net/tipc/bcast.c

   1 /*
   2  * net/tipc/bcast.c: TIPC broadcast code
   3  *
   4  * Copyright (c) 2004-2006, Ericsson AB
   5  * Copyright (c) 2004, Intel Corporation.
   6  * Copyright (c) 2005, Wind River Systems
   7  * All rights reserved.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions are met:
  11  *
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. Neither the names of the copyright holders nor the names of its
  18  *    contributors may be used to endorse or promote products derived from
  19  *    this software without specific prior written permission.
  20  *
  21  * Alternatively, this software may be distributed under the terms of the
  22  * GNU General Public License ("GPL") version 2 as published by the Free
  23  * Software Foundation.
  24  *
  25  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  26  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  28  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  29  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  35  * POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include "core.h"
  39 #include "link.h"
  40 #include "port.h"
  41 #include "name_distr.h"
  42 #include "bcast.h"
  43
  44 #define MAX_PKT_DEFAULT_MCAST 1500      /* bcast link max packet size (fixed) */
  45
  46 #define BCLINK_WIN_DEFAULT 20           /* bcast link window size (default) */
  47
  48 #define BCLINK_LOG_BUF_SIZE 0
  49
  50 /*
  51  * Loss rate for incoming broadcast frames; used to test retransmission code.
  52  * Set to N to cause every N'th frame to be discarded; 0 => don't discard any.
  53  */
  54
  55 #define TIPC_BCAST_LOSS_RATE 0
  56
  57 /**
  58  * struct bcbearer_pair - a pair of bearers used by broadcast link
  59  * @primary: pointer to primary bearer
  60  * @secondary: pointer to secondary bearer
  61  *
  62  * Bearers must have same priority and same set of reachable destinations
  63  * to be paired.
  64  */
  65
  66 struct bcbearer_pair {
  67         struct bearer *primary;
  68         struct bearer *secondary;
  69 };
  70
  71 /**
  72  * struct bcbearer - bearer used by broadcast link
  73  * @bearer: (non-standard) broadcast bearer structure
  74  * @media: (non-standard) broadcast media structure
  75  * @bpairs: array of bearer pairs
  76  * @bpairs_temp: temporary array of bearer pairs used by tipc_bcbearer_sort()
  77  * @remains: temporary node map used by tipc_bcbearer_send()
  78  * @remains_new: temporary node map used tipc_bcbearer_send()
  79  *
  80  * Note: The fields labelled "temporary" are incorporated into the bearer
  81  * to avoid consuming potentially limited stack space through the use of
  82  * large local variables within multicast routines.  Concurrent access is
  83  * prevented through use of the spinlock "bc_lock".
  84  */
  85
  86 struct bcbearer {
  87         struct bearer bearer;
  88         struct media media;
  89         struct bcbearer_pair bpairs[MAX_BEARERS];
  90         struct bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1];
  91         struct tipc_node_map remains;
  92         struct tipc_node_map remains_new;
  93 };
  94
  95 /**
  96  * struct bclink - link used for broadcast messages
  97  * @link: (non-standard) broadcast link structure
  98  * @node: (non-standard) node structure representing b'cast link's peer node
  99  *
 100  * Handles sequence numbering, fragmentation, bundling, etc.
 101  */
 102
 103 struct bclink {
 104         struct link link;
 105         struct tipc_node node;
 106 };
 107
 108
 109 static struct bcbearer *bcbearer = NULL;
 110 static struct bclink *bclink = NULL;
 111 static struct link *bcl = NULL;
 112 static DEFINE_SPINLOCK(bc_lock);
 113
 114 /* broadcast-capable node map */
 115 struct tipc_node_map tipc_bcast_nmap;
 116
 117 const char tipc_bclink_name[] = "broadcast-link";
 118
 119 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
 120                            struct tipc_node_map *nm_b,
 121                            struct tipc_node_map *nm_diff);
 122
 123 static u32 buf_seqno(struct sk_buff *buf)
 124 {
 125         return msg_seqno(buf_msg(buf));
 126 }
 127
 128 static u32 bcbuf_acks(struct sk_buff *buf)
 129 {
 130         return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
 131 }
 132
 133 static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
 134 {
 135         TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
 136 }
 137
 138 static void bcbuf_decr_acks(struct sk_buff *buf)
 139 {
 140         bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
 141 }
 142
 143
 144 static void bclink_set_last_sent(void)
 145 {
 146         if (bcl->next_out)
 147                 bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
 148         else
 149                 bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
 150 }
 151
 152 u32 tipc_bclink_get_last_sent(void)
 153 {
 154         return bcl->fsm_msg_cnt;
 155 }
 156
 157 /**
 158  * bclink_set_gap - set gap according to contents of current deferred pkt queue
 159  *
 160  * Called with 'node' locked, bc_lock unlocked
 161  */
 162
 163 static void bclink_set_gap(struct tipc_node *n_ptr)
 164 {
 165         struct sk_buff *buf = n_ptr->bclink.deferred_head;
 166
 167         n_ptr->bclink.gap_after = n_ptr->bclink.gap_to =
 168                 mod(n_ptr->bclink.last_in);
 169         if (unlikely(buf != NULL))
 170                 n_ptr->bclink.gap_to = mod(buf_seqno(buf) - 1);
 171 }
 172
 173 /**
 174  * bclink_ack_allowed - test if ACK or NACK message can be sent at this moment
 175  *
 176  * This mechanism endeavours to prevent all nodes in network from trying
 177  * to ACK or NACK at the same time.
 178  *
 179  * Note: TIPC uses a different trigger to distribute ACKs than it does to
 180  *       distribute NACKs, but tries to use the same spacing (divide by 16).
 181  */
 182
 183 static int bclink_ack_allowed(u32 n)
 184 {
 185         return (n % TIPC_MIN_LINK_WIN) == tipc_own_tag;
 186 }
 187
 188
 189 /**
 190  * bclink_retransmit_pkt - retransmit broadcast packets
 191  * @after: sequence number of last packet to *not* retransmit
 192  * @to: sequence number of last packet to retransmit
 193  *
 194  * Called with bc_lock locked
 195  */
 196
 197 static void bclink_retransmit_pkt(u32 after, u32 to)
 198 {
 199         struct sk_buff *buf;
 200
 201         buf = bcl->first_out;
 202         while (buf && less_eq(buf_seqno(buf), after)) {
 203                 buf = buf->next;
 204         }
 205         tipc_link_retransmit(bcl, buf, mod(to - after));
 206 }
 207
 208 /**
 209  * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
 210  * @n_ptr: node that sent acknowledgement info
 211  * @acked: broadcast sequence # that has been acknowledged
 212  *
 213  * Node is locked, bc_lock unlocked.
 214  */
 215
 216 void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
 217 {
 218         struct sk_buff *crs;
 219         struct sk_buff *next;
 220         unsigned int released = 0;
 221
 222         if (less_eq(acked, n_ptr->bclink.acked))
 223                 return;
 224
 225         spin_lock_bh(&bc_lock);
 226
 227         /* Skip over packets that node has previously acknowledged */
 228
 229         crs = bcl->first_out;
 230         while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked)) {
 231                 crs = crs->next;
 232         }
 233
 234         /* Update packets that node is now acknowledging */
 235
 236         while (crs && less_eq(buf_seqno(crs), acked)) {
 237                 next = crs->next;
 238                 bcbuf_decr_acks(crs);
 239                 if (bcbuf_acks(crs) == 0) {
 240                         bcl->first_out = next;
 241                         bcl->out_queue_size--;
 242                         buf_discard(crs);
 243                         released = 1;
 244                 }
 245                 crs = next;
 246         }
 247         n_ptr->bclink.acked = acked;
 248
 249         /* Try resolving broadcast link congestion, if necessary */
 250
 251         if (unlikely(bcl->next_out)) {
 252                 tipc_link_push_queue(bcl);
 253                 bclink_set_last_sent();
 254         }
 255         if (unlikely(released && !list_empty(&bcl->waiting_ports)))
 256                 tipc_link_wakeup_ports(bcl, 0);
 257         spin_unlock_bh(&bc_lock);
 258 }
 259
 260 /**
 261  * bclink_send_ack - unicast an ACK msg
 262  *
 263  * tipc_net_lock and node lock set
 264  */
 265
 266 static void bclink_send_ack(struct tipc_node *n_ptr)
 267 {
 268         struct link *l_ptr = n_ptr->active_links[n_ptr->addr & 1];
 269
 270         if (l_ptr != NULL)
 271                 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
 272 }
 273
 274 /**
 275  * bclink_send_nack- broadcast a NACK msg
 276  *
 277  * tipc_net_lock and node lock set
 278  */
 279
 280 static void bclink_send_nack(struct tipc_node *n_ptr)
 281 {
 282         struct sk_buff *buf;
 283         struct tipc_msg *msg;
 284
 285         if (!less(n_ptr->bclink.gap_after, n_ptr->bclink.gap_to))
 286                 return;
 287
 288         buf = tipc_buf_acquire(INT_H_SIZE);
 289         if (buf) {
 290                 msg = buf_msg(buf);
 291                 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
 292                          INT_H_SIZE, n_ptr->addr);
 293                 msg_set_mc_netid(msg, tipc_net_id);
 294                 msg_set_bcast_ack(msg, mod(n_ptr->bclink.last_in));
 295                 msg_set_bcgap_after(msg, n_ptr->bclink.gap_after);
 296                 msg_set_bcgap_to(msg, n_ptr->bclink.gap_to);
 297                 msg_set_bcast_tag(msg, tipc_own_tag);
 298
 299                 if (tipc_bearer_send(&bcbearer->bearer, buf, NULL)) {
 300                         bcl->stats.sent_nacks++;
 301                         buf_discard(buf);
 302                 } else {
 303                         tipc_bearer_schedule(bcl->b_ptr, bcl);
 304                         bcl->proto_msg_queue = buf;
 305                         bcl->stats.bearer_congs++;
 306                 }
 307
 308                 /*
 309                  * Ensure we doesn't send another NACK msg to the node
 310                  * until 16 more deferred messages arrive from it
 311                  * (i.e. helps prevent all nodes from NACK'ing at same time)
 312                  */
 313
 314                 n_ptr->bclink.nack_sync = tipc_own_tag;
 315         }
 316 }
 317
 318 /**
 319  * tipc_bclink_check_gap - send a NACK if a sequence gap exists
 320  *
 321  * tipc_net_lock and node lock set
 322  */
 323
 324 void tipc_bclink_check_gap(struct tipc_node *n_ptr, u32 last_sent)
 325 {
 326         if (!n_ptr->bclink.supported ||
 327             less_eq(last_sent, mod(n_ptr->bclink.last_in)))
 328                 return;
 329
 330         bclink_set_gap(n_ptr);
 331         if (n_ptr->bclink.gap_after == n_ptr->bclink.gap_to)
 332                 n_ptr->bclink.gap_to = last_sent;
 333         bclink_send_nack(n_ptr);
 334 }
 335
 336 /**
 337  * tipc_bclink_peek_nack - process a NACK msg meant for another node
 338  *
 339  * Only tipc_net_lock set.
 340  */
 341
 342 static void tipc_bclink_peek_nack(u32 dest, u32 sender_tag, u32 gap_after, u32 gap_to)
 343 {
 344         struct tipc_node *n_ptr = tipc_node_find(dest);
 345         u32 my_after, my_to;
 346
 347         if (unlikely(!n_ptr || !tipc_node_is_up(n_ptr)))
 348                 return;
 349         tipc_node_lock(n_ptr);
 350         /*
 351          * Modify gap to suppress unnecessary NACKs from this node
 352          */
 353         my_after = n_ptr->bclink.gap_after;
 354         my_to = n_ptr->bclink.gap_to;
 355
 356         if (less_eq(gap_after, my_after)) {
 357                 if (less(my_after, gap_to) && less(gap_to, my_to))
 358                         n_ptr->bclink.gap_after = gap_to;
 359                 else if (less_eq(my_to, gap_to))
 360                         n_ptr->bclink.gap_to = n_ptr->bclink.gap_after;
 361         } else if (less_eq(gap_after, my_to)) {
 362                 if (less_eq(my_to, gap_to))
 363                         n_ptr->bclink.gap_to = gap_after;
 364         } else {
 365                 /*
 366                  * Expand gap if missing bufs not in deferred queue:
 367                  */
 368                 struct sk_buff *buf = n_ptr->bclink.deferred_head;
 369                 u32 prev = n_ptr->bclink.gap_to;
 370
 371                 for (; buf; buf = buf->next) {
 372                         u32 seqno = buf_seqno(buf);
 373
 374                         if (mod(seqno - prev) != 1) {
 375                                 buf = NULL;
 376                                 break;
 377                         }
 378                         if (seqno == gap_after)
 379                                 break;
 380                         prev = seqno;
 381                 }
 382                 if (buf == NULL)
 383                         n_ptr->bclink.gap_to = gap_after;
 384         }
 385         /*
 386          * Some nodes may send a complementary NACK now:
 387          */
 388         if (bclink_ack_allowed(sender_tag + 1)) {
 389                 if (n_ptr->bclink.gap_to != n_ptr->bclink.gap_after) {
 390                         bclink_send_nack(n_ptr);
 391                         bclink_set_gap(n_ptr);
 392                 }
 393         }
 394         tipc_node_unlock(n_ptr);
 395 }
 396
 397 /**
 398  * tipc_bclink_send_msg - broadcast a packet to all nodes in cluster
 399  */
 400
 401 int tipc_bclink_send_msg(struct sk_buff *buf)
 402 {
 403         int res;
 404
 405         spin_lock_bh(&bc_lock);
 406
 407         res = tipc_link_send_buf(bcl, buf);
 408         if (unlikely(res == -ELINKCONG))
 409                 buf_discard(buf);
 410         else
 411                 bclink_set_last_sent();
 412
 413         if (bcl->out_queue_size > bcl->stats.max_queue_sz)
 414                 bcl->stats.max_queue_sz = bcl->out_queue_size;
 415         bcl->stats.queue_sz_counts++;
 416         bcl->stats.accu_queue_sz += bcl->out_queue_size;
 417
 418         spin_unlock_bh(&bc_lock);
 419         return res;
 420 }
 421
 422 /**
 423  * tipc_bclink_recv_pkt - receive a broadcast packet, and deliver upwards
 424  *
 425  * tipc_net_lock is read_locked, no other locks set
 426  */
 427
 428 void tipc_bclink_recv_pkt(struct sk_buff *buf)
 429 {
 430 #if (TIPC_BCAST_LOSS_RATE)
 431         static int rx_count = 0;
 432 #endif
 433         struct tipc_msg *msg = buf_msg(buf);
 434         struct tipc_node* node = tipc_node_find(msg_prevnode(msg));
 435         u32 next_in;
 436         u32 seqno;
 437         struct sk_buff *deferred;
 438
 439         if (unlikely(!node || !tipc_node_is_up(node) || !node->bclink.supported ||
 440                      (msg_mc_netid(msg) != tipc_net_id))) {
 441                 buf_discard(buf);
 442                 return;
 443         }
 444
 445         if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
 446                 if (msg_destnode(msg) == tipc_own_addr) {
 447                         tipc_node_lock(node);
 448                         tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
 449                         tipc_node_unlock(node);
 450                         spin_lock_bh(&bc_lock);
 451                         bcl->stats.recv_nacks++;
 452                         bcl->owner->next = node;   /* remember requestor */
 453                         bclink_retransmit_pkt(msg_bcgap_after(msg),
 454                                               msg_bcgap_to(msg));
 455                         bcl->owner->next = NULL;
 456                         spin_unlock_bh(&bc_lock);
 457                 } else {
 458                         tipc_bclink_peek_nack(msg_destnode(msg),
 459                                               msg_bcast_tag(msg),
 460                                               msg_bcgap_after(msg),
 461                                               msg_bcgap_to(msg));
 462                 }
 463                 buf_discard(buf);
 464                 return;
 465         }
 466
 467 #if (TIPC_BCAST_LOSS_RATE)
 468         if (++rx_count == TIPC_BCAST_LOSS_RATE) {
 469                 rx_count = 0;
 470                 buf_discard(buf);
 471                 return;
 472         }
 473 #endif
 474
 475         tipc_node_lock(node);
 476 receive:
 477         deferred = node->bclink.deferred_head;
 478         next_in = mod(node->bclink.last_in + 1);
 479         seqno = msg_seqno(msg);
 480
 481         if (likely(seqno == next_in)) {
 482                 bcl->stats.recv_info++;
 483                 node->bclink.last_in++;
 484                 bclink_set_gap(node);
 485                 if (unlikely(bclink_ack_allowed(seqno))) {
 486                         bclink_send_ack(node);
 487                         bcl->stats.sent_acks++;
 488                 }
 489                 if (likely(msg_isdata(msg))) {
 490                         tipc_node_unlock(node);
 491                         tipc_port_recv_mcast(buf, NULL);
 492                 } else if (msg_user(msg) == MSG_BUNDLER) {
 493                         bcl->stats.recv_bundles++;
 494                         bcl->stats.recv_bundled += msg_msgcnt(msg);
 495                         tipc_node_unlock(node);
 496                         tipc_link_recv_bundle(buf);
 497                 } else if (msg_user(msg) == MSG_FRAGMENTER) {
 498                         bcl->stats.recv_fragments++;
 499                         if (tipc_link_recv_fragment(&node->bclink.defragm,
 500                                                     &buf, &msg))
 501                                 bcl->stats.recv_fragmented++;
 502                         tipc_node_unlock(node);
 503                         tipc_net_route_msg(buf);
 504                 } else {
 505                         tipc_node_unlock(node);
 506                         tipc_net_route_msg(buf);
 507                 }
 508                 if (deferred && (buf_seqno(deferred) == mod(next_in + 1))) {
 509                         tipc_node_lock(node);
 510                         buf = deferred;
 511                         msg = buf_msg(buf);
 512                         node->bclink.deferred_head = deferred->next;
 513                         goto receive;
 514                 }
 515                 return;
 516         } else if (less(next_in, seqno)) {
 517                 u32 gap_after = node->bclink.gap_after;
 518                 u32 gap_to = node->bclink.gap_to;
 519
 520                 if (tipc_link_defer_pkt(&node->bclink.deferred_head,
 521                                         &node->bclink.deferred_tail,
 522                                         buf)) {
 523                         node->bclink.nack_sync++;
 524                         bcl->stats.deferred_recv++;
 525                         if (seqno == mod(gap_after + 1))
 526                                 node->bclink.gap_after = seqno;
 527                         else if (less(gap_after, seqno) && less(seqno, gap_to))
 528                                 node->bclink.gap_to = seqno;
 529                 }
 530                 if (bclink_ack_allowed(node->bclink.nack_sync)) {
 531                         if (gap_to != gap_after)
 532                                 bclink_send_nack(node);
 533                         bclink_set_gap(node);
 534                 }
 535         } else {
 536                 bcl->stats.duplicates++;
 537                 buf_discard(buf);
 538         }
 539         tipc_node_unlock(node);
 540 }
 541
 542 u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
 543 {
 544         return (n_ptr->bclink.supported &&
 545                 (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
 546 }
 547
 548
 549 /**
 550  * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
 551  *
 552  * Send through as many bearers as necessary to reach all nodes
 553  * that support TIPC multicasting.
 554  *
 555  * Returns 0 if packet sent successfully, non-zero if not
 556  */
 557
 558 static int tipc_bcbearer_send(struct sk_buff *buf,
 559                               struct tipc_bearer *unused1,
 560                               struct tipc_media_addr *unused2)
 561 {
 562         int bp_index;
 563
 564         /* Prepare buffer for broadcasting (if first time trying to send it) */
 565
 566         if (likely(!msg_non_seq(buf_msg(buf)))) {
 567                 struct tipc_msg *msg;
 568
 569                 assert(tipc_bcast_nmap.count != 0);
 570                 bcbuf_set_acks(buf, tipc_bcast_nmap.count);
 571                 msg = buf_msg(buf);
 572                 msg_set_non_seq(msg, 1);
 573                 msg_set_mc_netid(msg, tipc_net_id);
 574                 bcl->stats.sent_info++;
 575         }
 576
 577         /* Send buffer over bearers until all targets reached */
 578
 579         bcbearer->remains = tipc_bcast_nmap;
 580
 581         for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
 582                 struct bearer *p = bcbearer->bpairs[bp_index].primary;
 583                 struct bearer *s = bcbearer->bpairs[bp_index].secondary;
 584
 585                 if (!p)
 586                         break;  /* no more bearers to try */
 587
 588                 tipc_nmap_diff(&bcbearer->remains, &p->nodes, &bcbearer->remains_new);
 589                 if (bcbearer->remains_new.count == bcbearer->remains.count)
 590                         continue;       /* bearer pair doesn't add anything */
 591
 592                 if (p->publ.blocked ||
 593                     p->media->send_msg(buf, &p->publ, &p->media->bcast_addr)) {
 594                         /* unable to send on primary bearer */
 595                         if (!s || s->publ.blocked ||
 596                             s->media->send_msg(buf, &s->publ,
 597                                                &s->media->bcast_addr)) {
 598                                 /* unable to send on either bearer */
 599                                 continue;
 600                         }
 601                 }
 602
 603                 if (s) {
 604                         bcbearer->bpairs[bp_index].primary = s;
 605                         bcbearer->bpairs[bp_index].secondary = p;
 606                 }
 607
 608                 if (bcbearer->remains_new.count == 0)
 609                         return 0;
 610
 611                 bcbearer->remains = bcbearer->remains_new;
 612         }
 613
 614         /*
 615          * Unable to reach all targets (indicate success, since currently
 616          * there isn't code in place to properly block & unblock the
 617          * pseudo-bearer used by the broadcast link)
 618          */
 619
 620         return TIPC_OK;
 621 }
 622
 623 /**
 624  * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
 625  */
 626
 627 void tipc_bcbearer_sort(void)
 628 {
 629         struct bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
 630         struct bcbearer_pair *bp_curr;
 631         int b_index;
 632         int pri;
 633
 634         spin_lock_bh(&bc_lock);
 635
 636         /* Group bearers by priority (can assume max of two per priority) */
 637
 638         memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
 639
 640         for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
 641                 struct bearer *b = &tipc_bearers[b_index];
 642
 643                 if (!b->active || !b->nodes.count)
 644                         continue;
 645
 646                 if (!bp_temp[b->priority].primary)
 647                         bp_temp[b->priority].primary = b;
 648                 else
 649                         bp_temp[b->priority].secondary = b;
 650         }
 651
 652         /* Create array of bearer pairs for broadcasting */
 653
 654         bp_curr = bcbearer->bpairs;
 655         memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
 656
 657         for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
 658
 659                 if (!bp_temp[pri].primary)
 660                         continue;
 661
 662                 bp_curr->primary = bp_temp[pri].primary;
 663
 664                 if (bp_temp[pri].secondary) {
 665                         if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
 666                                             &bp_temp[pri].secondary->nodes)) {
 667                                 bp_curr->secondary = bp_temp[pri].secondary;
 668                         } else {
 669                                 bp_curr++;
 670                                 bp_curr->primary = bp_temp[pri].secondary;
 671                         }
 672                 }
 673
 674                 bp_curr++;
 675         }
 676
 677         spin_unlock_bh(&bc_lock);
 678 }
 679
 680 /**
 681  * tipc_bcbearer_push - resolve bearer congestion
 682  *
 683  * Forces bclink to push out any unsent packets, until all packets are gone
 684  * or congestion reoccurs.
 685  * No locks set when function called
 686  */
 687
 688 void tipc_bcbearer_push(void)
 689 {
 690         struct bearer *b_ptr;
 691
 692         spin_lock_bh(&bc_lock);
 693         b_ptr = &bcbearer->bearer;
 694         if (b_ptr->publ.blocked) {
 695                 b_ptr->publ.blocked = 0;
 696                 tipc_bearer_lock_push(b_ptr);
 697         }
 698         spin_unlock_bh(&bc_lock);
 699 }
 700
 701
 702 int tipc_bclink_stats(char *buf, const u32 buf_size)
 703 {
 704         struct print_buf pb;
 705
 706         if (!bcl)
 707                 return 0;
 708
 709         tipc_printbuf_init(&pb, buf, buf_size);
 710
 711         spin_lock_bh(&bc_lock);
 712
 713         tipc_printf(&pb, "Link <%s>\n"
 714                          "  Window:%u packets\n",
 715                     bcl->name, bcl->queue_limit[0]);
 716         tipc_printf(&pb, "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
 717                     bcl->stats.recv_info,
 718                     bcl->stats.recv_fragments,
 719                     bcl->stats.recv_fragmented,
 720                     bcl->stats.recv_bundles,
 721                     bcl->stats.recv_bundled);
 722         tipc_printf(&pb, "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
 723                     bcl->stats.sent_info,
 724                     bcl->stats.sent_fragments,
 725                     bcl->stats.sent_fragmented,
 726                     bcl->stats.sent_bundles,
 727                     bcl->stats.sent_bundled);
 728         tipc_printf(&pb, "  RX naks:%u defs:%u dups:%u\n",
 729                     bcl->stats.recv_nacks,
 730                     bcl->stats.deferred_recv,
 731                     bcl->stats.duplicates);
 732         tipc_printf(&pb, "  TX naks:%u acks:%u dups:%u\n",
 733                     bcl->stats.sent_nacks,
 734                     bcl->stats.sent_acks,
 735                     bcl->stats.retransmitted);
 736         tipc_printf(&pb, "  Congestion bearer:%u link:%u  Send queue max:%u avg:%u\n",
 737                     bcl->stats.bearer_congs,
 738                     bcl->stats.link_congs,
 739                     bcl->stats.max_queue_sz,
 740                     bcl->stats.queue_sz_counts
 741                     ? (bcl->stats.accu_queue_sz / bcl->stats.queue_sz_counts)
 742                     : 0);
 743
 744         spin_unlock_bh(&bc_lock);
 745         return tipc_printbuf_validate(&pb);
 746 }
 747
 748 int tipc_bclink_reset_stats(void)
 749 {
 750         if (!bcl)
 751                 return -ENOPROTOOPT;
 752
 753         spin_lock_bh(&bc_lock);
 754         memset(&bcl->stats, 0, sizeof(bcl->stats));
 755         spin_unlock_bh(&bc_lock);
 756         return 0;
 757 }
 758
 759 int tipc_bclink_set_queue_limits(u32 limit)
 760 {
 761         if (!bcl)
 762                 return -ENOPROTOOPT;
 763         if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
 764                 return -EINVAL;
 765
 766         spin_lock_bh(&bc_lock);
 767         tipc_link_set_queue_limits(bcl, limit);
 768         spin_unlock_bh(&bc_lock);
 769         return 0;
 770 }
 771
 772 int tipc_bclink_init(void)
 773 {
 774         bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC);
 775         bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC);
 776         if (!bcbearer || !bclink) {
 777  nomem:
 778                 warn("Multicast link creation failed, no memory\n");
 779                 kfree(bcbearer);
 780                 bcbearer = NULL;
 781                 kfree(bclink);
 782                 bclink = NULL;
 783                 return -ENOMEM;
 784         }
 785
 786         INIT_LIST_HEAD(&bcbearer->bearer.cong_links);
 787         bcbearer->bearer.media = &bcbearer->media;
 788         bcbearer->media.send_msg = tipc_bcbearer_send;
 789         sprintf(bcbearer->media.name, "tipc-multicast");
 790
 791         bcl = &bclink->link;
 792         INIT_LIST_HEAD(&bcl->waiting_ports);
 793         bcl->next_out_no = 1;
 794         spin_lock_init(&bclink->node.lock);
 795         bcl->owner = &bclink->node;
 796         bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
 797         tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
 798         bcl->b_ptr = &bcbearer->bearer;
 799         bcl->state = WORKING_WORKING;
 800         strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
 801
 802         if (BCLINK_LOG_BUF_SIZE) {
 803                 char *pb = kmalloc(BCLINK_LOG_BUF_SIZE, GFP_ATOMIC);
 804
 805                 if (!pb)
 806                         goto nomem;
 807                 tipc_printbuf_init(&bcl->print_buf, pb, BCLINK_LOG_BUF_SIZE);
 808         }
 809
 810         return 0;
 811 }
 812
 813 void tipc_bclink_stop(void)
 814 {
 815         spin_lock_bh(&bc_lock);
 816         if (bcbearer) {
 817                 tipc_link_stop(bcl);
 818                 if (BCLINK_LOG_BUF_SIZE)
 819                         kfree(bcl->print_buf.buf);
 820                 bcl = NULL;
 821                 kfree(bclink);
 822                 bclink = NULL;
 823                 kfree(bcbearer);
 824                 bcbearer = NULL;
 825         }
 826         spin_unlock_bh(&bc_lock);
 827 }
 828
 829
 830 /**
 831  * tipc_nmap_add - add a node to a node map
 832  */
 833
 834 void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
 835 {
 836         int n = tipc_node(node);
 837         int w = n / WSIZE;
 838         u32 mask = (1 << (n % WSIZE));
 839
 840         if ((nm_ptr->map[w] & mask) == 0) {
 841                 nm_ptr->count++;
 842                 nm_ptr->map[w] |= mask;
 843         }
 844 }
 845
 846 /**
 847  * tipc_nmap_remove - remove a node from a node map
 848  */
 849
 850 void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
 851 {
 852         int n = tipc_node(node);
 853         int w = n / WSIZE;
 854         u32 mask = (1 << (n % WSIZE));
 855
 856         if ((nm_ptr->map[w] & mask) != 0) {
 857                 nm_ptr->map[w] &= ~mask;
 858                 nm_ptr->count--;
 859         }
 860 }
 861
 862 /**
 863  * tipc_nmap_diff - find differences between node maps
 864  * @nm_a: input node map A
 865  * @nm_b: input node map B
 866  * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
 867  */
 868
 869 static void tipc_nmap_diff(struct tipc_node_map *nm_a,
 870                            struct tipc_node_map *nm_b,
 871                            struct tipc_node_map *nm_diff)
 872 {
 873         int stop = ARRAY_SIZE(nm_a->map);
 874         int w;
 875         int b;
 876         u32 map;
 877
 878         memset(nm_diff, 0, sizeof(*nm_diff));
 879         for (w = 0; w < stop; w++) {
 880                 map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
 881                 nm_diff->map[w] = map;
 882                 if (map != 0) {
 883                         for (b = 0 ; b < WSIZE; b++) {
 884                                 if (map & (1 << b))
 885                                         nm_diff->count++;
 886                         }
 887                 }
 888         }
 889 }
 890
 891 /**
 892  * tipc_port_list_add - add a port to a port list, ensuring no duplicates
 893  */
 894
 895 void tipc_port_list_add(struct port_list *pl_ptr, u32 port)
 896 {
 897         struct port_list *item = pl_ptr;
 898         int i;
 899         int item_sz = PLSIZE;
 900         int cnt = pl_ptr->count;
 901
 902         for (; ; cnt -= item_sz, item = item->next) {
 903                 if (cnt < PLSIZE)
 904                         item_sz = cnt;
 905                 for (i = 0; i < item_sz; i++)
 906                         if (item->ports[i] == port)
 907                                 return;
 908                 if (i < PLSIZE) {
 909                         item->ports[i] = port;
 910                         pl_ptr->count++;
 911                         return;
 912                 }
 913                 if (!item->next) {
 914                         item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
 915                         if (!item->next) {
 916                                 warn("Incomplete multicast delivery, no memory\n");
 917                                 return;
 918                         }
 919                         item->next->next = NULL;
 920                 }
 921         }
 922 }
 923
 924 /**
 925  * tipc_port_list_free - free dynamically created entries in port_list chain
 926  *
 927  */
 928
 929 void tipc_port_list_free(struct port_list *pl_ptr)
 930 {
 931         struct port_list *item;
 932         struct port_list *next;
 933
 934         for (item = pl_ptr->next; item; item = next) {
 935                 next = item->next;
 936                 kfree(item);
 937         }
 938 }
 939