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