net/tipc/bcast.c

   1 /*
   2  * net/tipc/bcast.c: TIPC broadcast code
   3  *
   4  * Copyright (c) 2004-2006, 2014-2015, 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 "socket.h"
  39 #include "msg.h"
  40 #include "bcast.h"
  41 #include "name_distr.h"
  42 #include "core.h"
  43
  44 #define MAX_PKT_DEFAULT_MCAST   1500    /* bcast link max packet size (fixed) */
  45 #define BCLINK_WIN_DEFAULT      20      /* bcast link window size (default) */
  46
  47 const char tipc_bclink_name[] = "broadcast-link";
  48
  49 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
  50                            struct tipc_node_map *nm_b,
  51                            struct tipc_node_map *nm_diff);
  52 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node);
  53 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node);
  54
  55 static void tipc_bclink_lock(struct net *net)
  56 {
  57         struct tipc_net *tn = net_generic(net, tipc_net_id);
  58
  59         spin_lock_bh(&tn->bclink->lock);
  60 }
  61
  62 static void tipc_bclink_unlock(struct net *net)
  63 {
  64         struct tipc_net *tn = net_generic(net, tipc_net_id);
  65         struct tipc_node *node = NULL;
  66
  67         if (likely(!tn->bclink->flags)) {
  68                 spin_unlock_bh(&tn->bclink->lock);
  69                 return;
  70         }
  71
  72         if (tn->bclink->flags & TIPC_BCLINK_RESET) {
  73                 tn->bclink->flags &= ~TIPC_BCLINK_RESET;
  74                 node = tipc_bclink_retransmit_to(net);
  75         }
  76         spin_unlock_bh(&tn->bclink->lock);
  77
  78         if (node)
  79                 tipc_link_reset_all(node);
  80 }
  81
  82 void tipc_bclink_input(struct net *net)
  83 {
  84         struct tipc_net *tn = net_generic(net, tipc_net_id);
  85
  86         tipc_sk_mcast_rcv(net, &tn->bclink->arrvq, &tn->bclink->inputq);
  87 }
  88
  89 uint  tipc_bclink_get_mtu(void)
  90 {
  91         return MAX_PKT_DEFAULT_MCAST;
  92 }
  93
  94 void tipc_bclink_set_flags(struct net *net, unsigned int flags)
  95 {
  96         struct tipc_net *tn = net_generic(net, tipc_net_id);
  97
  98         tn->bclink->flags |= flags;
  99 }
 100
 101 static u32 bcbuf_acks(struct sk_buff *buf)
 102 {
 103         return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
 104 }
 105
 106 static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
 107 {
 108         TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
 109 }
 110
 111 static void bcbuf_decr_acks(struct sk_buff *buf)
 112 {
 113         bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
 114 }
 115
 116 void tipc_bclink_add_node(struct net *net, u32 addr)
 117 {
 118         struct tipc_net *tn = net_generic(net, tipc_net_id);
 119
 120         tipc_bclink_lock(net);
 121         tipc_nmap_add(&tn->bclink->bcast_nodes, addr);
 122         tipc_bclink_unlock(net);
 123 }
 124
 125 void tipc_bclink_remove_node(struct net *net, u32 addr)
 126 {
 127         struct tipc_net *tn = net_generic(net, tipc_net_id);
 128
 129         tipc_bclink_lock(net);
 130         tipc_nmap_remove(&tn->bclink->bcast_nodes, addr);
 131         tipc_bclink_unlock(net);
 132 }
 133
 134 static void bclink_set_last_sent(struct net *net)
 135 {
 136         struct tipc_net *tn = net_generic(net, tipc_net_id);
 137         struct tipc_link *bcl = tn->bcl;
 138
 139         if (bcl->next_out)
 140                 bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
 141         else
 142                 bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
 143 }
 144
 145 u32 tipc_bclink_get_last_sent(struct net *net)
 146 {
 147         struct tipc_net *tn = net_generic(net, tipc_net_id);
 148
 149         return tn->bcl->fsm_msg_cnt;
 150 }
 151
 152 static void bclink_update_last_sent(struct tipc_node *node, u32 seqno)
 153 {
 154         node->bclink.last_sent = less_eq(node->bclink.last_sent, seqno) ?
 155                                                 seqno : node->bclink.last_sent;
 156 }
 157
 158
 159 /**
 160  * tipc_bclink_retransmit_to - get most recent node to request retransmission
 161  *
 162  * Called with bclink_lock locked
 163  */
 164 struct tipc_node *tipc_bclink_retransmit_to(struct net *net)
 165 {
 166         struct tipc_net *tn = net_generic(net, tipc_net_id);
 167
 168         return tn->bclink->retransmit_to;
 169 }
 170
 171 /**
 172  * bclink_retransmit_pkt - retransmit broadcast packets
 173  * @after: sequence number of last packet to *not* retransmit
 174  * @to: sequence number of last packet to retransmit
 175  *
 176  * Called with bclink_lock locked
 177  */
 178 static void bclink_retransmit_pkt(struct tipc_net *tn, u32 after, u32 to)
 179 {
 180         struct sk_buff *skb;
 181         struct tipc_link *bcl = tn->bcl;
 182
 183         skb_queue_walk(&bcl->outqueue, skb) {
 184                 if (more(buf_seqno(skb), after)) {
 185                         tipc_link_retransmit(bcl, skb, mod(to - after));
 186                         break;
 187                 }
 188         }
 189 }
 190
 191 /**
 192  * tipc_bclink_wakeup_users - wake up pending users
 193  *
 194  * Called with no locks taken
 195  */
 196 void tipc_bclink_wakeup_users(struct net *net)
 197 {
 198         struct tipc_net *tn = net_generic(net, tipc_net_id);
 199
 200         tipc_sk_rcv(net, &tn->bclink->link.wakeupq);
 201 }
 202
 203 /**
 204  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
 205  * @n_ptr: node that sent acknowledgement info
 206  * @acked: broadcast sequence # that has been acknowledged
 207  *
 208  * Node is locked, bclink_lock unlocked.
 209  */
 210 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
 211 {
 212         struct sk_buff *skb, *tmp;
 213         struct sk_buff *next;
 214         unsigned int released = 0;
 215         struct net *net = n_ptr->net;
 216         struct tipc_net *tn = net_generic(net, tipc_net_id);
 217
 218         if (unlikely(!n_ptr->bclink.recv_permitted))
 219                 return;
 220
 221         tipc_bclink_lock(net);
 222
 223         /* Bail out if tx queue is empty (no clean up is required) */
 224         skb = skb_peek(&tn->bcl->outqueue);
 225         if (!skb)
 226                 goto exit;
 227
 228         /* Determine which messages need to be acknowledged */
 229         if (acked == INVALID_LINK_SEQ) {
 230                 /*
 231                  * Contact with specified node has been lost, so need to
 232                  * acknowledge sent messages only (if other nodes still exist)
 233                  * or both sent and unsent messages (otherwise)
 234                  */
 235                 if (tn->bclink->bcast_nodes.count)
 236                         acked = tn->bcl->fsm_msg_cnt;
 237                 else
 238                         acked = tn->bcl->next_out_no;
 239         } else {
 240                 /*
 241                  * Bail out if specified sequence number does not correspond
 242                  * to a message that has been sent and not yet acknowledged
 243                  */
 244                 if (less(acked, buf_seqno(skb)) ||
 245                     less(tn->bcl->fsm_msg_cnt, acked) ||
 246                     less_eq(acked, n_ptr->bclink.acked))
 247                         goto exit;
 248         }
 249
 250         /* Skip over packets that node has previously acknowledged */
 251         skb_queue_walk(&tn->bcl->outqueue, skb) {
 252                 if (more(buf_seqno(skb), n_ptr->bclink.acked))
 253                         break;
 254         }
 255
 256         /* Update packets that node is now acknowledging */
 257         skb_queue_walk_from_safe(&tn->bcl->outqueue, skb, tmp) {
 258                 if (more(buf_seqno(skb), acked))
 259                         break;
 260
 261                 next = tipc_skb_queue_next(&tn->bcl->outqueue, skb);
 262                 if (skb != tn->bcl->next_out) {
 263                         bcbuf_decr_acks(skb);
 264                 } else {
 265                         bcbuf_set_acks(skb, 0);
 266                         tn->bcl->next_out = next;
 267                         bclink_set_last_sent(net);
 268                 }
 269
 270                 if (bcbuf_acks(skb) == 0) {
 271                         __skb_unlink(skb, &tn->bcl->outqueue);
 272                         kfree_skb(skb);
 273                         released = 1;
 274                 }
 275         }
 276         n_ptr->bclink.acked = acked;
 277
 278         /* Try resolving broadcast link congestion, if necessary */
 279         if (unlikely(tn->bcl->next_out)) {
 280                 tipc_link_push_packets(tn->bcl);
 281                 bclink_set_last_sent(net);
 282         }
 283         if (unlikely(released && !skb_queue_empty(&tn->bcl->wakeupq)))
 284                 n_ptr->action_flags |= TIPC_WAKEUP_BCAST_USERS;
 285 exit:
 286         tipc_bclink_unlock(net);
 287 }
 288
 289 /**
 290  * tipc_bclink_update_link_state - update broadcast link state
 291  *
 292  * RCU and node lock set
 293  */
 294 void tipc_bclink_update_link_state(struct tipc_node *n_ptr,
 295                                    u32 last_sent)
 296 {
 297         struct sk_buff *buf;
 298         struct net *net = n_ptr->net;
 299         struct tipc_net *tn = net_generic(net, tipc_net_id);
 300
 301         /* Ignore "stale" link state info */
 302         if (less_eq(last_sent, n_ptr->bclink.last_in))
 303                 return;
 304
 305         /* Update link synchronization state; quit if in sync */
 306         bclink_update_last_sent(n_ptr, last_sent);
 307
 308         if (n_ptr->bclink.last_sent == n_ptr->bclink.last_in)
 309                 return;
 310
 311         /* Update out-of-sync state; quit if loss is still unconfirmed */
 312         if ((++n_ptr->bclink.oos_state) == 1) {
 313                 if (n_ptr->bclink.deferred_size < (TIPC_MIN_LINK_WIN / 2))
 314                         return;
 315                 n_ptr->bclink.oos_state++;
 316         }
 317
 318         /* Don't NACK if one has been recently sent (or seen) */
 319         if (n_ptr->bclink.oos_state & 0x1)
 320                 return;
 321
 322         /* Send NACK */
 323         buf = tipc_buf_acquire(INT_H_SIZE);
 324         if (buf) {
 325                 struct tipc_msg *msg = buf_msg(buf);
 326                 struct sk_buff *skb = skb_peek(&n_ptr->bclink.deferred_queue);
 327                 u32 to = skb ? buf_seqno(skb) - 1 : n_ptr->bclink.last_sent;
 328
 329                 tipc_msg_init(tn->own_addr, msg, BCAST_PROTOCOL, STATE_MSG,
 330                               INT_H_SIZE, n_ptr->addr);
 331                 msg_set_non_seq(msg, 1);
 332                 msg_set_mc_netid(msg, tn->net_id);
 333                 msg_set_bcast_ack(msg, n_ptr->bclink.last_in);
 334                 msg_set_bcgap_after(msg, n_ptr->bclink.last_in);
 335                 msg_set_bcgap_to(msg, to);
 336
 337                 tipc_bclink_lock(net);
 338                 tipc_bearer_send(net, MAX_BEARERS, buf, NULL);
 339                 tn->bcl->stats.sent_nacks++;
 340                 tipc_bclink_unlock(net);
 341                 kfree_skb(buf);
 342
 343                 n_ptr->bclink.oos_state++;
 344         }
 345 }
 346
 347 /**
 348  * bclink_peek_nack - monitor retransmission requests sent by other nodes
 349  *
 350  * Delay any upcoming NACK by this node if another node has already
 351  * requested the first message this node is going to ask for.
 352  */
 353 static void bclink_peek_nack(struct net *net, struct tipc_msg *msg)
 354 {
 355         struct tipc_node *n_ptr = tipc_node_find(net, msg_destnode(msg));
 356
 357         if (unlikely(!n_ptr))
 358                 return;
 359
 360         tipc_node_lock(n_ptr);
 361
 362         if (n_ptr->bclink.recv_permitted &&
 363             (n_ptr->bclink.last_in != n_ptr->bclink.last_sent) &&
 364             (n_ptr->bclink.last_in == msg_bcgap_after(msg)))
 365                 n_ptr->bclink.oos_state = 2;
 366
 367         tipc_node_unlock(n_ptr);
 368 }
 369
 370 /* tipc_bclink_xmit - deliver buffer chain to all nodes in cluster
 371  *                    and to identified node local sockets
 372  * @net: the applicable net namespace
 373  * @list: chain of buffers containing message
 374  * Consumes the buffer chain, except when returning -ELINKCONG
 375  * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
 376  */
 377 int tipc_bclink_xmit(struct net *net, struct sk_buff_head *list)
 378 {
 379         struct tipc_net *tn = net_generic(net, tipc_net_id);
 380         struct tipc_link *bcl = tn->bcl;
 381         struct tipc_bclink *bclink = tn->bclink;
 382         int rc = 0;
 383         int bc = 0;
 384         struct sk_buff *skb;
 385         struct sk_buff_head arrvq;
 386         struct sk_buff_head inputq;
 387
 388         /* Prepare clone of message for local node */
 389         skb = tipc_msg_reassemble(list);
 390         if (unlikely(!skb)) {
 391                 __skb_queue_purge(list);
 392                 return -EHOSTUNREACH;
 393         }
 394
 395         /* Broadcast to all nodes */
 396         if (likely(bclink)) {
 397                 tipc_bclink_lock(net);
 398                 if (likely(bclink->bcast_nodes.count)) {
 399                         rc = __tipc_link_xmit(net, bcl, list);
 400                         if (likely(!rc)) {
 401                                 u32 len = skb_queue_len(&bcl->outqueue);
 402
 403                                 bclink_set_last_sent(net);
 404                                 bcl->stats.queue_sz_counts++;
 405                                 bcl->stats.accu_queue_sz += len;
 406                         }
 407                         bc = 1;
 408                 }
 409                 tipc_bclink_unlock(net);
 410         }
 411
 412         if (unlikely(!bc))
 413                 __skb_queue_purge(list);
 414
 415         if (unlikely(rc)) {
 416                 kfree_skb(skb);
 417                 return rc;
 418         }
 419         /* Deliver message clone */
 420         __skb_queue_head_init(&arrvq);
 421         skb_queue_head_init(&inputq);
 422         __skb_queue_tail(&arrvq, skb);
 423         tipc_sk_mcast_rcv(net, &arrvq, &inputq);
 424         return rc;
 425 }
 426
 427 /**
 428  * bclink_accept_pkt - accept an incoming, in-sequence broadcast packet
 429  *
 430  * Called with both sending node's lock and bclink_lock taken.
 431  */
 432 static void bclink_accept_pkt(struct tipc_node *node, u32 seqno)
 433 {
 434         struct tipc_net *tn = net_generic(node->net, tipc_net_id);
 435
 436         bclink_update_last_sent(node, seqno);
 437         node->bclink.last_in = seqno;
 438         node->bclink.oos_state = 0;
 439         tn->bcl->stats.recv_info++;
 440
 441         /*
 442          * Unicast an ACK periodically, ensuring that
 443          * all nodes in the cluster don't ACK at the same time
 444          */
 445         if (((seqno - tn->own_addr) % TIPC_MIN_LINK_WIN) == 0) {
 446                 tipc_link_proto_xmit(node->active_links[node->addr & 1],
 447                                      STATE_MSG, 0, 0, 0, 0, 0);
 448                 tn->bcl->stats.sent_acks++;
 449         }
 450 }
 451
 452 /**
 453  * tipc_bclink_rcv - receive a broadcast packet, and deliver upwards
 454  *
 455  * RCU is locked, no other locks set
 456  */
 457 void tipc_bclink_rcv(struct net *net, struct sk_buff *buf)
 458 {
 459         struct tipc_net *tn = net_generic(net, tipc_net_id);
 460         struct tipc_link *bcl = tn->bcl;
 461         struct tipc_msg *msg = buf_msg(buf);
 462         struct tipc_node *node;
 463         u32 next_in;
 464         u32 seqno;
 465         int deferred = 0;
 466         int pos = 0;
 467         struct sk_buff *iskb;
 468         struct sk_buff_head *arrvq, *inputq;
 469
 470         /* Screen out unwanted broadcast messages */
 471         if (msg_mc_netid(msg) != tn->net_id)
 472                 goto exit;
 473
 474         node = tipc_node_find(net, msg_prevnode(msg));
 475         if (unlikely(!node))
 476                 goto exit;
 477
 478         tipc_node_lock(node);
 479         if (unlikely(!node->bclink.recv_permitted))
 480                 goto unlock;
 481
 482         /* Handle broadcast protocol message */
 483         if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
 484                 if (msg_type(msg) != STATE_MSG)
 485                         goto unlock;
 486                 if (msg_destnode(msg) == tn->own_addr) {
 487                         tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
 488                         tipc_node_unlock(node);
 489                         tipc_bclink_lock(net);
 490                         bcl->stats.recv_nacks++;
 491                         tn->bclink->retransmit_to = node;
 492                         bclink_retransmit_pkt(tn, msg_bcgap_after(msg),
 493                                               msg_bcgap_to(msg));
 494                         tipc_bclink_unlock(net);
 495                 } else {
 496                         tipc_node_unlock(node);
 497                         bclink_peek_nack(net, msg);
 498                 }
 499                 goto exit;
 500         }
 501
 502         /* Handle in-sequence broadcast message */
 503         seqno = msg_seqno(msg);
 504         next_in = mod(node->bclink.last_in + 1);
 505         arrvq = &tn->bclink->arrvq;
 506         inputq = &tn->bclink->inputq;
 507
 508         if (likely(seqno == next_in)) {
 509 receive:
 510                 /* Deliver message to destination */
 511                 if (likely(msg_isdata(msg))) {
 512                         tipc_bclink_lock(net);
 513                         bclink_accept_pkt(node, seqno);
 514                         spin_lock_bh(&inputq->lock);
 515                         __skb_queue_tail(arrvq, buf);
 516                         spin_unlock_bh(&inputq->lock);
 517                         node->action_flags |= TIPC_BCAST_MSG_EVT;
 518                         tipc_bclink_unlock(net);
 519                         tipc_node_unlock(node);
 520                 } else if (msg_user(msg) == MSG_BUNDLER) {
 521                         tipc_bclink_lock(net);
 522                         bclink_accept_pkt(node, seqno);
 523                         bcl->stats.recv_bundles++;
 524                         bcl->stats.recv_bundled += msg_msgcnt(msg);
 525                         pos = 0;
 526                         while (tipc_msg_extract(buf, &iskb, &pos)) {
 527                                 spin_lock_bh(&inputq->lock);
 528                                 __skb_queue_tail(arrvq, iskb);
 529                                 spin_unlock_bh(&inputq->lock);
 530                         }
 531                         node->action_flags |= TIPC_BCAST_MSG_EVT;
 532                         tipc_bclink_unlock(net);
 533                         tipc_node_unlock(node);
 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(net);
 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(net);
 545                                 goto receive;
 546                         }
 547                         tipc_bclink_unlock(net);
 548                         tipc_node_unlock(node);
 549                 } else {
 550                         tipc_bclink_lock(net);
 551                         bclink_accept_pkt(node, seqno);
 552                         tipc_bclink_unlock(net);
 553                         tipc_node_unlock(node);
 554                         kfree_skb(buf);
 555                 }
 556                 buf = NULL;
 557
 558                 /* Determine new synchronization state */
 559                 tipc_node_lock(node);
 560                 if (unlikely(!tipc_node_is_up(node)))
 561                         goto unlock;
 562
 563                 if (node->bclink.last_in == node->bclink.last_sent)
 564                         goto unlock;
 565
 566                 if (skb_queue_empty(&node->bclink.deferred_queue)) {
 567                         node->bclink.oos_state = 1;
 568                         goto unlock;
 569                 }
 570
 571                 msg = buf_msg(skb_peek(&node->bclink.deferred_queue));
 572                 seqno = msg_seqno(msg);
 573                 next_in = mod(next_in + 1);
 574                 if (seqno != next_in)
 575                         goto unlock;
 576
 577                 /* Take in-sequence message from deferred queue & deliver it */
 578                 buf = __skb_dequeue(&node->bclink.deferred_queue);
 579                 goto receive;
 580         }
 581
 582         /* Handle out-of-sequence broadcast message */
 583         if (less(next_in, seqno)) {
 584                 deferred = tipc_link_defer_pkt(&node->bclink.deferred_queue,
 585                                                buf);
 586                 bclink_update_last_sent(node, seqno);
 587                 buf = NULL;
 588         }
 589
 590         tipc_bclink_lock(net);
 591
 592         if (deferred)
 593                 bcl->stats.deferred_recv++;
 594         else
 595                 bcl->stats.duplicates++;
 596
 597         tipc_bclink_unlock(net);
 598
 599 unlock:
 600         tipc_node_unlock(node);
 601 exit:
 602         kfree_skb(buf);
 603 }
 604
 605 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
 606 {
 607         return (n_ptr->bclink.recv_permitted &&
 608                 (tipc_bclink_get_last_sent(n_ptr->net) != n_ptr->bclink.acked));
 609 }
 610
 611
 612 /**
 613  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
 614  *
 615  * Send packet over as many bearers as necessary to reach all nodes
 616  * that have joined the broadcast link.
 617  *
 618  * Returns 0 (packet sent successfully) under all circumstances,
 619  * since the broadcast link's pseudo-bearer never blocks
 620  */
 621 static int tipc_bcbearer_send(struct net *net, struct sk_buff *buf,
 622                               struct tipc_bearer *unused1,
 623                               struct tipc_media_addr *unused2)
 624 {
 625         int bp_index;
 626         struct tipc_msg *msg = buf_msg(buf);
 627         struct tipc_net *tn = net_generic(net, tipc_net_id);
 628         struct tipc_bcbearer *bcbearer = tn->bcbearer;
 629         struct tipc_bclink *bclink = tn->bclink;
 630
 631         /* Prepare broadcast link message for reliable transmission,
 632          * if first time trying to send it;
 633          * preparation is skipped for broadcast link protocol messages
 634          * since they are sent in an unreliable manner and don't need it
 635          */
 636         if (likely(!msg_non_seq(buf_msg(buf)))) {
 637                 bcbuf_set_acks(buf, bclink->bcast_nodes.count);
 638                 msg_set_non_seq(msg, 1);
 639                 msg_set_mc_netid(msg, tn->net_id);
 640                 tn->bcl->stats.sent_info++;
 641
 642                 if (WARN_ON(!bclink->bcast_nodes.count)) {
 643                         dump_stack();
 644                         return 0;
 645                 }
 646         }
 647
 648         /* Send buffer over bearers until all targets reached */
 649         bcbearer->remains = bclink->bcast_nodes;
 650
 651         for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
 652                 struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
 653                 struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
 654                 struct tipc_bearer *bp[2] = {p, s};
 655                 struct tipc_bearer *b = bp[msg_link_selector(msg)];
 656                 struct sk_buff *tbuf;
 657
 658                 if (!p)
 659                         break; /* No more bearers to try */
 660                 if (!b)
 661                         b = p;
 662                 tipc_nmap_diff(&bcbearer->remains, &b->nodes,
 663                                &bcbearer->remains_new);
 664                 if (bcbearer->remains_new.count == bcbearer->remains.count)
 665                         continue; /* Nothing added by bearer pair */
 666
 667                 if (bp_index == 0) {
 668                         /* Use original buffer for first bearer */
 669                         tipc_bearer_send(net, b->identity, buf, &b->bcast_addr);
 670                 } else {
 671                         /* Avoid concurrent buffer access */
 672                         tbuf = pskb_copy_for_clone(buf, GFP_ATOMIC);
 673                         if (!tbuf)
 674                                 break;
 675                         tipc_bearer_send(net, b->identity, tbuf,
 676                                          &b->bcast_addr);
 677                         kfree_skb(tbuf); /* Bearer keeps a clone */
 678                 }
 679                 if (bcbearer->remains_new.count == 0)
 680                         break; /* All targets reached */
 681
 682                 bcbearer->remains = bcbearer->remains_new;
 683         }
 684
 685         return 0;
 686 }
 687
 688 /**
 689  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
 690  */
 691 void tipc_bcbearer_sort(struct net *net, struct tipc_node_map *nm_ptr,
 692                         u32 node, bool action)
 693 {
 694         struct tipc_net *tn = net_generic(net, tipc_net_id);
 695         struct tipc_bcbearer *bcbearer = tn->bcbearer;
 696         struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
 697         struct tipc_bcbearer_pair *bp_curr;
 698         struct tipc_bearer *b;
 699         int b_index;
 700         int pri;
 701
 702         tipc_bclink_lock(net);
 703
 704         if (action)
 705                 tipc_nmap_add(nm_ptr, node);
 706         else
 707                 tipc_nmap_remove(nm_ptr, node);
 708
 709         /* Group bearers by priority (can assume max of two per priority) */
 710         memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
 711
 712         rcu_read_lock();
 713         for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
 714                 b = rcu_dereference_rtnl(tn->bearer_list[b_index]);
 715                 if (!b || !b->nodes.count)
 716                         continue;
 717
 718                 if (!bp_temp[b->priority].primary)
 719                         bp_temp[b->priority].primary = b;
 720                 else
 721                         bp_temp[b->priority].secondary = b;
 722         }
 723         rcu_read_unlock();
 724
 725         /* Create array of bearer pairs for broadcasting */
 726         bp_curr = bcbearer->bpairs;
 727         memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
 728
 729         for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
 730
 731                 if (!bp_temp[pri].primary)
 732                         continue;
 733
 734                 bp_curr->primary = bp_temp[pri].primary;
 735
 736                 if (bp_temp[pri].secondary) {
 737                         if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
 738                                             &bp_temp[pri].secondary->nodes)) {
 739                                 bp_curr->secondary = bp_temp[pri].secondary;
 740                         } else {
 741                                 bp_curr++;
 742                                 bp_curr->primary = bp_temp[pri].secondary;
 743                         }
 744                 }
 745
 746                 bp_curr++;
 747         }
 748
 749         tipc_bclink_unlock(net);
 750 }
 751
 752 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
 753                                       struct tipc_stats *stats)
 754 {
 755         int i;
 756         struct nlattr *nest;
 757
 758         struct nla_map {
 759                 __u32 key;
 760                 __u32 val;
 761         };
 762
 763         struct nla_map map[] = {
 764                 {TIPC_NLA_STATS_RX_INFO, stats->recv_info},
 765                 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
 766                 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
 767                 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
 768                 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
 769                 {TIPC_NLA_STATS_TX_INFO, stats->sent_info},
 770                 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
 771                 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
 772                 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
 773                 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
 774                 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
 775                 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
 776                 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
 777                 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
 778                 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
 779                 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
 780                 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
 781                 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
 782                 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
 783                         (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
 784         };
 785
 786         nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
 787         if (!nest)
 788                 return -EMSGSIZE;
 789
 790         for (i = 0; i <  ARRAY_SIZE(map); i++)
 791                 if (nla_put_u32(skb, map[i].key, map[i].val))
 792                         goto msg_full;
 793
 794         nla_nest_end(skb, nest);
 795
 796         return 0;
 797 msg_full:
 798         nla_nest_cancel(skb, nest);
 799
 800         return -EMSGSIZE;
 801 }
 802
 803 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
 804 {
 805         int err;
 806         void *hdr;
 807         struct nlattr *attrs;
 808         struct nlattr *prop;
 809         struct tipc_net *tn = net_generic(net, tipc_net_id);
 810         struct tipc_link *bcl = tn->bcl;
 811
 812         if (!bcl)
 813                 return 0;
 814
 815         tipc_bclink_lock(net);
 816
 817         hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
 818                           NLM_F_MULTI, TIPC_NL_LINK_GET);
 819         if (!hdr)
 820                 return -EMSGSIZE;
 821
 822         attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
 823         if (!attrs)
 824                 goto msg_full;
 825
 826         /* The broadcast link is always up */
 827         if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
 828                 goto attr_msg_full;
 829
 830         if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
 831                 goto attr_msg_full;
 832         if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
 833                 goto attr_msg_full;
 834         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, bcl->next_in_no))
 835                 goto attr_msg_full;
 836         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, bcl->next_out_no))
 837                 goto attr_msg_full;
 838
 839         prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
 840         if (!prop)
 841                 goto attr_msg_full;
 842         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->queue_limit[0]))
 843                 goto prop_msg_full;
 844         nla_nest_end(msg->skb, prop);
 845
 846         err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
 847         if (err)
 848                 goto attr_msg_full;
 849
 850         tipc_bclink_unlock(net);
 851         nla_nest_end(msg->skb, attrs);
 852         genlmsg_end(msg->skb, hdr);
 853
 854         return 0;
 855
 856 prop_msg_full:
 857         nla_nest_cancel(msg->skb, prop);
 858 attr_msg_full:
 859         nla_nest_cancel(msg->skb, attrs);
 860 msg_full:
 861         tipc_bclink_unlock(net);
 862         genlmsg_cancel(msg->skb, hdr);
 863
 864         return -EMSGSIZE;
 865 }
 866
 867 int tipc_bclink_reset_stats(struct net *net)
 868 {
 869         struct tipc_net *tn = net_generic(net, tipc_net_id);
 870         struct tipc_link *bcl = tn->bcl;
 871
 872         if (!bcl)
 873                 return -ENOPROTOOPT;
 874
 875         tipc_bclink_lock(net);
 876         memset(&bcl->stats, 0, sizeof(bcl->stats));
 877         tipc_bclink_unlock(net);
 878         return 0;
 879 }
 880
 881 int tipc_bclink_set_queue_limits(struct net *net, u32 limit)
 882 {
 883         struct tipc_net *tn = net_generic(net, tipc_net_id);
 884         struct tipc_link *bcl = tn->bcl;
 885
 886         if (!bcl)
 887                 return -ENOPROTOOPT;
 888         if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
 889                 return -EINVAL;
 890
 891         tipc_bclink_lock(net);
 892         tipc_link_set_queue_limits(bcl, limit);
 893         tipc_bclink_unlock(net);
 894         return 0;
 895 }
 896
 897 int tipc_bclink_init(struct net *net)
 898 {
 899         struct tipc_net *tn = net_generic(net, tipc_net_id);
 900         struct tipc_bcbearer *bcbearer;
 901         struct tipc_bclink *bclink;
 902         struct tipc_link *bcl;
 903
 904         bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
 905         if (!bcbearer)
 906                 return -ENOMEM;
 907
 908         bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
 909         if (!bclink) {
 910                 kfree(bcbearer);
 911                 return -ENOMEM;
 912         }
 913
 914         bcl = &bclink->link;
 915         bcbearer->bearer.media = &bcbearer->media;
 916         bcbearer->media.send_msg = tipc_bcbearer_send;
 917         sprintf(bcbearer->media.name, "tipc-broadcast");
 918
 919         spin_lock_init(&bclink->lock);
 920         __skb_queue_head_init(&bcl->outqueue);
 921         __skb_queue_head_init(&bcl->deferred_queue);
 922         skb_queue_head_init(&bcl->wakeupq);
 923         bcl->next_out_no = 1;
 924         spin_lock_init(&bclink->node.lock);
 925         __skb_queue_head_init(&bclink->arrvq);
 926         skb_queue_head_init(&bclink->inputq);
 927         bcl->owner = &bclink->node;
 928         bcl->owner->net = net;
 929         bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
 930         tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
 931         bcl->bearer_id = MAX_BEARERS;
 932         rcu_assign_pointer(tn->bearer_list[MAX_BEARERS], &bcbearer->bearer);
 933         bcl->state = WORKING_WORKING;
 934         bcl->pmsg = (struct tipc_msg *)&bcl->proto_msg;
 935         msg_set_prevnode(bcl->pmsg, tn->own_addr);
 936         strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
 937         tn->bcbearer = bcbearer;
 938         tn->bclink = bclink;
 939         tn->bcl = bcl;
 940         return 0;
 941 }
 942
 943 void tipc_bclink_stop(struct net *net)
 944 {
 945         struct tipc_net *tn = net_generic(net, tipc_net_id);
 946
 947         tipc_bclink_lock(net);
 948         tipc_link_purge_queues(tn->bcl);
 949         tipc_bclink_unlock(net);
 950
 951         RCU_INIT_POINTER(tn->bearer_list[BCBEARER], NULL);
 952         synchronize_net();
 953         kfree(tn->bcbearer);
 954         kfree(tn->bclink);
 955 }
 956
 957 /**
 958  * tipc_nmap_add - add a node to a node map
 959  */
 960 static void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
 961 {
 962         int n = tipc_node(node);
 963         int w = n / WSIZE;
 964         u32 mask = (1 << (n % WSIZE));
 965
 966         if ((nm_ptr->map[w] & mask) == 0) {
 967                 nm_ptr->count++;
 968                 nm_ptr->map[w] |= mask;
 969         }
 970 }
 971
 972 /**
 973  * tipc_nmap_remove - remove a node from a node map
 974  */
 975 static void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
 976 {
 977         int n = tipc_node(node);
 978         int w = n / WSIZE;
 979         u32 mask = (1 << (n % WSIZE));
 980
 981         if ((nm_ptr->map[w] & mask) != 0) {
 982                 nm_ptr->map[w] &= ~mask;
 983                 nm_ptr->count--;
 984         }
 985 }
 986
 987 /**
 988  * tipc_nmap_diff - find differences between node maps
 989  * @nm_a: input node map A
 990  * @nm_b: input node map B
 991  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
 992  */
 993 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
 994                            struct tipc_node_map *nm_b,
 995                            struct tipc_node_map *nm_diff)
 996 {
 997         int stop = ARRAY_SIZE(nm_a->map);
 998         int w;
 999         int b;
1000         u32 map;
1001
1002         memset(nm_diff, 0, sizeof(*nm_diff));
1003         for (w = 0; w < stop; w++) {
1004                 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
1005                 nm_diff->map[w] = map;
1006                 if (map != 0) {
1007                         for (b = 0 ; b < WSIZE; b++) {
1008                                 if (map & (1 << b))
1009                                         nm_diff->count++;
1010                         }
1011                 }
1012         }
1013 }