net/tipc/bcast.c

   1 /*
   2  * net/tipc/bcast.c: TIPC broadcast code
   3  *
   4  * Copyright (c) 2004-2006, 2014, 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 "socket.h"
  42 #include "msg.h"
  43 #include "bcast.h"
  44 #include "name_distr.h"
  45
  46 #define MAX_PKT_DEFAULT_MCAST   1500    /* bcast link max packet size (fixed) */
  47 #define BCLINK_WIN_DEFAULT      20      /* bcast link window size (default) */
  48 #define BCBEARER                MAX_BEARERS
  49
  50 /**
  51  * struct tipc_bcbearer_pair - a pair of bearers used by broadcast link
  52  * @primary: pointer to primary bearer
  53  * @secondary: pointer to secondary bearer
  54  *
  55  * Bearers must have same priority and same set of reachable destinations
  56  * to be paired.
  57  */
  58
  59 struct tipc_bcbearer_pair {
  60         struct tipc_bearer *primary;
  61         struct tipc_bearer *secondary;
  62 };
  63
  64 /**
  65  * struct tipc_bcbearer - bearer used by broadcast link
  66  * @bearer: (non-standard) broadcast bearer structure
  67  * @media: (non-standard) broadcast media structure
  68  * @bpairs: array of bearer pairs
  69  * @bpairs_temp: temporary array of bearer pairs used by tipc_bcbearer_sort()
  70  * @remains: temporary node map used by tipc_bcbearer_send()
  71  * @remains_new: temporary node map used tipc_bcbearer_send()
  72  *
  73  * Note: The fields labelled "temporary" are incorporated into the bearer
  74  * to avoid consuming potentially limited stack space through the use of
  75  * large local variables within multicast routines.  Concurrent access is
  76  * prevented through use of the spinlock "bclink_lock".
  77  */
  78 struct tipc_bcbearer {
  79         struct tipc_bearer bearer;
  80         struct tipc_media media;
  81         struct tipc_bcbearer_pair bpairs[MAX_BEARERS];
  82         struct tipc_bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1];
  83         struct tipc_node_map remains;
  84         struct tipc_node_map remains_new;
  85 };
  86
  87 /**
  88  * struct tipc_bclink - link used for broadcast messages
  89  * @lock: spinlock governing access to structure
  90  * @link: (non-standard) broadcast link structure
  91  * @node: (non-standard) node structure representing b'cast link's peer node
  92  * @flags: represent bclink states
  93  * @bcast_nodes: map of broadcast-capable nodes
  94  * @retransmit_to: node that most recently requested a retransmit
  95  *
  96  * Handles sequence numbering, fragmentation, bundling, etc.
  97  */
  98 struct tipc_bclink {
  99         spinlock_t lock;
 100         struct tipc_link link;
 101         struct tipc_node node;
 102         unsigned int flags;
 103         struct tipc_node_map bcast_nodes;
 104         struct tipc_node *retransmit_to;
 105 };
 106
 107 static struct tipc_bcbearer *bcbearer;
 108 static struct tipc_bclink *bclink;
 109 static struct tipc_link *bcl;
 110
 111 const char tipc_bclink_name[] = "broadcast-link";
 112
 113 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
 114                            struct tipc_node_map *nm_b,
 115                            struct tipc_node_map *nm_diff);
 116 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node);
 117 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node);
 118
 119 static void tipc_bclink_lock(void)
 120 {
 121         spin_lock_bh(&bclink->lock);
 122 }
 123
 124 static void tipc_bclink_unlock(void)
 125 {
 126         struct tipc_node *node = NULL;
 127
 128         if (likely(!bclink->flags)) {
 129                 spin_unlock_bh(&bclink->lock);
 130                 return;
 131         }
 132
 133         if (bclink->flags & TIPC_BCLINK_RESET) {
 134                 bclink->flags &= ~TIPC_BCLINK_RESET;
 135                 node = tipc_bclink_retransmit_to();
 136         }
 137         spin_unlock_bh(&bclink->lock);
 138
 139         if (node)
 140                 tipc_link_reset_all(node);
 141 }
 142
 143 uint  tipc_bclink_get_mtu(void)
 144 {
 145         return MAX_PKT_DEFAULT_MCAST;
 146 }
 147
 148 void tipc_bclink_set_flags(unsigned int flags)
 149 {
 150         bclink->flags |= flags;
 151 }
 152
 153 static u32 bcbuf_acks(struct sk_buff *buf)
 154 {
 155         return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
 156 }
 157
 158 static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
 159 {
 160         TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
 161 }
 162
 163 static void bcbuf_decr_acks(struct sk_buff *buf)
 164 {
 165         bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
 166 }
 167
 168 void tipc_bclink_add_node(u32 addr)
 169 {
 170         tipc_bclink_lock();
 171         tipc_nmap_add(&bclink->bcast_nodes, addr);
 172         tipc_bclink_unlock();
 173 }
 174
 175 void tipc_bclink_remove_node(u32 addr)
 176 {
 177         tipc_bclink_lock();
 178         tipc_nmap_remove(&bclink->bcast_nodes, addr);
 179         tipc_bclink_unlock();
 180 }
 181
 182 static void bclink_set_last_sent(void)
 183 {
 184         if (bcl->next_out)
 185                 bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
 186         else
 187                 bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
 188 }
 189
 190 u32 tipc_bclink_get_last_sent(void)
 191 {
 192         return bcl->fsm_msg_cnt;
 193 }
 194
 195 static void bclink_update_last_sent(struct tipc_node *node, u32 seqno)
 196 {
 197         node->bclink.last_sent = less_eq(node->bclink.last_sent, seqno) ?
 198                                                 seqno : node->bclink.last_sent;
 199 }
 200
 201
 202 /**
 203  * tipc_bclink_retransmit_to - get most recent node to request retransmission
 204  *
 205  * Called with bclink_lock locked
 206  */
 207 struct tipc_node *tipc_bclink_retransmit_to(void)
 208 {
 209         return bclink->retransmit_to;
 210 }
 211
 212 /**
 213  * bclink_retransmit_pkt - retransmit broadcast packets
 214  * @after: sequence number of last packet to *not* retransmit
 215  * @to: sequence number of last packet to retransmit
 216  *
 217  * Called with bclink_lock locked
 218  */
 219 static void bclink_retransmit_pkt(u32 after, u32 to)
 220 {
 221         struct sk_buff *buf;
 222
 223         buf = bcl->first_out;
 224         while (buf && less_eq(buf_seqno(buf), after))
 225                 buf = buf->next;
 226         tipc_link_retransmit(bcl, buf, mod(to - after));
 227 }
 228
 229 /**
 230  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
 231  * @n_ptr: node that sent acknowledgement info
 232  * @acked: broadcast sequence # that has been acknowledged
 233  *
 234  * Node is locked, bclink_lock unlocked.
 235  */
 236 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
 237 {
 238         struct sk_buff *crs;
 239         struct sk_buff *next;
 240         unsigned int released = 0;
 241
 242         tipc_bclink_lock();
 243         /* Bail out if tx queue is empty (no clean up is required) */
 244         crs = bcl->first_out;
 245         if (!crs)
 246                 goto exit;
 247
 248         /* Determine which messages need to be acknowledged */
 249         if (acked == INVALID_LINK_SEQ) {
 250                 /*
 251                  * Contact with specified node has been lost, so need to
 252                  * acknowledge sent messages only (if other nodes still exist)
 253                  * or both sent and unsent messages (otherwise)
 254                  */
 255                 if (bclink->bcast_nodes.count)
 256                         acked = bcl->fsm_msg_cnt;
 257                 else
 258                         acked = bcl->next_out_no;
 259         } else {
 260                 /*
 261                  * Bail out if specified sequence number does not correspond
 262                  * to a message that has been sent and not yet acknowledged
 263                  */
 264                 if (less(acked, buf_seqno(crs)) ||
 265                     less(bcl->fsm_msg_cnt, acked) ||
 266                     less_eq(acked, n_ptr->bclink.acked))
 267                         goto exit;
 268         }
 269
 270         /* Skip over packets that node has previously acknowledged */
 271         while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked))
 272                 crs = crs->next;
 273
 274         /* Update packets that node is now acknowledging */
 275
 276         while (crs && less_eq(buf_seqno(crs), acked)) {
 277                 next = crs->next;
 278
 279                 if (crs != bcl->next_out)
 280                         bcbuf_decr_acks(crs);
 281                 else {
 282                         bcbuf_set_acks(crs, 0);
 283                         bcl->next_out = next;
 284                         bclink_set_last_sent();
 285                 }
 286
 287                 if (bcbuf_acks(crs) == 0) {
 288                         bcl->first_out = next;
 289                         bcl->out_queue_size--;
 290                         kfree_skb(crs);
 291                         released = 1;
 292                 }
 293                 crs = next;
 294         }
 295         n_ptr->bclink.acked = acked;
 296
 297         /* Try resolving broadcast link congestion, if necessary */
 298
 299         if (unlikely(bcl->next_out)) {
 300                 tipc_link_push_queue(bcl);
 301                 bclink_set_last_sent();
 302         }
 303         if (unlikely(released && !list_empty(&bcl->waiting_ports)))
 304                 tipc_link_wakeup_ports(bcl, 0);
 305 exit:
 306         tipc_bclink_unlock();
 307 }
 308
 309 /**
 310  * tipc_bclink_update_link_state - update broadcast link state
 311  *
 312  * RCU and node lock set
 313  */
 314 void tipc_bclink_update_link_state(struct tipc_node *n_ptr, u32 last_sent)
 315 {
 316         struct sk_buff *buf;
 317
 318         /* Ignore "stale" link state info */
 319
 320         if (less_eq(last_sent, n_ptr->bclink.last_in))
 321                 return;
 322
 323         /* Update link synchronization state; quit if in sync */
 324
 325         bclink_update_last_sent(n_ptr, last_sent);
 326
 327         if (n_ptr->bclink.last_sent == n_ptr->bclink.last_in)
 328                 return;
 329
 330         /* Update out-of-sync state; quit if loss is still unconfirmed */
 331
 332         if ((++n_ptr->bclink.oos_state) == 1) {
 333                 if (n_ptr->bclink.deferred_size < (TIPC_MIN_LINK_WIN / 2))
 334                         return;
 335                 n_ptr->bclink.oos_state++;
 336         }
 337
 338         /* Don't NACK if one has been recently sent (or seen) */
 339
 340         if (n_ptr->bclink.oos_state & 0x1)
 341                 return;
 342
 343         /* Send NACK */
 344
 345         buf = tipc_buf_acquire(INT_H_SIZE);
 346         if (buf) {
 347                 struct tipc_msg *msg = buf_msg(buf);
 348
 349                 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
 350                               INT_H_SIZE, n_ptr->addr);
 351                 msg_set_non_seq(msg, 1);
 352                 msg_set_mc_netid(msg, tipc_net_id);
 353                 msg_set_bcast_ack(msg, n_ptr->bclink.last_in);
 354                 msg_set_bcgap_after(msg, n_ptr->bclink.last_in);
 355                 msg_set_bcgap_to(msg, n_ptr->bclink.deferred_head
 356                                  ? buf_seqno(n_ptr->bclink.deferred_head) - 1
 357                                  : n_ptr->bclink.last_sent);
 358
 359                 tipc_bclink_lock();
 360                 tipc_bearer_send(MAX_BEARERS, buf, NULL);
 361                 bcl->stats.sent_nacks++;
 362                 tipc_bclink_unlock();
 363                 kfree_skb(buf);
 364
 365                 n_ptr->bclink.oos_state++;
 366         }
 367 }
 368
 369 /**
 370  * bclink_peek_nack - monitor retransmission requests sent by other nodes
 371  *
 372  * Delay any upcoming NACK by this node if another node has already
 373  * requested the first message this node is going to ask for.
 374  */
 375 static void bclink_peek_nack(struct tipc_msg *msg)
 376 {
 377         struct tipc_node *n_ptr = tipc_node_find(msg_destnode(msg));
 378
 379         if (unlikely(!n_ptr))
 380                 return;
 381
 382         tipc_node_lock(n_ptr);
 383
 384         if (n_ptr->bclink.recv_permitted &&
 385             (n_ptr->bclink.last_in != n_ptr->bclink.last_sent) &&
 386             (n_ptr->bclink.last_in == msg_bcgap_after(msg)))
 387                 n_ptr->bclink.oos_state = 2;
 388
 389         tipc_node_unlock(n_ptr);
 390 }
 391
 392 /*
 393  * tipc_bclink_xmit - broadcast a packet to all nodes in cluster
 394  */
 395 int tipc_bclink_xmit(struct sk_buff *buf)
 396 {
 397         int res;
 398
 399         tipc_bclink_lock();
 400
 401         if (!bclink->bcast_nodes.count) {
 402                 res = msg_data_sz(buf_msg(buf));
 403                 kfree_skb(buf);
 404                 goto exit;
 405         }
 406
 407         res = __tipc_link_xmit(bcl, buf);
 408         if (likely(res >= 0)) {
 409                 bclink_set_last_sent();
 410                 bcl->stats.queue_sz_counts++;
 411                 bcl->stats.accu_queue_sz += bcl->out_queue_size;
 412         }
 413 exit:
 414         tipc_bclink_unlock();
 415         return res;
 416 }
 417
 418 /* tipc_bclink_xmit2 - broadcast buffer chain to all nodes in cluster
 419  *                     and to identified node local sockets
 420  * @buf: chain of buffers containing message
 421  * Consumes the buffer chain, except when returning -ELINKCONG
 422  * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
 423  */
 424 int tipc_bclink_xmit2(struct sk_buff *buf)
 425 {
 426         int rc = 0;
 427         int bc = 0;
 428         struct sk_buff *clbuf;
 429
 430         /* Prepare clone of message for local node */
 431         clbuf = tipc_msg_reassemble(buf);
 432         if (unlikely(!clbuf)) {
 433                 kfree_skb_list(buf);
 434                 return -EHOSTUNREACH;
 435         }
 436
 437         /* Broadcast to all other nodes */
 438         if (likely(bclink)) {
 439                 tipc_bclink_lock();
 440                 if (likely(bclink->bcast_nodes.count)) {
 441                         rc = __tipc_link_xmit(bcl, buf);
 442                         if (likely(!rc)) {
 443                                 bclink_set_last_sent();
 444                                 bcl->stats.queue_sz_counts++;
 445                                 bcl->stats.accu_queue_sz += bcl->out_queue_size;
 446                         }
 447                         bc = 1;
 448                 }
 449                 tipc_bclink_unlock();
 450         }
 451
 452         if (unlikely(!bc))
 453                 kfree_skb_list(buf);
 454
 455         /* Deliver message clone */
 456         if (likely(!rc))
 457                 tipc_sk_mcast_rcv(clbuf);
 458         else
 459                 kfree_skb(clbuf);
 460
 461         return rc;
 462 }
 463
 464 /**
 465  * bclink_accept_pkt - accept an incoming, in-sequence broadcast packet
 466  *
 467  * Called with both sending node's lock and bclink_lock taken.
 468  */
 469 static void bclink_accept_pkt(struct tipc_node *node, u32 seqno)
 470 {
 471         bclink_update_last_sent(node, seqno);
 472         node->bclink.last_in = seqno;
 473         node->bclink.oos_state = 0;
 474         bcl->stats.recv_info++;
 475
 476         /*
 477          * Unicast an ACK periodically, ensuring that
 478          * all nodes in the cluster don't ACK at the same time
 479          */
 480
 481         if (((seqno - tipc_own_addr) % TIPC_MIN_LINK_WIN) == 0) {
 482                 tipc_link_proto_xmit(node->active_links[node->addr & 1],
 483                                      STATE_MSG, 0, 0, 0, 0, 0);
 484                 bcl->stats.sent_acks++;
 485         }
 486 }
 487
 488 /**
 489  * tipc_bclink_rcv - receive a broadcast packet, and deliver upwards
 490  *
 491  * RCU is locked, no other locks set
 492  */
 493 void tipc_bclink_rcv(struct sk_buff *buf)
 494 {
 495         struct tipc_msg *msg = buf_msg(buf);
 496         struct tipc_node *node;
 497         u32 next_in;
 498         u32 seqno;
 499         int deferred;
 500
 501         /* Screen out unwanted broadcast messages */
 502
 503         if (msg_mc_netid(msg) != tipc_net_id)
 504                 goto exit;
 505
 506         node = tipc_node_find(msg_prevnode(msg));
 507         if (unlikely(!node))
 508                 goto exit;
 509
 510         tipc_node_lock(node);
 511         if (unlikely(!node->bclink.recv_permitted))
 512                 goto unlock;
 513
 514         /* Handle broadcast protocol message */
 515
 516         if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
 517                 if (msg_type(msg) != STATE_MSG)
 518                         goto unlock;
 519                 if (msg_destnode(msg) == tipc_own_addr) {
 520                         tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
 521                         tipc_node_unlock(node);
 522                         tipc_bclink_lock();
 523                         bcl->stats.recv_nacks++;
 524                         bclink->retransmit_to = node;
 525                         bclink_retransmit_pkt(msg_bcgap_after(msg),
 526                                               msg_bcgap_to(msg));
 527                         tipc_bclink_unlock();
 528                 } else {
 529                         tipc_node_unlock(node);
 530                         bclink_peek_nack(msg);
 531                 }
 532                 goto exit;
 533         }
 534
 535         /* Handle in-sequence broadcast message */
 536
 537         seqno = msg_seqno(msg);
 538         next_in = mod(node->bclink.last_in + 1);
 539
 540         if (likely(seqno == next_in)) {
 541 receive:
 542                 /* Deliver message to destination */
 543
 544                 if (likely(msg_isdata(msg))) {
 545                         tipc_bclink_lock();
 546                         bclink_accept_pkt(node, seqno);
 547                         tipc_bclink_unlock();
 548                         tipc_node_unlock(node);
 549                         if (likely(msg_mcast(msg)))
 550                                 tipc_port_mcast_rcv(buf, NULL);
 551                         else
 552                                 kfree_skb(buf);
 553                 } else if (msg_user(msg) == MSG_BUNDLER) {
 554                         tipc_bclink_lock();
 555                         bclink_accept_pkt(node, seqno);
 556                         bcl->stats.recv_bundles++;
 557                         bcl->stats.recv_bundled += msg_msgcnt(msg);
 558                         tipc_bclink_unlock();
 559                         tipc_node_unlock(node);
 560                         tipc_link_bundle_rcv(buf);
 561                 } else if (msg_user(msg) == MSG_FRAGMENTER) {
 562                         tipc_buf_append(&node->bclink.reasm_buf, &buf);
 563                         if (unlikely(!buf && !node->bclink.reasm_buf))
 564                                 goto unlock;
 565                         tipc_bclink_lock();
 566                         bclink_accept_pkt(node, seqno);
 567                         bcl->stats.recv_fragments++;
 568                         if (buf) {
 569                                 bcl->stats.recv_fragmented++;
 570                                 msg = buf_msg(buf);
 571                                 tipc_bclink_unlock();
 572                                 goto receive;
 573                         }
 574                         tipc_bclink_unlock();
 575                         tipc_node_unlock(node);
 576                 } else if (msg_user(msg) == NAME_DISTRIBUTOR) {
 577                         tipc_bclink_lock();
 578                         bclink_accept_pkt(node, seqno);
 579                         tipc_bclink_unlock();
 580                         tipc_node_unlock(node);
 581                         tipc_named_rcv(buf);
 582                 } else {
 583                         tipc_bclink_lock();
 584                         bclink_accept_pkt(node, seqno);
 585                         tipc_bclink_unlock();
 586                         tipc_node_unlock(node);
 587                         kfree_skb(buf);
 588                 }
 589                 buf = NULL;
 590
 591                 /* Determine new synchronization state */
 592
 593                 tipc_node_lock(node);
 594                 if (unlikely(!tipc_node_is_up(node)))
 595                         goto unlock;
 596
 597                 if (node->bclink.last_in == node->bclink.last_sent)
 598                         goto unlock;
 599
 600                 if (!node->bclink.deferred_head) {
 601                         node->bclink.oos_state = 1;
 602                         goto unlock;
 603                 }
 604
 605                 msg = buf_msg(node->bclink.deferred_head);
 606                 seqno = msg_seqno(msg);
 607                 next_in = mod(next_in + 1);
 608                 if (seqno != next_in)
 609                         goto unlock;
 610
 611                 /* Take in-sequence message from deferred queue & deliver it */
 612
 613                 buf = node->bclink.deferred_head;
 614                 node->bclink.deferred_head = buf->next;
 615                 buf->next = NULL;
 616                 node->bclink.deferred_size--;
 617                 goto receive;
 618         }
 619
 620         /* Handle out-of-sequence broadcast message */
 621
 622         if (less(next_in, seqno)) {
 623                 deferred = tipc_link_defer_pkt(&node->bclink.deferred_head,
 624                                                &node->bclink.deferred_tail,
 625                                                buf);
 626                 node->bclink.deferred_size += deferred;
 627                 bclink_update_last_sent(node, seqno);
 628                 buf = NULL;
 629         } else
 630                 deferred = 0;
 631
 632         tipc_bclink_lock();
 633
 634         if (deferred)
 635                 bcl->stats.deferred_recv++;
 636         else
 637                 bcl->stats.duplicates++;
 638
 639         tipc_bclink_unlock();
 640
 641 unlock:
 642         tipc_node_unlock(node);
 643 exit:
 644         kfree_skb(buf);
 645 }
 646
 647 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
 648 {
 649         return (n_ptr->bclink.recv_permitted &&
 650                 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
 651 }
 652
 653
 654 /**
 655  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
 656  *
 657  * Send packet over as many bearers as necessary to reach all nodes
 658  * that have joined the broadcast link.
 659  *
 660  * Returns 0 (packet sent successfully) under all circumstances,
 661  * since the broadcast link's pseudo-bearer never blocks
 662  */
 663 static int tipc_bcbearer_send(struct sk_buff *buf, struct tipc_bearer *unused1,
 664                               struct tipc_media_addr *unused2)
 665 {
 666         int bp_index;
 667
 668         /* Prepare broadcast link message for reliable transmission,
 669          * if first time trying to send it;
 670          * preparation is skipped for broadcast link protocol messages
 671          * since they are sent in an unreliable manner and don't need it
 672          */
 673         if (likely(!msg_non_seq(buf_msg(buf)))) {
 674                 struct tipc_msg *msg;
 675
 676                 bcbuf_set_acks(buf, bclink->bcast_nodes.count);
 677                 msg = buf_msg(buf);
 678                 msg_set_non_seq(msg, 1);
 679                 msg_set_mc_netid(msg, tipc_net_id);
 680                 bcl->stats.sent_info++;
 681
 682                 if (WARN_ON(!bclink->bcast_nodes.count)) {
 683                         dump_stack();
 684                         return 0;
 685                 }
 686         }
 687
 688         /* Send buffer over bearers until all targets reached */
 689         bcbearer->remains = bclink->bcast_nodes;
 690
 691         for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
 692                 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
 693                 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
 694                 struct tipc_bearer *b = p;
 695                 struct sk_buff *tbuf;
 696
 697                 if (!p)
 698                         break; /* No more bearers to try */
 699
 700                 tipc_nmap_diff(&bcbearer->remains, &b->nodes,
 701                                &bcbearer->remains_new);
 702                 if (bcbearer->remains_new.count == bcbearer->remains.count)
 703                         continue; /* Nothing added by bearer pair */
 704
 705                 if (bp_index == 0) {
 706                         /* Use original buffer for first bearer */
 707                         tipc_bearer_send(b->identity, buf, &b->bcast_addr);
 708                 } else {
 709                         /* Avoid concurrent buffer access */
 710                         tbuf = pskb_copy_for_clone(buf, GFP_ATOMIC);
 711                         if (!tbuf)
 712                                 break;
 713                         tipc_bearer_send(b->identity, tbuf, &b->bcast_addr);
 714                         kfree_skb(tbuf); /* Bearer keeps a clone */
 715                 }
 716
 717                 /* Swap bearers for next packet */
 718                 if (s) {
 719                         bcbearer->bpairs[bp_index].primary = s;
 720                         bcbearer->bpairs[bp_index].secondary = p;
 721                 }
 722
 723                 if (bcbearer->remains_new.count == 0)
 724                         break; /* All targets reached */
 725
 726                 bcbearer->remains = bcbearer->remains_new;
 727         }
 728
 729         return 0;
 730 }
 731
 732 /**
 733  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
 734  */
 735 void tipc_bcbearer_sort(struct tipc_node_map *nm_ptr, u32 node, bool action)
 736 {
 737         struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
 738         struct tipc_bcbearer_pair *bp_curr;
 739         struct tipc_bearer *b;
 740         int b_index;
 741         int pri;
 742
 743         tipc_bclink_lock();
 744
 745         if (action)
 746                 tipc_nmap_add(nm_ptr, node);
 747         else
 748                 tipc_nmap_remove(nm_ptr, node);
 749
 750         /* Group bearers by priority (can assume max of two per priority) */
 751         memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
 752
 753         rcu_read_lock();
 754         for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
 755                 b = rcu_dereference_rtnl(bearer_list[b_index]);
 756                 if (!b || !b->nodes.count)
 757                         continue;
 758
 759                 if (!bp_temp[b->priority].primary)
 760                         bp_temp[b->priority].primary = b;
 761                 else
 762                         bp_temp[b->priority].secondary = b;
 763         }
 764         rcu_read_unlock();
 765
 766         /* Create array of bearer pairs for broadcasting */
 767         bp_curr = bcbearer->bpairs;
 768         memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
 769
 770         for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
 771
 772                 if (!bp_temp[pri].primary)
 773                         continue;
 774
 775                 bp_curr->primary = bp_temp[pri].primary;
 776
 777                 if (bp_temp[pri].secondary) {
 778                         if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
 779                                             &bp_temp[pri].secondary->nodes)) {
 780                                 bp_curr->secondary = bp_temp[pri].secondary;
 781                         } else {
 782                                 bp_curr++;
 783                                 bp_curr->primary = bp_temp[pri].secondary;
 784                         }
 785                 }
 786
 787                 bp_curr++;
 788         }
 789
 790         tipc_bclink_unlock();
 791 }
 792
 793
 794 int tipc_bclink_stats(char *buf, const u32 buf_size)
 795 {
 796         int ret;
 797         struct tipc_stats *s;
 798
 799         if (!bcl)
 800                 return 0;
 801
 802         tipc_bclink_lock();
 803
 804         s = &bcl->stats;
 805
 806         ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
 807                             "  Window:%u packets\n",
 808                             bcl->name, bcl->queue_limit[0]);
 809         ret += tipc_snprintf(buf + ret, buf_size - ret,
 810                              "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
 811                              s->recv_info, s->recv_fragments,
 812                              s->recv_fragmented, s->recv_bundles,
 813                              s->recv_bundled);
 814         ret += tipc_snprintf(buf + ret, buf_size - ret,
 815                              "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
 816                              s->sent_info, s->sent_fragments,
 817                              s->sent_fragmented, s->sent_bundles,
 818                              s->sent_bundled);
 819         ret += tipc_snprintf(buf + ret, buf_size - ret,
 820                              "  RX naks:%u defs:%u dups:%u\n",
 821                              s->recv_nacks, s->deferred_recv, s->duplicates);
 822         ret += tipc_snprintf(buf + ret, buf_size - ret,
 823                              "  TX naks:%u acks:%u dups:%u\n",
 824                              s->sent_nacks, s->sent_acks, s->retransmitted);
 825         ret += tipc_snprintf(buf + ret, buf_size - ret,
 826                              "  Congestion link:%u  Send queue max:%u avg:%u\n",
 827                              s->link_congs, s->max_queue_sz,
 828                              s->queue_sz_counts ?
 829                              (s->accu_queue_sz / s->queue_sz_counts) : 0);
 830
 831         tipc_bclink_unlock();
 832         return ret;
 833 }
 834
 835 int tipc_bclink_reset_stats(void)
 836 {
 837         if (!bcl)
 838                 return -ENOPROTOOPT;
 839
 840         tipc_bclink_lock();
 841         memset(&bcl->stats, 0, sizeof(bcl->stats));
 842         tipc_bclink_unlock();
 843         return 0;
 844 }
 845
 846 int tipc_bclink_set_queue_limits(u32 limit)
 847 {
 848         if (!bcl)
 849                 return -ENOPROTOOPT;
 850         if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
 851                 return -EINVAL;
 852
 853         tipc_bclink_lock();
 854         tipc_link_set_queue_limits(bcl, limit);
 855         tipc_bclink_unlock();
 856         return 0;
 857 }
 858
 859 int tipc_bclink_init(void)
 860 {
 861         bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
 862         if (!bcbearer)
 863                 return -ENOMEM;
 864
 865         bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
 866         if (!bclink) {
 867                 kfree(bcbearer);
 868                 return -ENOMEM;
 869         }
 870
 871         bcl = &bclink->link;
 872         bcbearer->bearer.media = &bcbearer->media;
 873         bcbearer->media.send_msg = tipc_bcbearer_send;
 874         sprintf(bcbearer->media.name, "tipc-broadcast");
 875
 876         spin_lock_init(&bclink->lock);
 877         INIT_LIST_HEAD(&bcl->waiting_ports);
 878         bcl->next_out_no = 1;
 879         spin_lock_init(&bclink->node.lock);
 880         bcl->owner = &bclink->node;
 881         bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
 882         tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
 883         bcl->bearer_id = MAX_BEARERS;
 884         rcu_assign_pointer(bearer_list[MAX_BEARERS], &bcbearer->bearer);
 885         bcl->state = WORKING_WORKING;
 886         strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
 887         return 0;
 888 }
 889
 890 void tipc_bclink_stop(void)
 891 {
 892         tipc_bclink_lock();
 893         tipc_link_purge_queues(bcl);
 894         tipc_bclink_unlock();
 895
 896         RCU_INIT_POINTER(bearer_list[BCBEARER], NULL);
 897         synchronize_net();
 898         kfree(bcbearer);
 899         kfree(bclink);
 900 }
 901
 902 /**
 903  * tipc_nmap_add - add a node to a node map
 904  */
 905 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
 906 {
 907         int n = tipc_node(node);
 908         int w = n / WSIZE;
 909         u32 mask = (1 << (n % WSIZE));
 910
 911         if ((nm_ptr->map[w] & mask) == 0) {
 912                 nm_ptr->count++;
 913                 nm_ptr->map[w] |= mask;
 914         }
 915 }
 916
 917 /**
 918  * tipc_nmap_remove - remove a node from a node map
 919  */
 920 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
 921 {
 922         int n = tipc_node(node);
 923         int w = n / WSIZE;
 924         u32 mask = (1 << (n % WSIZE));
 925
 926         if ((nm_ptr->map[w] & mask) != 0) {
 927                 nm_ptr->map[w] &= ~mask;
 928                 nm_ptr->count--;
 929         }
 930 }
 931
 932 /**
 933  * tipc_nmap_diff - find differences between node maps
 934  * @nm_a: input node map A
 935  * @nm_b: input node map B
 936  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
 937  */
 938 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
 939                            struct tipc_node_map *nm_b,
 940                            struct tipc_node_map *nm_diff)
 941 {
 942         int stop = ARRAY_SIZE(nm_a->map);
 943         int w;
 944         int b;
 945         u32 map;
 946
 947         memset(nm_diff, 0, sizeof(*nm_diff));
 948         for (w = 0; w < stop; w++) {
 949                 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
 950                 nm_diff->map[w] = map;
 951                 if (map != 0) {
 952                         for (b = 0 ; b < WSIZE; b++) {
 953                                 if (map & (1 << b))
 954                                         nm_diff->count++;
 955                         }
 956                 }
 957         }
 958 }
 959
 960 /**
 961  * tipc_port_list_add - add a port to a port list, ensuring no duplicates
 962  */
 963 void tipc_port_list_add(struct tipc_port_list *pl_ptr, u32 port)
 964 {
 965         struct tipc_port_list *item = pl_ptr;
 966         int i;
 967         int item_sz = PLSIZE;
 968         int cnt = pl_ptr->count;
 969
 970         for (; ; cnt -= item_sz, item = item->next) {
 971                 if (cnt < PLSIZE)
 972                         item_sz = cnt;
 973                 for (i = 0; i < item_sz; i++)
 974                         if (item->ports[i] == port)
 975                                 return;
 976                 if (i < PLSIZE) {
 977                         item->ports[i] = port;
 978                         pl_ptr->count++;
 979                         return;
 980                 }
 981                 if (!item->next) {
 982                         item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
 983                         if (!item->next) {
 984                                 pr_warn("Incomplete multicast delivery, no memory\n");
 985                                 return;
 986                         }
 987                         item->next->next = NULL;
 988                 }
 989         }
 990 }
 991
 992 /**
 993  * tipc_port_list_free - free dynamically created entries in port_list chain
 994  *
 995  */
 996 void tipc_port_list_free(struct tipc_port_list *pl_ptr)
 997 {
 998         struct tipc_port_list *item;
 999         struct tipc_port_list *next;
1000
1001         for (item = pl_ptr->next; item; item = next) {
1002                 next = item->next;
1003                 kfree(item);
1004         }
1005 }