcbearer->remains_new;
 669         }
 670
 671         return 0;
 672 }
 673
 674 /**
 675  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
 676  */
 677
 678 void tipc_bcbearer_sort(void)
 679 {
 680         struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
 681         struct tipc_bcbearer_pair *bp_curr;
 682         int b_index;
 683         int pri;
 684
 685         spin_lock_bh(&bc_lock);
 686
 687         /* Group bearers by priority (can assume max of two per priority) */
 688
 689         memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
 690
 691         for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
 692                 struct tipc_bearer *b = &tipc_bearers[b_index];
 693
 694                 if (!b->active || !b->nodes.count)
 695                         continue;
 696
 697                 if (!bp_temp[b->priority].primary)
 698                         bp_temp[b->priority].primary = b;
 699                 else
 700                         bp_temp[b->priority].secondary = b;
 701         }
 702
 703         /* Create array of bearer pairs for broadcasting */
 704
 705         bp_curr = bcbearer->bpairs;
 706         memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
 707
 708         for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
 709
 710                 if (!bp_temp[pri].primary)
 711                         continue;
 712
 713                 bp_curr->primary = bp_temp[pri].primary;
 714
 715                 if (bp_temp[pri].secondary) {
 716                         if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
 717                                             &bp_temp[pri].secondary->nodes)) {
 718                                 bp_curr->secondary = bp_temp[pri].secondary;
 719                         } else {
 720                                 bp_curr++;
 721                                 bp_curr->primary = bp_temp[pri].secondary;
 722                         }
 723                 }
 724
 725                 bp_curr++;
 726         }
 727
 728         spin_unlock_bh(&bc_lock);
 729 }
 730
 731
 732 int tipc_bclink_stats(char *buf, const u32 buf_size)
 733 {
 734         struct print_buf pb;
 735
 736         if (!bcl)
 737                 return 0;
 738
 739         tipc_printbuf_init(&pb, buf, buf_size);
 740
 741         spin_lock_bh(&bc_lock);
 742
 743         tipc_printf(&pb, "Link <%s>\n"
 744                          "  Window:%u packets\n",
 745                     bcl->name, bcl->queue_limit[0]);
 746         tipc_printf(&pb, "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
 747                     bcl->stats.recv_info,
 748                     bcl->stats.recv_fragments,
 749                     bcl->stats.recv_fragmented,
 750                     bcl->stats.recv_bundles,
 751                     bcl->stats.recv_bundled);
 752         tipc_printf(&pb, "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
 753                     bcl->stats.sent_info,
 754                     bcl->stats.sent_fragments,
 755                     bcl->stats.sent_fragmented,
 756                     bcl->stats.sent_bundles,
 757                     bcl->stats.sent_bundled);
 758         tipc_printf(&pb, "  RX naks:%u defs:%u dups:%u\n",
 759                     bcl->stats.recv_nacks,
 760                     bcl->stats.deferred_recv,
 761                     bcl->stats.duplicates);
 762         tipc_printf(&pb, "  TX naks:%u acks:%u dups:%u\n",
 763                     bcl->stats.sent_nacks,
 764                     bcl->stats.sent_acks,
 765                     bcl->stats.retransmitted);
 766         tipc_printf(&pb, "  Congestion bearer:%u link:%u  Send queue max:%u avg:%u\n",
 767                     bcl->stats.bearer_congs,
 768                     bcl->stats.link_congs,
 769                     bcl->stats.max_queue_sz,
 770                     bcl->stats.queue_sz_counts
 771                     ? (bcl->stats.accu_queue_sz / bcl->stats.queue_sz_counts)
 772                     : 0);
 773
 774         spin_unlock_bh(&bc_lock);
 775         return tipc_printbuf_validate(&pb);
 776 }
 777
 778 int tipc_bclink_reset_stats(void)
 779 {
 780         if (!bcl)
 781                 return -ENOPROTOOPT;
 782
 783         spin_lock_bh(&bc_lock);
 784         memset(&bcl->stats, 0, sizeof(bcl->stats));
 785         spin_unlock_bh(&bc_lock);
 786         return 0;
 787 }
 788
 789 int tipc_bclink_set_queue_limits(u32 limit)
 790 {
 791         if (!bcl)
 792                 return -ENOPROTOOPT;
 793         if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
 794                 return -EINVAL;
 795
 796         spin_lock_bh(&bc_lock);
 797         tipc_link_set_queue_limits(bcl, limit);
 798         spin_unlock_bh(&bc_lock);
 799         return 0;
 800 }
 801
 802 void tipc_bclink_init(void)
 803 {
 804         INIT_LIST_HEAD(&bcbearer->bearer.cong_links);
 805         bcbearer->bearer.media = &bcbearer->media;
 806         bcbearer->media.send_msg = tipc_bcbearer_send;
 807         sprintf(bcbearer->media.name, "tipc-broadcast");
 808
 809         INIT_LIST_HEAD(&bcl->waiting_ports);
 810         bcl->next_out_no = 1;
 811         spin_lock_init(&bclink->node.lock);
 812         bcl->owner = &bclink->node;
 813         bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
 814         tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
 815         bcl->b_ptr = &bcbearer->bearer;
 816         bcl->state = WORKING_WORKING;
 817         strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
 818 }
 819
 820 void tipc_bclink_stop(void)
 821 {
 822         spin_lock_bh(&bc_lock);
 823         tipc_link_stop(bcl);
 824         spin_unlock_bh(&bc_lock);
 825
 826         memset(bclink, 0, sizeof(*bclink));
 827         memset(bcbearer, 0, sizeof(*bcbearer));
 828 }
 829
 830
 831 /**
 832  * tipc_nmap_add - add a node to a node map
 833  */
 834
 835 void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
 836 {
 837         int n = tipc_node(node);
 838         int w = n / WSIZE;
 839         u32 mask = (1 << (n % WSIZE));
 840
 841         if ((nm_ptr->map[w] & mask) == 0) {
 842                 nm_ptr->count++;
 843                 nm_ptr->map[w] |= mask;
 844         }
 845 }
 846
 847 /**
 848  * tipc_nmap_remove - remove a node from a node map
 849  */
 850
 851 void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
 852 {
 853         int n = tipc_node(node);
 854         int w = n / WSIZE;
 855         u32 mask = (1 << (n % WSIZE));
 856
 857         if ((nm_ptr->map[w] & mask) != 0) {
 858                 nm_ptr->map[w] &= ~mask;
 859                 nm_ptr->count--;
 860         }
 861 }
 862
 863 /**
 864  * tipc_nmap_diff - find differences between node maps
 865  * @nm_a: input node map A
 866  * @nm_b: input node map B
 867  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
 868  */
 869
 870 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
 871                            struct tipc_node_map *nm_b,
 872                            struct tipc_node_map *nm_diff)
 873 {
 874         int stop = ARRAY_SIZE(nm_a->map);
 875         int w;
 876         int b;
 877         u32 map;
 878
 879         memset(nm_diff, 0, sizeof(*nm_diff));
 880         for (w = 0; w < stop; w++) {
 881                 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
 882                 nm_diff->map[w] = map;
 883                 if (map != 0) {
 884                         for (b = 0 ; b < WSIZE; b++) {
 885                                 if (map & (1 << b))
 886                                         nm_diff->count++;
 887                         }
 888                 }
 889         }
 890 }
 891
 892 /**
 893  * tipc_port_list_add - add a port to a port list, ensuring no duplicates
 894  */
 895
 896 void tipc_port_list_add(struct tipc_port_list *pl_ptr, u32 port)
 897 {
 898         struct tipc_port_list *item = pl_ptr;
 899         int i;
 900         int item_sz = PLSIZE;
 901         int cnt = pl_ptr->count;
 902
 903         for (; ; cnt -= item_sz, item = item->next) {
 904                 if (cnt < PLSIZE)
 905                         item_sz = cnt;
 906                 for (i = 0; i < item_sz; i++)
 907                         if (item->ports[i] == port)
 908                                 return;
 909                 if (i < PLSIZE) {
 910                         item->ports[i] = port;
 911                         pl_ptr->count++;
 912                         return;
 913                 }
 914                 if (!item->next) {
 915                         item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
 916                         if (!item->next) {
 917                                 warn("Incomplete multicast delivery, no memory\n");
 918                                 return;
 919                         }
 920                         item->next->next = NULL;
 921                 }
 922         }
 923 }
 924
 925 /**
 926  * tipc_port_list_free - free dynamically created entries in port_list chain
 927  *
 928  */
 929
 930 void tipc_port_list_free(struct tipc_port_list *pl_ptr)
 931 {
 932         struct tipc_port_list *item;
 933         struct tipc_port_list *next;
 934
 935         for (item = pl_ptr->next; item; item = next) {
 936                 next = item->next;
 937                 kfree(item);
 938         }
 939 }
 940