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_wakeup_users - wake up pending users
 230  *
 231  * Called with no locks taken
 232  */
 233 void tipc_bclink_wakeup_users(void)
 234 {
 235         while (skb_queue_len(&bclink->link.waiting_sks))
 236                 tipc_sk_rcv(skb_dequeue(&bclink->link.waiting_sks));
 237 }
 238
 239 /**
 240  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
 241  * @n_ptr: node that sent acknowledgement info
 242  * @acked: broadcast sequence # that has been acknowledged
 243  *
 244  * Node is locked, bclink_lock unlocked.
 245  */
 246 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
 247 {
 248         struct sk_buff *crs;
 249         struct sk_buff *next;
 250         unsigned int released = 0;
 251
 252         tipc_bclink_lock();
 253         /* Bail out if tx queue is empty (no clean up is required) */
 254         crs = bcl->first_out;
 255         if (!crs)
 256                 goto exit;
 257
 258         /* Determine which messages need to be acknowledged */
 259         if (acked == INVALID_LINK_SEQ) {
 260                 /*
 261                  * Contact with specified node has been lost, so need to
 262                  * acknowledge sent messages only (if other nodes still exist)
 263                  * or both sent and unsent messages (otherwise)
 264                  */
 265                 if (bclink->bcast_nodes.count)
 266                         acked = bcl->fsm_msg_cnt;
 267                 else
 268                         acked = bcl->next_out_no;
 269         } else {
 270                 /*
 271                  * Bail out if specified sequence number does not correspond
 272                  * to a message that has been sent and not yet acknowledged
 273                  */
 274                 if (less(acked, buf_seqno(crs)) ||
 275                     less(bcl->fsm_msg_cnt, acked) ||
 276                     less_eq(acked, n_ptr->bclink.acked))
 277                         goto exit;
 278         }
 279
 280         /* Skip over packets that node has previously acknowledged */
 281         while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked))
 282                 crs = crs->next;
 283
 284         /* Update packets that node is now acknowledging */
 285
 286         while (crs && less_eq(buf_seqno(crs), acked)) {
 287                 next = crs->next;
 288
 289                 if (crs != bcl->next_out)
 290                         bcbuf_decr_acks(crs);
 291                 else {
 292                         bcbuf_set_acks(crs, 0);
 293                         bcl->next_out = next;
 294                         bclink_set_last_sent();
 295                 }
 296
 297                 if (bcbuf_acks(crs) == 0) {
 298                         bcl->first_out = next;
 299                         bcl->out_queue_size--;
 300                         kfree_skb(crs);
 301                         released = 1;
 302                 }
 303                 crs = next;
 304         }
 305         n_ptr->bclink.acked = acked;
 306
 307         /* Try resolving broadcast link congestion, if necessary */
 308
 309         if (unlikely(bcl->next_out)) {
 310                 tipc_link_push_queue(bcl);
 311                 bclink_set_last_sent();
 312         }
 313         if (unlikely(released && !skb_queue_empty(&bcl->waiting_sks)))
 314                 n_ptr->action_flags |= TIPC_WAKEUP_BCAST_USERS;
 315
 316 exit:
 317         tipc_bclink_unlock();
 318 }
 319
 320 /**
 321  * tipc_bclink_update_link_state - update broadcast link state
 322  *
 323  * RCU and node lock set
 324  */
 325 void tipc_bclink_update_link_state(struct tipc_node *n_ptr, u32 last_sent)
 326 {
 327         struct sk_buff *buf;
 328
 329         /* Ignore "stale" link state info */
 330
 331         if (less_eq(last_sent, n_ptr->bclink.last_in))
 332                 return;
 333
 334         /* Update link synchronization state; quit if in sync */
 335
 336         bclink_update_last_sent(n_ptr, last_sent);
 337
 338         if (n_ptr->bclink.last_sent == n_ptr->bclink.last_in)
 339                 return;
 340
 341         /* Update out-of-sync state; quit if loss is still unconfirmed */
 342
 343         if ((++n_ptr->bclink.oos_state) == 1) {
 344                 if (n_ptr->bclink.deferred_size < (TIPC_MIN_LINK_WIN / 2))
 345                         return;
 346                 n_ptr->bclink.oos_state++;
 347         }
 348
 349         /* Don't NACK if one has been recently sent (or seen) */
 350
 351         if (n_ptr->bclink.oos_state & 0x1)
 352                 return;
 353
 354         /* Send NACK */
 355
 356         buf = tipc_buf_acquire(INT_H_SIZE);
 357         if (buf) {
 358                 struct tipc_msg *msg = buf_msg(buf);
 359
 360                 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
 361                               INT_H_SIZE, n_ptr->addr);
 362                 msg_set_non_seq(msg, 1);
 363                 msg_set_mc_netid(msg, tipc_net_id);
 364                 msg_set_bcast_ack(msg, n_ptr->bclink.last_in);
 365                 msg_set_bcgap_after(msg, n_ptr->bclink.last_in);
 366                 msg_set_bcgap_to(msg, n_ptr->bclink.deferred_head
 367                                  ? buf_seqno(n_ptr->bclink.deferred_head) - 1
 368                                  : n_ptr->bclink.last_sent);
 369
 370                 tipc_bclink_lock();
 371                 tipc_bearer_send(MAX_BEARERS, buf, NULL);
 372                 bcl->stats.sent_nacks++;
 373                 tipc_bclink_unlock();
 374                 kfree_skb(buf);
 375
 376                 n_ptr->bclink.oos_state++;
 377         }
 378 }
 379
 380 /**
 381  * bclink_peek_nack - monitor retransmission requests sent by other nodes
 382  *
 383  * Delay any upcoming NACK by this node if another node has already
 384  * requested the first message this node is going to ask for.
 385  */
 386 static void bclink_peek_nack(struct tipc_msg *msg)
 387 {
 388         struct tipc_node *n_ptr = tipc_node_find(msg_destnode(msg));
 389
 390         if (unlikely(!n_ptr))
 391                 return;
 392
 393         tipc_node_lock(n_ptr);
 394
 395         if (n_ptr->bclink.recv_permitted &&
 396             (n_ptr->bclink.last_in != n_ptr->bclink.last_sent) &&
 397             (n_ptr->bclink.last_in == msg_bcgap_after(msg)))
 398                 n_ptr->bclink.oos_state = 2;
 399
 400         tipc_node_unlock(n_ptr);
 401 }
 402
 403 /* tipc_bclink_xmit - broadcast buffer chain to all nodes in cluster
 404  *                    and to identified node local sockets
 405  * @buf: chain of buffers containing message
 406  * Consumes the buffer chain, except when returning -ELINKCONG
 407  * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
 408  */
 409 int tipc_bclink_xmit(struct sk_buff *buf)
 410 {
 411         int rc = 0;
 412         int bc = 0;
 413         struct sk_buff *clbuf;
 414
 415         /* Prepare clone of message for local node */
 416         clbuf = tipc_msg_reassemble(buf);
 417         if (unlikely(!clbuf)) {
 418                 kfree_skb_list(buf);
 419                 return -EHOSTUNREACH;
 420         }
 421
 422         /* Broadcast to all other nodes */
 423         if (likely(bclink)) {
 424                 tipc_bclink_lock();
 425                 if (likely(bclink->bcast_nodes.count)) {
 426                         rc = __tipc_link_xmit(bcl, buf);
 427                         if (likely(!rc)) {
 428                                 bclink_set_last_sent();
 429                                 bcl->stats.queue_sz_counts++;
 430                                 bcl->stats.accu_queue_sz += bcl->out_queue_size;
 431                         }
 432                         bc = 1;
 433                 }
 434                 tipc_bclink_unlock();
 435         }
 436
 437         if (unlikely(!bc))
 438                 kfree_skb_list(buf);
 439
 440         /* Deliver message clone */
 441         if (likely(!rc))
 442                 tipc_sk_mcast_rcv(clbuf);
 443         else
 444                 kfree_skb(clbuf);
 445
 446         return rc;
 447 }
 448
 449 /**
 450  * bclink_accept_pkt - accept an incoming, in-sequence broadcast packet
 451  *
 452  * Called with both sending node's lock and bclink_lock taken.
 453  */
 454 static void bclink_accept_pkt(struct tipc_node *node, u32 seqno)
 455 {
 456         bclink_update_last_sent(node, seqno);
 457         node->bclink.last_in = seqno;
 458         node->bclink.oos_state = 0;
 459         bcl->stats.recv_info++;
 460
 461         /*
 462          * Unicast an ACK periodically, ensuring that
 463          * all nodes in the cluster don't ACK at the same time
 464          */
 465
 466         if (((seqno - tipc_own_addr) % TIPC_MIN_LINK_WIN) == 0) {
 467                 tipc_link_proto_xmit(node->active_links[node->addr & 1],
 468                                      STATE_MSG, 0, 0, 0, 0, 0);
 469                 bcl->stats.sent_acks++;
 470         }
 471 }
 472
 473 /**
 474  * tipc_bclink_rcv - receive a broadcast packet, and deliver upwards
 475  *
 476  * RCU is locked, no other locks set
 477  */
 478 void tipc_bclink_rcv(struct sk_buff *buf)
 479 {
 480         struct tipc_msg *msg = buf_msg(buf);
 481         struct tipc_node *node;
 482         u32 next_in;
 483         u32 seqno;
 484         int deferred = 0;
 485
 486         /* Screen out unwanted broadcast messages */
 487
 488         if (msg_mc_netid(msg) != tipc_net_id)
 489                 goto exit;
 490
 491         node = tipc_node_find(msg_prevnode(msg));
 492         if (unlikely(!node))
 493                 goto exit;
 494
 495         tipc_node_lock(node);
 496         if (unlikely(!node->bclink.recv_permitted))
 497                 goto unlock;
 498
 499         /* Handle broadcast protocol message */
 500
 501         if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
 502                 if (msg_type(msg) != STATE_MSG)
 503                         goto unlock;
 504                 if (msg_destnode(msg) == tipc_own_addr) {
 505                         tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
 506                         tipc_node_unlock(node);
 507                         tipc_bclink_lock();
 508                         bcl->stats.recv_nacks++;
 509                         bclink->retransmit_to = node;
 510                         bclink_retransmit_pkt(msg_bcgap_after(msg),
 511                                               msg_bcgap_to(msg));
 512                         tipc_bclink_unlock();
 513                 } else {
 514                         tipc_node_unlock(node);
 515                         bclink_peek_nack(msg);
 516                 }
 517                 goto exit;
 518         }
 519
 520         /* Handle in-sequence broadcast message */
 521
 522         seqno = msg_seqno(msg);
 523         next_in = mod(node->bclink.last_in + 1);
 524
 525         if (likely(seqno == next_in)) {
 526 receive:
 527                 /* Deliver message to destination */
 528
 529                 if (likely(msg_isdata(msg))) {
 530                         tipc_bclink_lock();
 531                         bclink_accept_pkt(node, seqno);
 532                         tipc_bclink_unlock();
 533                         tipc_node_unlock(node);
 534                         if (likely(msg_mcast(msg)))
 535                                 tipc_sk_mcast_rcv(buf);
 536                         else
 537                                 kfree_skb(buf);
 538                 } else if (msg_user(msg) == MSG_BUNDLER) {
 539                         tipc_bclink_lock();
 540                         bclink_accept_pkt(node, seqno);
 541                         bcl->stats.recv_bundles++;
 542                         bcl->stats.recv_bundled += msg_msgcnt(msg);
 543                         tipc_bclink_unlock();
 544                         tipc_node_unlock(node);
 545                         tipc_link_bundle_rcv(buf);
 546                 } else if (msg_user(msg) == MSG_FRAGMENTER) {
 547                         tipc_buf_append(&node->bclink.reasm_buf, &buf);
 548                         if (unlikely(!buf && !node->bclink.reasm_buf))
 549                                 goto unlock;
 550                         tipc_bclink_lock();
 551                         bclink_accept_pkt(node, seqno);
 552                         bcl->stats.recv_fragments++;
 553                         if (buf) {
 554                                 bcl->stats.recv_fragmented++;
 555                                 msg = buf_msg(buf);
 556                                 tipc_bclink_unlock();
 557                                 goto receive;
 558                         }
 559                         tipc_bclink_unlock();
 560                         tipc_node_unlock(node);
 561                 } else if (msg_user(msg) == NAME_DISTRIBUTOR) {
 562                         tipc_bclink_lock();
 563                         bclink_accept_pkt(node, seqno);
 564                         tipc_bclink_unlock();
 565                         tipc_node_unlock(node);
 566                         tipc_named_rcv(buf);
 567                 } else {
 568                         tipc_bclink_lock();
 569                         bclink_accept_pkt(node, seqno);
 570                         tipc_bclink_unlock();
 571                         tipc_node_unlock(node);
 572                         kfree_skb(buf);
 573                 }
 574                 buf = NULL;
 575
 576                 /* Determine new synchronization state */
 577
 578                 tipc_node_lock(node);
 579                 if (unlikely(!tipc_node_is_up(node)))
 580                         goto unlock;
 581
 582                 if (node->bclink.last_in == node->bclink.last_sent)
 583                         goto unlock;
 584
 585                 if (!node->bclink.deferred_head) {
 586                         node->bclink.oos_state = 1;
 587                         goto unlock;
 588                 }
 589
 590                 msg = buf_msg(node->bclink.deferred_head);
 591                 seqno = msg_seqno(msg);
 592                 next_in = mod(next_in + 1);
 593                 if (seqno != next_in)
 594                         goto unlock;
 595
 596                 /* Take in-sequence message from deferred queue & deliver it */
 597
 598                 buf = node->bclink.deferred_head;
 599                 node->bclink.deferred_head = buf->next;
 600                 buf->next = NULL;
 601                 node->bclink.deferred_size--;
 602                 goto receive;
 603         }
 604
 605         /* Handle out-of-sequence broadcast message */
 606
 607         if (less(next_in, seqno)) {
 608                 deferred = tipc_link_defer_pkt(&node->bclink.deferred_head,
 609                                                &node->bclink.deferred_tail,
 610                                                buf);
 611                 node->bclink.deferred_size += deferred;
 612                 bclink_update_last_sent(node, seqno);
 613                 buf = NULL;
 614         }
 615
 616         tipc_bclink_lock();
 617
 618         if (deferred)
 619                 bcl->stats.deferred_recv++;
 620         else
 621                 bcl->stats.duplicates++;
 622
 623         tipc_bclink_unlock();
 624
 625 unlock:
 626         tipc_node_unlock(node);
 627 exit:
 628         kfree_skb(buf);
 629 }
 630
 631 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
 632 {
 633         return (n_ptr->bclink.recv_permitted &&
 634                 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
 635 }
 636
 637
 638 /**
 639  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
 640  *
 641  * Send packet over as many bearers as necessary to reach all nodes
 642  * that have joined the broadcast link.
 643  *
 644  * Returns 0 (packet sent successfully) under all circumstances,
 645  * since the broadcast link's pseudo-bearer never blocks
 646  */
 647 static int tipc_bcbearer_send(struct sk_buff *buf, struct tipc_bearer *unused1,
 648                               struct tipc_media_addr *unused2)
 649 {
 650         int bp_index;
 651         struct tipc_msg *msg = buf_msg(buf);
 652
 653         /* Prepare broadcast link message for reliable transmission,
 654          * if first time trying to send it;
 655          * preparation is skipped for broadcast link protocol messages
 656          * since they are sent in an unreliable manner and don't need it
 657          */
 658         if (likely(!msg_non_seq(buf_msg(buf)))) {
 659                 bcbuf_set_acks(buf, bclink->bcast_nodes.count);
 660                 msg_set_non_seq(msg, 1);
 661                 msg_set_mc_netid(msg, tipc_net_id);
 662                 bcl->stats.sent_info++;
 663
 664                 if (WARN_ON(!bclink->bcast_nodes.count)) {
 665                         dump_stack();
 666                         return 0;
 667                 }
 668         }
 669
 670         /* Send buffer over bearers until all targets reached */
 671         bcbearer->remains = bclink->bcast_nodes;
 672
 673         for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
 674                 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
 675                 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
 676                 struct tipc_bearer *bp[2] = {p, s};
 677                 struct tipc_bearer *b = bp[msg_link_selector(msg)];
 678                 struct sk_buff *tbuf;
 679
 680                 if (!p)
 681                         break; /* No more bearers to try */
 682                 if (!b)
 683                         b = p;
 684                 tipc_nmap_diff(&bcbearer->remains, &b->nodes,
 685                                &bcbearer->remains_new);
 686                 if (bcbearer->remains_new.count == bcbearer->remains.count)
 687                         continue; /* Nothing added by bearer pair */
 688
 689                 if (bp_index == 0) {
 690                         /* Use original buffer for first bearer */
 691                         tipc_bearer_send(b->identity, buf, &b->bcast_addr);
 692                 } else {
 693                         /* Avoid concurrent buffer access */
 694                         tbuf = pskb_copy_for_clone(buf, GFP_ATOMIC);
 695                         if (!tbuf)
 696                                 break;
 697                         tipc_bearer_send(b->identity, tbuf, &b->bcast_addr);
 698                         kfree_skb(tbuf); /* Bearer keeps a clone */
 699                 }
 700                 if (bcbearer->remains_new.count == 0)
 701                         break; /* All targets reached */
 702
 703                 bcbearer->remains = bcbearer->remains_new;
 704         }
 705
 706         return 0;
 707 }
 708
 709 /**
 710  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
 711  */
 712 void tipc_bcbearer_sort(struct tipc_node_map *nm_ptr, u32 node, bool action)
 713 {
 714         struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
 715         struct tipc_bcbearer_pair *bp_curr;
 716         struct tipc_bearer *b;
 717         int b_index;
 718         int pri;
 719
 720         tipc_bclink_lock();
 721
 722         if (action)
 723                 tipc_nmap_add(nm_ptr, node);
 724         else
 725                 tipc_nmap_remove(nm_ptr, node);
 726
 727         /* Group bearers by priority (can assume max of two per priority) */
 728         memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
 729
 730         rcu_read_lock();
 731         for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
 732                 b = rcu_dereference_rtnl(bearer_list[b_index]);
 733                 if (!b || !b->nodes.count)
 734                         continue;
 735
 736                 if (!bp_temp[b->priority].primary)
 737                         bp_temp[b->priority].primary = b;
 738                 else
 739                         bp_temp[b->priority].secondary = b;
 740         }
 741         rcu_read_unlock();
 742
 743         /* Create array of bearer pairs for broadcasting */
 744         bp_curr = bcbearer->bpairs;
 745         memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
 746
 747         for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
 748
 749                 if (!bp_temp[pri].primary)
 750                         continue;
 751
 752                 bp_curr->primary = bp_temp[pri].primary;
 753
 754                 if (bp_temp[pri].secondary) {
 755                         if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
 756                                             &bp_temp[pri].secondary->nodes)) {
 757                                 bp_curr->secondary = bp_temp[pri].secondary;
 758                         } else {
 759                                 bp_curr++;
 760                                 bp_curr->primary = bp_temp[pri].secondary;
 761                         }
 762                 }
 763
 764                 bp_curr++;
 765         }
 766
 767         tipc_bclink_unlock();
 768 }
 769
 770
 771 int tipc_bclink_stats(char *buf, const u32 buf_size)
 772 {
 773         int ret;
 774         struct tipc_stats *s;
 775
 776         if (!bcl)
 777                 return 0;
 778
 779         tipc_bclink_lock();
 780
 781         s = &bcl->stats;
 782
 783         ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
 784                             "  Window:%u packets\n",
 785                             bcl->name, bcl->queue_limit[0]);
 786         ret += tipc_snprintf(buf + ret, buf_size - ret,
 787                              "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
 788                              s->recv_info, s->recv_fragments,
 789                              s->recv_fragmented, s->recv_bundles,
 790                              s->recv_bundled);
 791         ret += tipc_snprintf(buf + ret, buf_size - ret,
 792                              "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
 793                              s->sent_info, s->sent_fragments,
 794                              s->sent_fragmented, s->sent_bundles,
 795                              s->sent_bundled);
 796         ret += tipc_snprintf(buf + ret, buf_size - ret,
 797                              "  RX naks:%u defs:%u dups:%u\n",
 798                              s->recv_nacks, s->deferred_recv, s->duplicates);
 799         ret += tipc_snprintf(buf + ret, buf_size - ret,
 800                              "  TX naks:%u acks:%u dups:%u\n",
 801                              s->sent_nacks, s->sent_acks, s->retransmitted);
 802         ret += tipc_snprintf(buf + ret, buf_size - ret,
 803                              "  Congestion link:%u  Send queue max:%u avg:%u\n",
 804                              s->link_congs, s->max_queue_sz,
 805                              s->queue_sz_counts ?
 806                              (s->accu_queue_sz / s->queue_sz_counts) : 0);
 807
 808         tipc_bclink_unlock();
 809         return ret;
 810 }
 811
 812 int tipc_bclink_reset_stats(void)
 813 {
 814         if (!bcl)
 815                 return -ENOPROTOOPT;
 816
 817         tipc_bclink_lock();
 818         memset(&bcl->stats, 0, sizeof(bcl->stats));
 819         tipc_bclink_unlock();
 820         return 0;
 821 }
 822
 823 int tipc_bclink_set_queue_limits(u32 limit)
 824 {
 825         if (!bcl)
 826                 return -ENOPROTOOPT;
 827         if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
 828                 return -EINVAL;
 829
 830         tipc_bclink_lock();
 831         tipc_link_set_queue_limits(bcl, limit);
 832         tipc_bclink_unlock();
 833         return 0;
 834 }
 835
 836 int tipc_bclink_init(void)
 837 {
 838         bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
 839         if (!bcbearer)
 840                 return -ENOMEM;
 841
 842         bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
 843         if (!bclink) {
 844                 kfree(bcbearer);
 845                 return -ENOMEM;
 846         }
 847
 848         bcl = &bclink->link;
 849         bcbearer->bearer.media = &bcbearer->media;
 850         bcbearer->media.send_msg = tipc_bcbearer_send;
 851         sprintf(bcbearer->media.name, "tipc-broadcast");
 852
 853         spin_lock_init(&bclink->lock);
 854         __skb_queue_head_init(&bcl->waiting_sks);
 855         bcl->next_out_no = 1;
 856         spin_lock_init(&bclink->node.lock);
 857         __skb_queue_head_init(&bclink->node.waiting_sks);
 858         bcl->owner = &bclink->node;
 859         bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
 860         tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
 861         bcl->bearer_id = MAX_BEARERS;
 862         rcu_assign_pointer(bearer_list[MAX_BEARERS], &bcbearer->bearer);
 863         bcl->state = WORKING_WORKING;
 864         strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
 865         return 0;
 866 }
 867
 868 void tipc_bclink_stop(void)
 869 {
 870         tipc_bclink_lock();
 871         tipc_link_purge_queues(bcl);
 872         tipc_bclink_unlock();
 873
 874         RCU_INIT_POINTER(bearer_list[BCBEARER], NULL);
 875         synchronize_net();
 876         kfree(bcbearer);
 877         kfree(bclink);
 878 }
 879
 880 /**
 881  * tipc_nmap_add - add a node to a node map
 882  */
 883 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
 884 {
 885         int n = tipc_node(node);
 886         int w = n / WSIZE;
 887         u32 mask = (1 << (n % WSIZE));
 888
 889         if ((nm_ptr->map[w] & mask) == 0) {
 890                 nm_ptr->count++;
 891                 nm_ptr->map[w] |= mask;
 892         }
 893 }
 894
 895 /**
 896  * tipc_nmap_remove - remove a node from a node map
 897  */
 898 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
 899 {
 900         int n = tipc_node(node);
 901         int w = n / WSIZE;
 902         u32 mask = (1 << (n % WSIZE));
 903
 904         if ((nm_ptr->map[w] & mask) != 0) {
 905                 nm_ptr->map[w] &= ~mask;
 906                 nm_ptr->count--;
 907         }
 908 }
 909
 910 /**
 911  * tipc_nmap_diff - find differences between node maps
 912  * @nm_a: input node map A
 913  * @nm_b: input node map B
 914  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
 915  */
 916 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
 917                            struct tipc_node_map *nm_b,
 918                            struct tipc_node_map *nm_diff)
 919 {
 920         int stop = ARRAY_SIZE(nm_a->map);
 921         int w;
 922         int b;
 923         u32 map;
 924
 925         memset(nm_diff, 0, sizeof(*nm_diff));
 926         for (w = 0; w < stop; w++) {
 927                 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
 928                 nm_diff->map[w] = map;
 929                 if (map != 0) {
 930                         for (b = 0 ; b < WSIZE; b++) {
 931                                 if (map & (1 << b))
 932                                         nm_diff->count++;
 933                         }
 934                 }
 935         }
 936 }
 937
 938 /**
 939  * tipc_port_list_add - add a port to a port list, ensuring no duplicates
 940  */
 941 void tipc_port_list_add(struct tipc_port_list *pl_ptr, u32 port)
 942 {
 943         struct tipc_port_list *item = pl_ptr;
 944         int i;
 945         int item_sz = PLSIZE;
 946         int cnt = pl_ptr->count;
 947
 948         for (; ; cnt -= item_sz, item = item->next) {
 949                 if (cnt < PLSIZE)
 950                         item_sz = cnt;
 951                 for (i = 0; i < item_sz; i++)
 952                         if (item->ports[i] == port)
 953                                 return;
 954                 if (i < PLSIZE) {
 955                         item->ports[i] = port;
 956                         pl_ptr->count++;
 957                         return;
 958                 }
 959                 if (!item->next) {
 960                         item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
 961                         if (!item->next) {
 962                                 pr_warn("Incomplete multicast delivery, no memory\n");
 963                                 return;
 964                         }
 965                         item->next->next = NULL;
 966                 }
 967         }
 968 }
 969
 970 /**
 971  * tipc_port_list_free - free dynamically created entries in port_list chain
 972  *
 973  */
 974 void tipc_port_list_free(struct tipc_port_list *pl_ptr)
 975 {
 976         struct tipc_port_list *item;
 977         struct tipc_port_list *next;
 978
 979         for (item = pl_ptr->next; item; item = next) {
 980                 next = item->next;
 981                 kfree(item);
 982         }
 983 }