net/tipc/bcast.c

   1 /*
   2  * net/tipc/bcast.c: TIPC broadcast code
   3  *
   4  * Copyright (c) 2004-2006, 2014-2016, 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 <linux/tipc_config.h>
  39 #include "socket.h"
  40 #include "msg.h"
  41 #include "bcast.h"
  42 #include "link.h"
  43 #include "name_table.h"
  44
  45 #define BCLINK_WIN_DEFAULT      50      /* bcast link window size (default) */
  46 #define BCLINK_WIN_MIN          32      /* bcast minimum link window size */
  47
  48 const char tipc_bclink_name[] = "broadcast-link";
  49
  50 /**
  51  * struct tipc_bc_base - base structure for keeping broadcast send state
  52  * @link: broadcast send link structure
  53  * @inputq: data input queue; will only carry SOCK_WAKEUP messages
  54  * @dest: array keeping number of reachable destinations per bearer
  55  * @primary_bearer: a bearer having links to all broadcast destinations, if any
  56  * @bcast_support: indicates if primary bearer, if any, supports broadcast
  57  * @rcast_support: indicates if all peer nodes support replicast
  58  * @rc_ratio: dest count as percentage of cluster size where send method changes
  59  * @bc_threshold: calculated drom rc_ratio; if dests > threshold use broadcast
  60  */
  61 struct tipc_bc_base {
  62         struct tipc_link *link;
  63         struct sk_buff_head inputq;
  64         int dests[MAX_BEARERS];
  65         int primary_bearer;
  66         bool bcast_support;
  67         bool rcast_support;
  68         int rc_ratio;
  69         int bc_threshold;
  70 };
  71
  72 static struct tipc_bc_base *tipc_bc_base(struct net *net)
  73 {
  74         return tipc_net(net)->bcbase;
  75 }
  76
  77 int tipc_bcast_get_mtu(struct net *net)
  78 {
  79         return tipc_link_mtu(tipc_bc_sndlink(net)) - INT_H_SIZE;
  80 }
  81
  82 void tipc_bcast_disable_rcast(struct net *net)
  83 {
  84         tipc_bc_base(net)->rcast_support = false;
  85 }
  86
  87 static void tipc_bcbase_calc_bc_threshold(struct net *net)
  88 {
  89         struct tipc_bc_base *bb = tipc_bc_base(net);
  90         int cluster_size = tipc_link_bc_peers(tipc_bc_sndlink(net));
  91
  92         bb->bc_threshold = 1 + (cluster_size * bb->rc_ratio / 100);
  93 }
  94
  95 /* tipc_bcbase_select_primary(): find a bearer with links to all destinations,
  96  *                               if any, and make it primary bearer
  97  */
  98 static void tipc_bcbase_select_primary(struct net *net)
  99 {
 100         struct tipc_bc_base *bb = tipc_bc_base(net);
 101         int all_dests =  tipc_link_bc_peers(bb->link);
 102         int i, mtu, prim;
 103
 104         bb->primary_bearer = INVALID_BEARER_ID;
 105         bb->bcast_support = true;
 106
 107         if (!all_dests)
 108                 return;
 109
 110         for (i = 0; i < MAX_BEARERS; i++) {
 111                 if (!bb->dests[i])
 112                         continue;
 113
 114                 mtu = tipc_bearer_mtu(net, i);
 115                 if (mtu < tipc_link_mtu(bb->link))
 116                         tipc_link_set_mtu(bb->link, mtu);
 117                 bb->bcast_support &= tipc_bearer_bcast_support(net, i);
 118                 if (bb->dests[i] < all_dests)
 119                         continue;
 120
 121                 bb->primary_bearer = i;
 122
 123                 /* Reduce risk that all nodes select same primary */
 124                 if ((i ^ tipc_own_addr(net)) & 1)
 125                         break;
 126         }
 127         prim = bb->primary_bearer;
 128         if (prim != INVALID_BEARER_ID)
 129                 bb->bcast_support = tipc_bearer_bcast_support(net, prim);
 130 }
 131
 132 void tipc_bcast_inc_bearer_dst_cnt(struct net *net, int bearer_id)
 133 {
 134         struct tipc_bc_base *bb = tipc_bc_base(net);
 135
 136         tipc_bcast_lock(net);
 137         bb->dests[bearer_id]++;
 138         tipc_bcbase_select_primary(net);
 139         tipc_bcast_unlock(net);
 140 }
 141
 142 void tipc_bcast_dec_bearer_dst_cnt(struct net *net, int bearer_id)
 143 {
 144         struct tipc_bc_base *bb = tipc_bc_base(net);
 145
 146         tipc_bcast_lock(net);
 147         bb->dests[bearer_id]--;
 148         tipc_bcbase_select_primary(net);
 149         tipc_bcast_unlock(net);
 150 }
 151
 152 /* tipc_bcbase_xmit - broadcast a packet queue across one or more bearers
 153  *
 154  * Note that number of reachable destinations, as indicated in the dests[]
 155  * array, may transitionally differ from the number of destinations indicated
 156  * in each sent buffer. We can sustain this. Excess destination nodes will
 157  * drop and never acknowledge the unexpected packets, and missing destinations
 158  * will either require retransmission (if they are just about to be added to
 159  * the bearer), or be removed from the buffer's 'ackers' counter (if they
 160  * just went down)
 161  */
 162 static void tipc_bcbase_xmit(struct net *net, struct sk_buff_head *xmitq)
 163 {
 164         int bearer_id;
 165         struct tipc_bc_base *bb = tipc_bc_base(net);
 166         struct sk_buff *skb, *_skb;
 167         struct sk_buff_head _xmitq;
 168
 169         if (skb_queue_empty(xmitq))
 170                 return;
 171
 172         /* The typical case: at least one bearer has links to all nodes */
 173         bearer_id = bb->primary_bearer;
 174         if (bearer_id >= 0) {
 175                 tipc_bearer_bc_xmit(net, bearer_id, xmitq);
 176                 return;
 177         }
 178
 179         /* We have to transmit across all bearers */
 180         skb_queue_head_init(&_xmitq);
 181         for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
 182                 if (!bb->dests[bearer_id])
 183                         continue;
 184
 185                 skb_queue_walk(xmitq, skb) {
 186                         _skb = pskb_copy_for_clone(skb, GFP_ATOMIC);
 187                         if (!_skb)
 188                                 break;
 189                         __skb_queue_tail(&_xmitq, _skb);
 190                 }
 191                 tipc_bearer_bc_xmit(net, bearer_id, &_xmitq);
 192         }
 193         __skb_queue_purge(xmitq);
 194         __skb_queue_purge(&_xmitq);
 195 }
 196
 197 static void tipc_bcast_select_xmit_method(struct net *net, int dests,
 198                                           struct tipc_mc_method *method)
 199 {
 200         struct tipc_bc_base *bb = tipc_bc_base(net);
 201         unsigned long exp = method->expires;
 202
 203         /* Broadcast supported by used bearer/bearers? */
 204         if (!bb->bcast_support) {
 205                 method->rcast = true;
 206                 return;
 207         }
 208         /* Any destinations which don't support replicast ? */
 209         if (!bb->rcast_support) {
 210                 method->rcast = false;
 211                 return;
 212         }
 213         /* Can current method be changed ? */
 214         method->expires = jiffies + TIPC_METHOD_EXPIRE;
 215         if (method->mandatory || time_before(jiffies, exp))
 216                 return;
 217
 218         /* Determine method to use now */
 219         method->rcast = dests <= bb->bc_threshold;
 220 }
 221
 222 /* tipc_bcast_xmit - broadcast the buffer chain to all external nodes
 223  * @net: the applicable net namespace
 224  * @pkts: chain of buffers containing message
 225  * @cong_link_cnt: set to 1 if broadcast link is congested, otherwise 0
 226  * Consumes the buffer chain.
 227  * Returns 0 if success, otherwise errno: -EHOSTUNREACH,-EMSGSIZE
 228  */
 229 static int tipc_bcast_xmit(struct net *net, struct sk_buff_head *pkts,
 230                            u16 *cong_link_cnt)
 231 {
 232         struct tipc_link *l = tipc_bc_sndlink(net);
 233         struct sk_buff_head xmitq;
 234         int rc = 0;
 235
 236         __skb_queue_head_init(&xmitq);
 237         tipc_bcast_lock(net);
 238         if (tipc_link_bc_peers(l))
 239                 rc = tipc_link_xmit(l, pkts, &xmitq);
 240         tipc_bcast_unlock(net);
 241         tipc_bcbase_xmit(net, &xmitq);
 242         __skb_queue_purge(pkts);
 243         if (rc == -ELINKCONG) {
 244                 *cong_link_cnt = 1;
 245                 rc = 0;
 246         }
 247         return rc;
 248 }
 249
 250 /* tipc_rcast_xmit - replicate and send a message to given destination nodes
 251  * @net: the applicable net namespace
 252  * @pkts: chain of buffers containing message
 253  * @dests: list of destination nodes
 254  * @cong_link_cnt: returns number of congested links
 255  * @cong_links: returns identities of congested links
 256  * Returns 0 if success, otherwise errno
 257  */
 258 static int tipc_rcast_xmit(struct net *net, struct sk_buff_head *pkts,
 259                            struct tipc_nlist *dests, u16 *cong_link_cnt)
 260 {
 261         struct sk_buff_head _pkts;
 262         struct u32_item *n, *tmp;
 263         u32 dst, selector;
 264
 265         selector = msg_link_selector(buf_msg(skb_peek(pkts)));
 266         __skb_queue_head_init(&_pkts);
 267
 268         list_for_each_entry_safe(n, tmp, &dests->list, list) {
 269                 dst = n->value;
 270                 if (!tipc_msg_pskb_copy(dst, pkts, &_pkts))
 271                         return -ENOMEM;
 272
 273                 /* Any other return value than -ELINKCONG is ignored */
 274                 if (tipc_node_xmit(net, &_pkts, dst, selector) == -ELINKCONG)
 275                         (*cong_link_cnt)++;
 276         }
 277         return 0;
 278 }
 279
 280 /* tipc_mcast_xmit - deliver message to indicated destination nodes
 281  *                   and to identified node local sockets
 282  * @net: the applicable net namespace
 283  * @pkts: chain of buffers containing message
 284  * @method: send method to be used
 285  * @dests: destination nodes for message.
 286  * @cong_link_cnt: returns number of encountered congested destination links
 287  * Consumes buffer chain.
 288  * Returns 0 if success, otherwise errno
 289  */
 290 int tipc_mcast_xmit(struct net *net, struct sk_buff_head *pkts,
 291                     struct tipc_mc_method *method, struct tipc_nlist *dests,
 292                     u16 *cong_link_cnt)
 293 {
 294         struct sk_buff_head inputq, localq;
 295         int rc = 0;
 296
 297         skb_queue_head_init(&inputq);
 298         skb_queue_head_init(&localq);
 299
 300         /* Clone packets before they are consumed by next call */
 301         if (dests->local && !tipc_msg_reassemble(pkts, &localq)) {
 302                 rc = -ENOMEM;
 303                 goto exit;
 304         }
 305         /* Send according to determined transmit method */
 306         if (dests->remote) {
 307                 tipc_bcast_select_xmit_method(net, dests->remote, method);
 308                 if (method->rcast)
 309                         rc = tipc_rcast_xmit(net, pkts, dests, cong_link_cnt);
 310                 else
 311                         rc = tipc_bcast_xmit(net, pkts, cong_link_cnt);
 312         }
 313
 314         if (dests->local)
 315                 tipc_sk_mcast_rcv(net, &localq, &inputq);
 316 exit:
 317         /* This queue should normally be empty by now */
 318         __skb_queue_purge(pkts);
 319         return rc;
 320 }
 321
 322 /* tipc_bcast_rcv - receive a broadcast packet, and deliver to rcv link
 323  *
 324  * RCU is locked, no other locks set
 325  */
 326 int tipc_bcast_rcv(struct net *net, struct tipc_link *l, struct sk_buff *skb)
 327 {
 328         struct tipc_msg *hdr = buf_msg(skb);
 329         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 330         struct sk_buff_head xmitq;
 331         int rc;
 332
 333         __skb_queue_head_init(&xmitq);
 334
 335         if (msg_mc_netid(hdr) != tipc_netid(net) || !tipc_link_is_up(l)) {
 336                 kfree_skb(skb);
 337                 return 0;
 338         }
 339
 340         tipc_bcast_lock(net);
 341         if (msg_user(hdr) == BCAST_PROTOCOL)
 342                 rc = tipc_link_bc_nack_rcv(l, skb, &xmitq);
 343         else
 344                 rc = tipc_link_rcv(l, skb, NULL);
 345         tipc_bcast_unlock(net);
 346
 347         tipc_bcbase_xmit(net, &xmitq);
 348
 349         /* Any socket wakeup messages ? */
 350         if (!skb_queue_empty(inputq))
 351                 tipc_sk_rcv(net, inputq);
 352
 353         return rc;
 354 }
 355
 356 /* tipc_bcast_ack_rcv - receive and handle a broadcast acknowledge
 357  *
 358  * RCU is locked, no other locks set
 359  */
 360 void tipc_bcast_ack_rcv(struct net *net, struct tipc_link *l,
 361                         struct tipc_msg *hdr)
 362 {
 363         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 364         u16 acked = msg_bcast_ack(hdr);
 365         struct sk_buff_head xmitq;
 366
 367         /* Ignore bc acks sent by peer before bcast synch point was received */
 368         if (msg_bc_ack_invalid(hdr))
 369                 return;
 370
 371         __skb_queue_head_init(&xmitq);
 372
 373         tipc_bcast_lock(net);
 374         tipc_link_bc_ack_rcv(l, acked, &xmitq);
 375         tipc_bcast_unlock(net);
 376
 377         tipc_bcbase_xmit(net, &xmitq);
 378
 379         /* Any socket wakeup messages ? */
 380         if (!skb_queue_empty(inputq))
 381                 tipc_sk_rcv(net, inputq);
 382 }
 383
 384 /* tipc_bcast_synch_rcv -  check and update rcv link with peer's send state
 385  *
 386  * RCU is locked, no other locks set
 387  */
 388 int tipc_bcast_sync_rcv(struct net *net, struct tipc_link *l,
 389                         struct tipc_msg *hdr)
 390 {
 391         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 392         struct sk_buff_head xmitq;
 393         int rc = 0;
 394
 395         __skb_queue_head_init(&xmitq);
 396
 397         tipc_bcast_lock(net);
 398         if (msg_type(hdr) != STATE_MSG) {
 399                 tipc_link_bc_init_rcv(l, hdr);
 400         } else if (!msg_bc_ack_invalid(hdr)) {
 401                 tipc_link_bc_ack_rcv(l, msg_bcast_ack(hdr), &xmitq);
 402                 rc = tipc_link_bc_sync_rcv(l, hdr, &xmitq);
 403         }
 404         tipc_bcast_unlock(net);
 405
 406         tipc_bcbase_xmit(net, &xmitq);
 407
 408         /* Any socket wakeup messages ? */
 409         if (!skb_queue_empty(inputq))
 410                 tipc_sk_rcv(net, inputq);
 411         return rc;
 412 }
 413
 414 /* tipc_bcast_add_peer - add a peer node to broadcast link and bearer
 415  *
 416  * RCU is locked, node lock is set
 417  */
 418 void tipc_bcast_add_peer(struct net *net, struct tipc_link *uc_l,
 419                          struct sk_buff_head *xmitq)
 420 {
 421         struct tipc_link *snd_l = tipc_bc_sndlink(net);
 422
 423         tipc_bcast_lock(net);
 424         tipc_link_add_bc_peer(snd_l, uc_l, xmitq);
 425         tipc_bcbase_select_primary(net);
 426         tipc_bcbase_calc_bc_threshold(net);
 427         tipc_bcast_unlock(net);
 428 }
 429
 430 /* tipc_bcast_remove_peer - remove a peer node from broadcast link and bearer
 431  *
 432  * RCU is locked, node lock is set
 433  */
 434 void tipc_bcast_remove_peer(struct net *net, struct tipc_link *rcv_l)
 435 {
 436         struct tipc_link *snd_l = tipc_bc_sndlink(net);
 437         struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
 438         struct sk_buff_head xmitq;
 439
 440         __skb_queue_head_init(&xmitq);
 441
 442         tipc_bcast_lock(net);
 443         tipc_link_remove_bc_peer(snd_l, rcv_l, &xmitq);
 444         tipc_bcbase_select_primary(net);
 445         tipc_bcbase_calc_bc_threshold(net);
 446         tipc_bcast_unlock(net);
 447
 448         tipc_bcbase_xmit(net, &xmitq);
 449
 450         /* Any socket wakeup messages ? */
 451         if (!skb_queue_empty(inputq))
 452                 tipc_sk_rcv(net, inputq);
 453 }
 454
 455 int tipc_bclink_reset_stats(struct net *net)
 456 {
 457         struct tipc_link *l = tipc_bc_sndlink(net);
 458
 459         if (!l)
 460                 return -ENOPROTOOPT;
 461
 462         tipc_bcast_lock(net);
 463         tipc_link_reset_stats(l);
 464         tipc_bcast_unlock(net);
 465         return 0;
 466 }
 467
 468 static int tipc_bc_link_set_queue_limits(struct net *net, u32 limit)
 469 {
 470         struct tipc_link *l = tipc_bc_sndlink(net);
 471
 472         if (!l)
 473                 return -ENOPROTOOPT;
 474         if (limit < BCLINK_WIN_MIN)
 475                 limit = BCLINK_WIN_MIN;
 476         if (limit > TIPC_MAX_LINK_WIN)
 477                 return -EINVAL;
 478         tipc_bcast_lock(net);
 479         tipc_link_set_queue_limits(l, limit);
 480         tipc_bcast_unlock(net);
 481         return 0;
 482 }
 483
 484 int tipc_nl_bc_link_set(struct net *net, struct nlattr *attrs[])
 485 {
 486         int err;
 487         u32 win;
 488         struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
 489
 490         if (!attrs[TIPC_NLA_LINK_PROP])
 491                 return -EINVAL;
 492
 493         err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP], props);
 494         if (err)
 495                 return err;
 496
 497         if (!props[TIPC_NLA_PROP_WIN])
 498                 return -EOPNOTSUPP;
 499
 500         win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
 501
 502         return tipc_bc_link_set_queue_limits(net, win);
 503 }
 504
 505 int tipc_bcast_init(struct net *net)
 506 {
 507         struct tipc_net *tn = tipc_net(net);
 508         struct tipc_bc_base *bb = NULL;
 509         struct tipc_link *l = NULL;
 510
 511         bb = kzalloc(sizeof(*bb), GFP_ATOMIC);
 512         if (!bb)
 513                 goto enomem;
 514         tn->bcbase = bb;
 515         spin_lock_init(&tipc_net(net)->bclock);
 516
 517         if (!tipc_link_bc_create(net, 0, 0,
 518                                  U16_MAX,
 519                                  BCLINK_WIN_DEFAULT,
 520                                  0,
 521                                  &bb->inputq,
 522                                  NULL,
 523                                  NULL,
 524                                  &l))
 525                 goto enomem;
 526         bb->link = l;
 527         tn->bcl = l;
 528         bb->rc_ratio = 25;
 529         bb->rcast_support = true;
 530         return 0;
 531 enomem:
 532         kfree(bb);
 533         kfree(l);
 534         return -ENOMEM;
 535 }
 536
 537 void tipc_bcast_stop(struct net *net)
 538 {
 539         struct tipc_net *tn = net_generic(net, tipc_net_id);
 540
 541         synchronize_net();
 542         kfree(tn->bcbase);
 543         kfree(tn->bcl);
 544 }
 545
 546 void tipc_nlist_init(struct tipc_nlist *nl, u32 self)
 547 {
 548         memset(nl, 0, sizeof(*nl));
 549         INIT_LIST_HEAD(&nl->list);
 550         nl->self = self;
 551 }
 552
 553 void tipc_nlist_add(struct tipc_nlist *nl, u32 node)
 554 {
 555         if (node == nl->self)
 556                 nl->local = true;
 557         else if (u32_push(&nl->list, node))
 558                 nl->remote++;
 559 }
 560
 561 void tipc_nlist_del(struct tipc_nlist *nl, u32 node)
 562 {
 563         if (node == nl->self)
 564                 nl->local = false;
 565         else if (u32_del(&nl->list, node))
 566                 nl->remote--;
 567 }
 568
 569 void tipc_nlist_purge(struct tipc_nlist *nl)
 570 {
 571         u32_list_purge(&nl->list);
 572         nl->remote = 0;
 573         nl->local = 0;
 574 }