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 *skb;
 221
 222         skb_queue_walk(&bcl->outqueue, skb) {
 223                 if (more(buf_seqno(skb), after)) {
 224                         tipc_link_retransmit(bcl, skb, mod(to - after));
 225                         break;
 226                 }
 227         }
 228 }
 229
 230 /**
 231  * tipc_bclink_wakeup_users - wake up pending users
 232  *
 233  * Called with no locks taken
 234  */
 235 void tipc_bclink_wakeup_users(struct net *net)
 236 {
 237         struct sk_buff *skb;
 238
 239         while ((skb = skb_dequeue(&bclink->link.waiting_sks)))
 240                 tipc_sk_rcv(net, skb);
 241 }
 242
 243 /**
 244  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
 245  * @n_ptr: node that sent acknowledgement info
 246  * @acked: broadcast sequence # that has been acknowledged
 247  *
 248  * Node is locked, bclink_lock unlocked.
 249  */
 250 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
 251 {
 252         struct sk_buff *skb, *tmp;
 253         struct sk_buff *next;
 254         unsigned int released = 0;
 255
 256         tipc_bclink_lock();
 257         /* Bail out if tx queue is empty (no clean up is required) */
 258         skb = skb_peek(&bcl->outqueue);
 259         if (!skb)
 260                 goto exit;
 261
 262         /* Determine which messages need to be acknowledged */
 263         if (acked == INVALID_LINK_SEQ) {
 264                 /*
 265                  * Contact with specified node has been lost, so need to
 266                  * acknowledge sent messages only (if other nodes still exist)
 267                  * or both sent and unsent messages (otherwise)
 268                  */
 269                 if (bclink->bcast_nodes.count)
 270                         acked = bcl->fsm_msg_cnt;
 271                 else
 272                         acked = bcl->next_out_no;
 273         } else {
 274                 /*
 275                  * Bail out if specified sequence number does not correspond
 276                  * to a message that has been sent and not yet acknowledged
 277                  */
 278                 if (less(acked, buf_seqno(skb)) ||
 279                     less(bcl->fsm_msg_cnt, acked) ||
 280                     less_eq(acked, n_ptr->bclink.acked))
 281                         goto exit;
 282         }
 283
 284         /* Skip over packets that node has previously acknowledged */
 285         skb_queue_walk(&bcl->outqueue, skb) {
 286                 if (more(buf_seqno(skb), n_ptr->bclink.acked))
 287                         break;
 288         }
 289
 290         /* Update packets that node is now acknowledging */
 291         skb_queue_walk_from_safe(&bcl->outqueue, skb, tmp) {
 292                 if (more(buf_seqno(skb), acked))
 293                         break;
 294
 295                 next = tipc_skb_queue_next(&bcl->outqueue, skb);
 296                 if (skb != bcl->next_out) {
 297                         bcbuf_decr_acks(skb);
 298                 } else {
 299                         bcbuf_set_acks(skb, 0);
 300                         bcl->next_out = next;
 301                         bclink_set_last_sent();
 302                 }
 303
 304                 if (bcbuf_acks(skb) == 0) {
 305                         __skb_unlink(skb, &bcl->outqueue);
 306                         kfree_skb(skb);
 307                         released = 1;
 308                 }
 309         }
 310         n_ptr->bclink.acked = acked;
 311
 312         /* Try resolving broadcast link congestion, if necessary */
 313         if (unlikely(bcl->next_out)) {
 314                 tipc_link_push_packets(bcl);
 315                 bclink_set_last_sent();
 316         }
 317         if (unlikely(released && !skb_queue_empty(&bcl->waiting_sks)))
 318                 n_ptr->action_flags |= TIPC_WAKEUP_BCAST_USERS;
 319
 320 exit:
 321         tipc_bclink_unlock();
 322 }
 323
 324 /**
 325  * tipc_bclink_update_link_state - update broadcast link state
 326  *
 327  * RCU and node lock set
 328  */
 329 void tipc_bclink_update_link_state(struct net *net, struct tipc_node *n_ptr,
 330                                    u32 last_sent)
 331 {
 332         struct sk_buff *buf;
 333         struct tipc_net *tn = net_generic(net, tipc_net_id);
 334
 335         /* Ignore "stale" link state info */
 336         if (less_eq(last_sent, n_ptr->bclink.last_in))
 337                 return;
 338
 339         /* Update link synchronization state; quit if in sync */
 340         bclink_update_last_sent(n_ptr, last_sent);
 341
 342         if (n_ptr->bclink.last_sent == n_ptr->bclink.last_in)
 343                 return;
 344
 345         /* Update out-of-sync state; quit if loss is still unconfirmed */
 346         if ((++n_ptr->bclink.oos_state) == 1) {
 347                 if (n_ptr->bclink.deferred_size < (TIPC_MIN_LINK_WIN / 2))
 348                         return;
 349                 n_ptr->bclink.oos_state++;
 350         }
 351
 352         /* Don't NACK if one has been recently sent (or seen) */
 353         if (n_ptr->bclink.oos_state & 0x1)
 354                 return;
 355
 356         /* Send NACK */
 357         buf = tipc_buf_acquire(INT_H_SIZE);
 358         if (buf) {
 359                 struct tipc_msg *msg = buf_msg(buf);
 360                 struct sk_buff *skb = skb_peek(&n_ptr->bclink.deferred_queue);
 361                 u32 to = skb ? buf_seqno(skb) - 1 : n_ptr->bclink.last_sent;
 362
 363                 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
 364                               INT_H_SIZE, n_ptr->addr);
 365                 msg_set_non_seq(msg, 1);
 366                 msg_set_mc_netid(msg, tn->net_id);
 367                 msg_set_bcast_ack(msg, n_ptr->bclink.last_in);
 368                 msg_set_bcgap_after(msg, n_ptr->bclink.last_in);
 369                 msg_set_bcgap_to(msg, to);
 370
 371                 tipc_bclink_lock();
 372                 tipc_bearer_send(MAX_BEARERS, buf, NULL);
 373                 bcl->stats.sent_nacks++;
 374                 tipc_bclink_unlock();
 375                 kfree_skb(buf);
 376
 377                 n_ptr->bclink.oos_state++;
 378         }
 379 }
 380
 381 /**
 382  * bclink_peek_nack - monitor retransmission requests sent by other nodes
 383  *
 384  * Delay any upcoming NACK by this node if another node has already
 385  * requested the first message this node is going to ask for.
 386  */
 387 static void bclink_peek_nack(struct net *net, struct tipc_msg *msg)
 388 {
 389         struct tipc_node *n_ptr = tipc_node_find(net, msg_destnode(msg));
 390
 391         if (unlikely(!n_ptr))
 392                 return;
 393
 394         tipc_node_lock(n_ptr);
 395
 396         if (n_ptr->bclink.recv_permitted &&
 397             (n_ptr->bclink.last_in != n_ptr->bclink.last_sent) &&
 398             (n_ptr->bclink.last_in == msg_bcgap_after(msg)))
 399                 n_ptr->bclink.oos_state = 2;
 400
 401         tipc_node_unlock(n_ptr);
 402 }
 403
 404 /* tipc_bclink_xmit - broadcast buffer chain to all nodes in cluster
 405  *                    and to identified node local sockets
 406  * @net: the applicable net namespace
 407  * @list: chain of buffers containing message
 408  * Consumes the buffer chain, except when returning -ELINKCONG
 409  * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
 410  */
 411 int tipc_bclink_xmit(struct net *net, struct sk_buff_head *list)
 412 {
 413         int rc = 0;
 414         int bc = 0;
 415         struct sk_buff *skb;
 416
 417         /* Prepare clone of message for local node */
 418         skb = tipc_msg_reassemble(list);
 419         if (unlikely(!skb)) {
 420                 __skb_queue_purge(list);
 421                 return -EHOSTUNREACH;
 422         }
 423
 424         /* Broadcast to all other nodes */
 425         if (likely(bclink)) {
 426                 tipc_bclink_lock();
 427                 if (likely(bclink->bcast_nodes.count)) {
 428                         rc = __tipc_link_xmit(bcl, list);
 429                         if (likely(!rc)) {
 430                                 u32 len = skb_queue_len(&bcl->outqueue);
 431
 432                                 bclink_set_last_sent();
 433                                 bcl->stats.queue_sz_counts++;
 434                                 bcl->stats.accu_queue_sz += len;
 435                         }
 436                         bc = 1;
 437                 }
 438                 tipc_bclink_unlock();
 439         }
 440
 441         if (unlikely(!bc))
 442                 __skb_queue_purge(list);
 443
 444         /* Deliver message clone */
 445         if (likely(!rc))
 446                 tipc_sk_mcast_rcv(net, skb);
 447         else
 448                 kfree_skb(skb);
 449
 450         return rc;
 451 }
 452
 453 /**
 454  * bclink_accept_pkt - accept an incoming, in-sequence broadcast packet
 455  *
 456  * Called with both sending node's lock and bclink_lock taken.
 457  */
 458 static void bclink_accept_pkt(struct tipc_node *node, u32 seqno)
 459 {
 460         bclink_update_last_sent(node, seqno);
 461         node->bclink.last_in = seqno;
 462         node->bclink.oos_state = 0;
 463         bcl->stats.recv_info++;
 464
 465         /*
 466          * Unicast an ACK periodically, ensuring that
 467          * all nodes in the cluster don't ACK at the same time
 468          */
 469         if (((seqno - tipc_own_addr) % TIPC_MIN_LINK_WIN) == 0) {
 470                 tipc_link_proto_xmit(node->active_links[node->addr & 1],
 471                                      STATE_MSG, 0, 0, 0, 0, 0);
 472                 bcl->stats.sent_acks++;
 473         }
 474 }
 475
 476 /**
 477  * tipc_bclink_rcv - receive a broadcast packet, and deliver upwards
 478  *
 479  * RCU is locked, no other locks set
 480  */
 481 void tipc_bclink_rcv(struct net *net, struct sk_buff *buf)
 482 {
 483         struct tipc_net *tn = net_generic(net, tipc_net_id);
 484         struct tipc_msg *msg = buf_msg(buf);
 485         struct tipc_node *node;
 486         u32 next_in;
 487         u32 seqno;
 488         int deferred = 0;
 489
 490         /* Screen out unwanted broadcast messages */
 491         if (msg_mc_netid(msg) != tn->net_id)
 492                 goto exit;
 493
 494         node = tipc_node_find(net, msg_prevnode(msg));
 495         if (unlikely(!node))
 496                 goto exit;
 497
 498         tipc_node_lock(node);
 499         if (unlikely(!node->bclink.recv_permitted))
 500                 goto unlock;
 501
 502         /* Handle broadcast protocol message */
 503         if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
 504                 if (msg_type(msg) != STATE_MSG)
 505                         goto unlock;
 506                 if (msg_destnode(msg) == tipc_own_addr) {
 507                         tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
 508                         tipc_node_unlock(node);
 509                         tipc_bclink_lock();
 510                         bcl->stats.recv_nacks++;
 511                         bclink->retransmit_to = node;
 512                         bclink_retransmit_pkt(msg_bcgap_after(msg),
 513                                               msg_bcgap_to(msg));
 514                         tipc_bclink_unlock();
 515                 } else {
 516                         tipc_node_unlock(node);
 517                         bclink_peek_nack(net, msg);
 518                 }
 519                 goto exit;
 520         }
 521
 522         /* Handle in-sequence broadcast message */
 523         seqno = msg_seqno(msg);
 524         next_in = mod(node->bclink.last_in + 1);
 525
 526         if (likely(seqno == next_in)) {
 527 receive:
 528                 /* Deliver message to destination */
 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(net, 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(net, 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(net, 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                 tipc_node_lock(node);
 578                 if (unlikely(!tipc_node_is_up(node)))
 579                         goto unlock;
 580
 581                 if (node->bclink.last_in == node->bclink.last_sent)
 582                         goto unlock;
 583
 584                 if (skb_queue_empty(&node->bclink.deferred_queue)) {
 585                         node->bclink.oos_state = 1;
 586                         goto unlock;
 587                 }
 588
 589                 msg = buf_msg(skb_peek(&node->bclink.deferred_queue));
 590                 seqno = msg_seqno(msg);
 591                 next_in = mod(next_in + 1);
 592                 if (seqno != next_in)
 593                         goto unlock;
 594
 595                 /* Take in-sequence message from deferred queue & deliver it */
 596                 buf = __skb_dequeue(&node->bclink.deferred_queue);
 597                 goto receive;
 598         }
 599
 600         /* Handle out-of-sequence broadcast message */
 601         if (less(next_in, seqno)) {
 602                 deferred = tipc_link_defer_pkt(&node->bclink.deferred_queue,
 603                                                buf);
 604                 bclink_update_last_sent(node, seqno);
 605                 buf = NULL;
 606         }
 607
 608         tipc_bclink_lock();
 609
 610         if (deferred)
 611                 bcl->stats.deferred_recv++;
 612         else
 613                 bcl->stats.duplicates++;
 614
 615         tipc_bclink_unlock();
 616
 617 unlock:
 618         tipc_node_unlock(node);
 619 exit:
 620         kfree_skb(buf);
 621 }
 622
 623 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
 624 {
 625         return (n_ptr->bclink.recv_permitted &&
 626                 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
 627 }
 628
 629
 630 /**
 631  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
 632  *
 633  * Send packet over as many bearers as necessary to reach all nodes
 634  * that have joined the broadcast link.
 635  *
 636  * Returns 0 (packet sent successfully) under all circumstances,
 637  * since the broadcast link's pseudo-bearer never blocks
 638  */
 639 static int tipc_bcbearer_send(struct sk_buff *buf, struct tipc_bearer *unused1,
 640                               struct tipc_media_addr *unused2)
 641 {
 642         int bp_index;
 643         struct tipc_msg *msg = buf_msg(buf);
 644         struct net *net = sock_net(buf->sk);
 645         struct tipc_net *tn = net_generic(net, tipc_net_id);
 646
 647         /* Prepare broadcast link message for reliable transmission,
 648          * if first time trying to send it;
 649          * preparation is skipped for broadcast link protocol messages
 650          * since they are sent in an unreliable manner and don't need it
 651          */
 652         if (likely(!msg_non_seq(buf_msg(buf)))) {
 653                 bcbuf_set_acks(buf, bclink->bcast_nodes.count);
 654                 msg_set_non_seq(msg, 1);
 655                 msg_set_mc_netid(msg, tn->net_id);
 656                 bcl->stats.sent_info++;
 657
 658                 if (WARN_ON(!bclink->bcast_nodes.count)) {
 659                         dump_stack();
 660                         return 0;
 661                 }
 662         }
 663
 664         /* Send buffer over bearers until all targets reached */
 665         bcbearer->remains = bclink->bcast_nodes;
 666
 667         for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
 668                 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
 669                 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
 670                 struct tipc_bearer *bp[2] = {p, s};
 671                 struct tipc_bearer *b = bp[msg_link_selector(msg)];
 672                 struct sk_buff *tbuf;
 673
 674                 if (!p)
 675                         break; /* No more bearers to try */
 676                 if (!b)
 677                         b = p;
 678                 tipc_nmap_diff(&bcbearer->remains, &b->nodes,
 679                                &bcbearer->remains_new);
 680                 if (bcbearer->remains_new.count == bcbearer->remains.count)
 681                         continue; /* Nothing added by bearer pair */
 682
 683                 if (bp_index == 0) {
 684                         /* Use original buffer for first bearer */
 685                         tipc_bearer_send(b->identity, buf, &b->bcast_addr);
 686                 } else {
 687                         /* Avoid concurrent buffer access */
 688                         tbuf = pskb_copy_for_clone(buf, GFP_ATOMIC);
 689                         if (!tbuf)
 690                                 break;
 691                         tipc_bearer_send(b->identity, tbuf, &b->bcast_addr);
 692                         kfree_skb(tbuf); /* Bearer keeps a clone */
 693                 }
 694                 if (bcbearer->remains_new.count == 0)
 695                         break; /* All targets reached */
 696
 697                 bcbearer->remains = bcbearer->remains_new;
 698         }
 699
 700         return 0;
 701 }
 702
 703 /**
 704  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
 705  */
 706 void tipc_bcbearer_sort(struct tipc_node_map *nm_ptr, u32 node, bool action)
 707 {
 708         struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
 709         struct tipc_bcbearer_pair *bp_curr;
 710         struct tipc_bearer *b;
 711         int b_index;
 712         int pri;
 713
 714         tipc_bclink_lock();
 715
 716         if (action)
 717                 tipc_nmap_add(nm_ptr, node);
 718         else
 719                 tipc_nmap_remove(nm_ptr, node);
 720
 721         /* Group bearers by priority (can assume max of two per priority) */
 722         memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
 723
 724         rcu_read_lock();
 725         for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
 726                 b = rcu_dereference_rtnl(bearer_list[b_index]);
 727                 if (!b || !b->nodes.count)
 728                         continue;
 729
 730                 if (!bp_temp[b->priority].primary)
 731                         bp_temp[b->priority].primary = b;
 732                 else
 733                         bp_temp[b->priority].secondary = b;
 734         }
 735         rcu_read_unlock();
 736
 737         /* Create array of bearer pairs for broadcasting */
 738         bp_curr = bcbearer->bpairs;
 739         memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
 740
 741         for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
 742
 743                 if (!bp_temp[pri].primary)
 744                         continue;
 745
 746                 bp_curr->primary = bp_temp[pri].primary;
 747
 748                 if (bp_temp[pri].secondary) {
 749                         if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
 750                                             &bp_temp[pri].secondary->nodes)) {
 751                                 bp_curr->secondary = bp_temp[pri].secondary;
 752                         } else {
 753                                 bp_curr++;
 754                                 bp_curr->primary = bp_temp[pri].secondary;
 755                         }
 756                 }
 757
 758                 bp_curr++;
 759         }
 760
 761         tipc_bclink_unlock();
 762 }
 763
 764 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
 765                                       struct tipc_stats *stats)
 766 {
 767         int i;
 768         struct nlattr *nest;
 769
 770         struct nla_map {
 771                 __u32 key;
 772                 __u32 val;
 773         };
 774
 775         struct nla_map map[] = {
 776                 {TIPC_NLA_STATS_RX_INFO, stats->recv_info},
 777                 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
 778                 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
 779                 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
 780                 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
 781                 {TIPC_NLA_STATS_TX_INFO, stats->sent_info},
 782                 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
 783                 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
 784                 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
 785                 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
 786                 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
 787                 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
 788                 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
 789                 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
 790                 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
 791                 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
 792                 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
 793                 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
 794                 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
 795                         (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
 796         };
 797
 798         nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
 799         if (!nest)
 800                 return -EMSGSIZE;
 801
 802         for (i = 0; i <  ARRAY_SIZE(map); i++)
 803                 if (nla_put_u32(skb, map[i].key, map[i].val))
 804                         goto msg_full;
 805
 806         nla_nest_end(skb, nest);
 807
 808         return 0;
 809 msg_full:
 810         nla_nest_cancel(skb, nest);
 811
 812         return -EMSGSIZE;
 813 }
 814
 815 int tipc_nl_add_bc_link(struct tipc_nl_msg *msg)
 816 {
 817         int err;
 818         void *hdr;
 819         struct nlattr *attrs;
 820         struct nlattr *prop;
 821
 822         if (!bcl)
 823                 return 0;
 824
 825         tipc_bclink_lock();
 826
 827         hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_v2_family,
 828                           NLM_F_MULTI, TIPC_NL_LINK_GET);
 829         if (!hdr)
 830                 return -EMSGSIZE;
 831
 832         attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
 833         if (!attrs)
 834                 goto msg_full;
 835
 836         /* The broadcast link is always up */
 837         if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
 838                 goto attr_msg_full;
 839
 840         if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
 841                 goto attr_msg_full;
 842         if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
 843                 goto attr_msg_full;
 844         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, bcl->next_in_no))
 845                 goto attr_msg_full;
 846         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, bcl->next_out_no))
 847                 goto attr_msg_full;
 848
 849         prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
 850         if (!prop)
 851                 goto attr_msg_full;
 852         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->queue_limit[0]))
 853                 goto prop_msg_full;
 854         nla_nest_end(msg->skb, prop);
 855
 856         err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
 857         if (err)
 858                 goto attr_msg_full;
 859
 860         tipc_bclink_unlock();
 861         nla_nest_end(msg->skb, attrs);
 862         genlmsg_end(msg->skb, hdr);
 863
 864         return 0;
 865
 866 prop_msg_full:
 867         nla_nest_cancel(msg->skb, prop);
 868 attr_msg_full:
 869         nla_nest_cancel(msg->skb, attrs);
 870 msg_full:
 871         tipc_bclink_unlock();
 872         genlmsg_cancel(msg->skb, hdr);
 873
 874         return -EMSGSIZE;
 875 }
 876
 877 int tipc_bclink_stats(char *buf, const u32 buf_size)
 878 {
 879         int ret;
 880         struct tipc_stats *s;
 881
 882         if (!bcl)
 883                 return 0;
 884
 885         tipc_bclink_lock();
 886
 887         s = &bcl->stats;
 888
 889         ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
 890                             "  Window:%u packets\n",
 891                             bcl->name, bcl->queue_limit[0]);
 892         ret += tipc_snprintf(buf + ret, buf_size - ret,
 893                              "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
 894                              s->recv_info, s->recv_fragments,
 895                              s->recv_fragmented, s->recv_bundles,
 896                              s->recv_bundled);
 897         ret += tipc_snprintf(buf + ret, buf_size - ret,
 898                              "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
 899                              s->sent_info, s->sent_fragments,
 900                              s->sent_fragmented, s->sent_bundles,
 901                              s->sent_bundled);
 902         ret += tipc_snprintf(buf + ret, buf_size - ret,
 903                              "  RX naks:%u defs:%u dups:%u\n",
 904                              s->recv_nacks, s->deferred_recv, s->duplicates);
 905         ret += tipc_snprintf(buf + ret, buf_size - ret,
 906                              "  TX naks:%u acks:%u dups:%u\n",
 907                              s->sent_nacks, s->sent_acks, s->retransmitted);
 908         ret += tipc_snprintf(buf + ret, buf_size - ret,
 909                              "  Congestion link:%u  Send queue max:%u avg:%u\n",
 910                              s->link_congs, s->max_queue_sz,
 911                              s->queue_sz_counts ?
 912                              (s->accu_queue_sz / s->queue_sz_counts) : 0);
 913
 914         tipc_bclink_unlock();
 915         return ret;
 916 }
 917
 918 int tipc_bclink_reset_stats(void)
 919 {
 920         if (!bcl)
 921                 return -ENOPROTOOPT;
 922
 923         tipc_bclink_lock();
 924         memset(&bcl->stats, 0, sizeof(bcl->stats));
 925         tipc_bclink_unlock();
 926         return 0;
 927 }
 928
 929 int tipc_bclink_set_queue_limits(u32 limit)
 930 {
 931         if (!bcl)
 932                 return -ENOPROTOOPT;
 933         if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
 934                 return -EINVAL;
 935
 936         tipc_bclink_lock();
 937         tipc_link_set_queue_limits(bcl, limit);
 938         tipc_bclink_unlock();
 939         return 0;
 940 }
 941
 942 int tipc_bclink_init(void)
 943 {
 944         bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
 945         if (!bcbearer)
 946                 return -ENOMEM;
 947
 948         bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
 949         if (!bclink) {
 950                 kfree(bcbearer);
 951                 return -ENOMEM;
 952         }
 953
 954         bcl = &bclink->link;
 955         bcbearer->bearer.media = &bcbearer->media;
 956         bcbearer->media.send_msg = tipc_bcbearer_send;
 957         sprintf(bcbearer->media.name, "tipc-broadcast");
 958
 959         spin_lock_init(&bclink->lock);
 960         __skb_queue_head_init(&bcl->outqueue);
 961         __skb_queue_head_init(&bcl->deferred_queue);
 962         skb_queue_head_init(&bcl->waiting_sks);
 963         bcl->next_out_no = 1;
 964         spin_lock_init(&bclink->node.lock);
 965         __skb_queue_head_init(&bclink->node.waiting_sks);
 966         bcl->owner = &bclink->node;
 967         bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
 968         tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
 969         bcl->bearer_id = MAX_BEARERS;
 970         rcu_assign_pointer(bearer_list[MAX_BEARERS], &bcbearer->bearer);
 971         bcl->state = WORKING_WORKING;
 972         strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
 973         return 0;
 974 }
 975
 976 void tipc_bclink_stop(void)
 977 {
 978         tipc_bclink_lock();
 979         tipc_link_purge_queues(bcl);
 980         tipc_bclink_unlock();
 981
 982         RCU_INIT_POINTER(bearer_list[BCBEARER], NULL);
 983         synchronize_net();
 984         kfree(bcbearer);
 985         kfree(bclink);
 986 }
 987
 988 /**
 989  * tipc_nmap_add - add a node to a node map
 990  */
 991 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
 992 {
 993         int n = tipc_node(node);
 994         int w = n / WSIZE;
 995         u32 mask = (1 << (n % WSIZE));
 996
 997         if ((nm_ptr->map[w] & mask) == 0) {
 998                 nm_ptr->count++;
 999                 nm_ptr->map[w] |= mask;
1000         }
1001 }
1002
1003 /**
1004  * tipc_nmap_remove - remove a node from a node map
1005  */
1006 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
1007 {
1008         int n = tipc_node(node);
1009         int w = n / WSIZE;
1010         u32 mask = (1 << (n % WSIZE));
1011
1012         if ((nm_ptr->map[w] & mask) != 0) {
1013                 nm_ptr->map[w] &= ~mask;
1014                 nm_ptr->count--;
1015         }
1016 }
1017
1018 /**
1019  * tipc_nmap_diff - find differences between node maps
1020  * @nm_a: input node map A
1021  * @nm_b: input node map B
1022  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
1023  */
1024 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
1025                            struct tipc_node_map *nm_b,
1026                            struct tipc_node_map *nm_diff)
1027 {
1028         int stop = ARRAY_SIZE(nm_a->map);
1029         int w;
1030         int b;
1031         u32 map;
1032
1033         memset(nm_diff, 0, sizeof(*nm_diff));
1034         for (w = 0; w < stop; w++) {
1035                 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
1036                 nm_diff->map[w] = map;
1037                 if (map != 0) {
1038                         for (b = 0 ; b < WSIZE; b++) {
1039                                 if (map & (1 << b))
1040                                         nm_diff->count++;
1041                         }
1042                 }
1043         }
1044 }
1045
1046 /**
1047  * tipc_port_list_add - add a port to a port list, ensuring no duplicates
1048  */
1049 void tipc_port_list_add(struct tipc_port_list *pl_ptr, u32 port)
1050 {
1051         struct tipc_port_list *item = pl_ptr;
1052         int i;
1053         int item_sz = PLSIZE;
1054         int cnt = pl_ptr->count;
1055
1056         for (; ; cnt -= item_sz, item = item->next) {
1057                 if (cnt < PLSIZE)
1058                         item_sz = cnt;
1059                 for (i = 0; i < item_sz; i++)
1060                         if (item->ports[i] == port)
1061                                 return;
1062                 if (i < PLSIZE) {
1063                         item->ports[i] = port;
1064                         pl_ptr->count++;
1065                         return;
1066                 }
1067                 if (!item->next) {
1068                         item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
1069                         if (!item->next) {
1070                                 pr_warn("Incomplete multicast delivery, no memory\n");
1071                                 return;
1072                         }
1073                         item->next->next = NULL;
1074                 }
1075         }
1076 }
1077
1078 /**
1079  * tipc_port_list_free - free dynamically created entries in port_list chain
1080  *
1081  */
1082 void tipc_port_list_free(struct tipc_port_list *pl_ptr)
1083 {
1084         struct tipc_port_list *item;
1085         struct tipc_port_list *next;
1086
1087         for (item = pl_ptr->next; item; item = next) {
1088                 next = item->next;
1089                 kfree(item);
1090         }
1091 }