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