net/tipc/link.c

   1 /*
   2  * net/tipc/link.c: TIPC link code
   3  *
   4  * Copyright (c) 1996-2007, 2012-2014, Ericsson AB
   5  * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
   6  * All rights reserved.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions are met:
  10  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the names of the copyright holders nor the names of its
  17  *    contributors may be used to endorse or promote products derived from
  18  *    this software without specific prior written permission.
  19  *
  20  * Alternatively, this software may be distributed under the terms of the
  21  * GNU General Public License ("GPL") version 2 as published by the Free
  22  * Software Foundation.
  23  *
  24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  34  * POSSIBILITY OF SUCH DAMAGE.
  35  */
  36
  37 #include "core.h"
  38 #include "link.h"
  39 #include "port.h"
  40 #include "socket.h"
  41 #include "name_distr.h"
  42 #include "discover.h"
  43 #include "config.h"
  44
  45 #include <linux/pkt_sched.h>
  46
  47 /*
  48  * Error message prefixes
  49  */
  50 static const char *link_co_err = "Link changeover error, ";
  51 static const char *link_rst_msg = "Resetting link ";
  52 static const char *link_unk_evt = "Unknown link event ";
  53
  54 /*
  55  * Out-of-range value for link session numbers
  56  */
  57 #define INVALID_SESSION 0x10000
  58
  59 /*
  60  * Link state events:
  61  */
  62 #define  STARTING_EVT    856384768      /* link processing trigger */
  63 #define  TRAFFIC_MSG_EVT 560815u        /* rx'd ??? */
  64 #define  TIMEOUT_EVT     560817u        /* link timer expired */
  65
  66 /*
  67  * The following two 'message types' is really just implementation
  68  * data conveniently stored in the message header.
  69  * They must not be considered part of the protocol
  70  */
  71 #define OPEN_MSG   0
  72 #define CLOSED_MSG 1
  73
  74 /*
  75  * State value stored in 'exp_msg_count'
  76  */
  77 #define START_CHANGEOVER 100000u
  78
  79 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
  80                                        struct sk_buff *buf);
  81 static void tipc_link_proto_rcv(struct tipc_link *l_ptr, struct sk_buff *buf);
  82 static int  tipc_link_tunnel_rcv(struct tipc_node *n_ptr,
  83                                  struct sk_buff **buf);
  84 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance);
  85 static int  tipc_link_iovec_long_xmit(struct tipc_port *sender,
  86                                       struct iovec const *msg_sect,
  87                                       unsigned int len, u32 destnode);
  88 static void link_state_event(struct tipc_link *l_ptr, u32 event);
  89 static void link_reset_statistics(struct tipc_link *l_ptr);
  90 static void link_print(struct tipc_link *l_ptr, const char *str);
  91 static int tipc_link_frag_xmit(struct tipc_link *l_ptr, struct sk_buff *buf);
  92 static void tipc_link_sync_xmit(struct tipc_link *l);
  93 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf);
  94
  95 /*
  96  *  Simple link routines
  97  */
  98 static unsigned int align(unsigned int i)
  99 {
 100         return (i + 3) & ~3u;
 101 }
 102
 103 static void link_init_max_pkt(struct tipc_link *l_ptr)
 104 {
 105         struct tipc_bearer *b_ptr;
 106         u32 max_pkt;
 107
 108         rcu_read_lock();
 109         b_ptr = rcu_dereference_rtnl(bearer_list[l_ptr->bearer_id]);
 110         if (!b_ptr) {
 111                 rcu_read_unlock();
 112                 return;
 113         }
 114         max_pkt = (b_ptr->mtu & ~3);
 115         rcu_read_unlock();
 116
 117         if (max_pkt > MAX_MSG_SIZE)
 118                 max_pkt = MAX_MSG_SIZE;
 119
 120         l_ptr->max_pkt_target = max_pkt;
 121         if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
 122                 l_ptr->max_pkt = l_ptr->max_pkt_target;
 123         else
 124                 l_ptr->max_pkt = MAX_PKT_DEFAULT;
 125
 126         l_ptr->max_pkt_probes = 0;
 127 }
 128
 129 static u32 link_next_sent(struct tipc_link *l_ptr)
 130 {
 131         if (l_ptr->next_out)
 132                 return buf_seqno(l_ptr->next_out);
 133         return mod(l_ptr->next_out_no);
 134 }
 135
 136 static u32 link_last_sent(struct tipc_link *l_ptr)
 137 {
 138         return mod(link_next_sent(l_ptr) - 1);
 139 }
 140
 141 /*
 142  *  Simple non-static link routines (i.e. referenced outside this file)
 143  */
 144 int tipc_link_is_up(struct tipc_link *l_ptr)
 145 {
 146         if (!l_ptr)
 147                 return 0;
 148         return link_working_working(l_ptr) || link_working_unknown(l_ptr);
 149 }
 150
 151 int tipc_link_is_active(struct tipc_link *l_ptr)
 152 {
 153         return  (l_ptr->owner->active_links[0] == l_ptr) ||
 154                 (l_ptr->owner->active_links[1] == l_ptr);
 155 }
 156
 157 /**
 158  * link_timeout - handle expiration of link timer
 159  * @l_ptr: pointer to link
 160  */
 161 static void link_timeout(struct tipc_link *l_ptr)
 162 {
 163         tipc_node_lock(l_ptr->owner);
 164
 165         /* update counters used in statistical profiling of send traffic */
 166         l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
 167         l_ptr->stats.queue_sz_counts++;
 168
 169         if (l_ptr->first_out) {
 170                 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
 171                 u32 length = msg_size(msg);
 172
 173                 if ((msg_user(msg) == MSG_FRAGMENTER) &&
 174                     (msg_type(msg) == FIRST_FRAGMENT)) {
 175                         length = msg_size(msg_get_wrapped(msg));
 176                 }
 177                 if (length) {
 178                         l_ptr->stats.msg_lengths_total += length;
 179                         l_ptr->stats.msg_length_counts++;
 180                         if (length <= 64)
 181                                 l_ptr->stats.msg_length_profile[0]++;
 182                         else if (length <= 256)
 183                                 l_ptr->stats.msg_length_profile[1]++;
 184                         else if (length <= 1024)
 185                                 l_ptr->stats.msg_length_profile[2]++;
 186                         else if (length <= 4096)
 187                                 l_ptr->stats.msg_length_profile[3]++;
 188                         else if (length <= 16384)
 189                                 l_ptr->stats.msg_length_profile[4]++;
 190                         else if (length <= 32768)
 191                                 l_ptr->stats.msg_length_profile[5]++;
 192                         else
 193                                 l_ptr->stats.msg_length_profile[6]++;
 194                 }
 195         }
 196
 197         /* do all other link processing performed on a periodic basis */
 198
 199         link_state_event(l_ptr, TIMEOUT_EVT);
 200
 201         if (l_ptr->next_out)
 202                 tipc_link_push_queue(l_ptr);
 203
 204         tipc_node_unlock(l_ptr->owner);
 205 }
 206
 207 static void link_set_timer(struct tipc_link *l_ptr, u32 time)
 208 {
 209         k_start_timer(&l_ptr->timer, time);
 210 }
 211
 212 /**
 213  * tipc_link_create - create a new link
 214  * @n_ptr: pointer to associated node
 215  * @b_ptr: pointer to associated bearer
 216  * @media_addr: media address to use when sending messages over link
 217  *
 218  * Returns pointer to link.
 219  */
 220 struct tipc_link *tipc_link_create(struct tipc_node *n_ptr,
 221                                    struct tipc_bearer *b_ptr,
 222                                    const struct tipc_media_addr *media_addr)
 223 {
 224         struct tipc_link *l_ptr;
 225         struct tipc_msg *msg;
 226         char *if_name;
 227         char addr_string[16];
 228         u32 peer = n_ptr->addr;
 229
 230         if (n_ptr->link_cnt >= 2) {
 231                 tipc_addr_string_fill(addr_string, n_ptr->addr);
 232                 pr_err("Attempt to establish third link to %s\n", addr_string);
 233                 return NULL;
 234         }
 235
 236         if (n_ptr->links[b_ptr->identity]) {
 237                 tipc_addr_string_fill(addr_string, n_ptr->addr);
 238                 pr_err("Attempt to establish second link on <%s> to %s\n",
 239                        b_ptr->name, addr_string);
 240                 return NULL;
 241         }
 242
 243         l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
 244         if (!l_ptr) {
 245                 pr_warn("Link creation failed, no memory\n");
 246                 return NULL;
 247         }
 248
 249         l_ptr->addr = peer;
 250         if_name = strchr(b_ptr->name, ':') + 1;
 251         sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
 252                 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
 253                 tipc_node(tipc_own_addr),
 254                 if_name,
 255                 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
 256                 /* note: peer i/f name is updated by reset/activate message */
 257         memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
 258         l_ptr->owner = n_ptr;
 259         l_ptr->checkpoint = 1;
 260         l_ptr->peer_session = INVALID_SESSION;
 261         l_ptr->bearer_id = b_ptr->identity;
 262         link_set_supervision_props(l_ptr, b_ptr->tolerance);
 263         l_ptr->state = RESET_UNKNOWN;
 264
 265         l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
 266         msg = l_ptr->pmsg;
 267         tipc_msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
 268         msg_set_size(msg, sizeof(l_ptr->proto_msg));
 269         msg_set_session(msg, (tipc_random & 0xffff));
 270         msg_set_bearer_id(msg, b_ptr->identity);
 271         strcpy((char *)msg_data(msg), if_name);
 272
 273         l_ptr->priority = b_ptr->priority;
 274         tipc_link_set_queue_limits(l_ptr, b_ptr->window);
 275
 276         l_ptr->net_plane = b_ptr->net_plane;
 277         link_init_max_pkt(l_ptr);
 278
 279         l_ptr->next_out_no = 1;
 280         INIT_LIST_HEAD(&l_ptr->waiting_ports);
 281
 282         link_reset_statistics(l_ptr);
 283
 284         tipc_node_attach_link(n_ptr, l_ptr);
 285
 286         k_init_timer(&l_ptr->timer, (Handler)link_timeout,
 287                      (unsigned long)l_ptr);
 288
 289         link_state_event(l_ptr, STARTING_EVT);
 290
 291         return l_ptr;
 292 }
 293
 294 void tipc_link_delete_list(unsigned int bearer_id, bool shutting_down)
 295 {
 296         struct tipc_link *l_ptr;
 297         struct tipc_node *n_ptr;
 298
 299         rcu_read_lock();
 300         list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
 301                 tipc_node_lock(n_ptr);
 302                 l_ptr = n_ptr->links[bearer_id];
 303                 if (l_ptr) {
 304                         tipc_link_reset(l_ptr);
 305                         if (shutting_down || !tipc_node_is_up(n_ptr)) {
 306                                 tipc_node_detach_link(l_ptr->owner, l_ptr);
 307                                 tipc_link_reset_fragments(l_ptr);
 308                                 tipc_node_unlock(n_ptr);
 309
 310                                 /* Nobody else can access this link now: */
 311                                 del_timer_sync(&l_ptr->timer);
 312                                 kfree(l_ptr);
 313                         } else {
 314                                 /* Detach/delete when failover is finished: */
 315                                 l_ptr->flags |= LINK_STOPPED;
 316                                 tipc_node_unlock(n_ptr);
 317                                 del_timer_sync(&l_ptr->timer);
 318                         }
 319                         continue;
 320                 }
 321                 tipc_node_unlock(n_ptr);
 322         }
 323         rcu_read_unlock();
 324 }
 325
 326 /**
 327  * link_schedule_port - schedule port for deferred sending
 328  * @l_ptr: pointer to link
 329  * @origport: reference to sending port
 330  * @sz: amount of data to be sent
 331  *
 332  * Schedules port for renewed sending of messages after link congestion
 333  * has abated.
 334  */
 335 static int link_schedule_port(struct tipc_link *l_ptr, u32 origport, u32 sz)
 336 {
 337         struct tipc_port *p_ptr;
 338
 339         spin_lock_bh(&tipc_port_list_lock);
 340         p_ptr = tipc_port_lock(origport);
 341         if (p_ptr) {
 342                 if (!list_empty(&p_ptr->wait_list))
 343                         goto exit;
 344                 p_ptr->congested = 1;
 345                 p_ptr->waiting_pkts = 1 + ((sz - 1) / l_ptr->max_pkt);
 346                 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
 347                 l_ptr->stats.link_congs++;
 348 exit:
 349                 tipc_port_unlock(p_ptr);
 350         }
 351         spin_unlock_bh(&tipc_port_list_lock);
 352         return -ELINKCONG;
 353 }
 354
 355 void tipc_link_wakeup_ports(struct tipc_link *l_ptr, int all)
 356 {
 357         struct tipc_port *p_ptr;
 358         struct tipc_port *temp_p_ptr;
 359         int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
 360
 361         if (all)
 362                 win = 100000;
 363         if (win <= 0)
 364                 return;
 365         if (!spin_trylock_bh(&tipc_port_list_lock))
 366                 return;
 367         if (link_congested(l_ptr))
 368                 goto exit;
 369         list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
 370                                  wait_list) {
 371                 if (win <= 0)
 372                         break;
 373                 list_del_init(&p_ptr->wait_list);
 374                 spin_lock_bh(p_ptr->lock);
 375                 p_ptr->congested = 0;
 376                 tipc_port_wakeup(p_ptr);
 377                 win -= p_ptr->waiting_pkts;
 378                 spin_unlock_bh(p_ptr->lock);
 379         }
 380
 381 exit:
 382         spin_unlock_bh(&tipc_port_list_lock);
 383 }
 384
 385 /**
 386  * link_release_outqueue - purge link's outbound message queue
 387  * @l_ptr: pointer to link
 388  */
 389 static void link_release_outqueue(struct tipc_link *l_ptr)
 390 {
 391         kfree_skb_list(l_ptr->first_out);
 392         l_ptr->first_out = NULL;
 393         l_ptr->out_queue_size = 0;
 394 }
 395
 396 /**
 397  * tipc_link_reset_fragments - purge link's inbound message fragments queue
 398  * @l_ptr: pointer to link
 399  */
 400 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
 401 {
 402         kfree_skb(l_ptr->reasm_buf);
 403         l_ptr->reasm_buf = NULL;
 404 }
 405
 406 /**
 407  * tipc_link_purge_queues - purge all pkt queues associated with link
 408  * @l_ptr: pointer to link
 409  */
 410 void tipc_link_purge_queues(struct tipc_link *l_ptr)
 411 {
 412         kfree_skb_list(l_ptr->oldest_deferred_in);
 413         kfree_skb_list(l_ptr->first_out);
 414         tipc_link_reset_fragments(l_ptr);
 415         kfree_skb(l_ptr->proto_msg_queue);
 416         l_ptr->proto_msg_queue = NULL;
 417 }
 418
 419 void tipc_link_reset(struct tipc_link *l_ptr)
 420 {
 421         u32 prev_state = l_ptr->state;
 422         u32 checkpoint = l_ptr->next_in_no;
 423         int was_active_link = tipc_link_is_active(l_ptr);
 424
 425         msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
 426
 427         /* Link is down, accept any session */
 428         l_ptr->peer_session = INVALID_SESSION;
 429
 430         /* Prepare for max packet size negotiation */
 431         link_init_max_pkt(l_ptr);
 432
 433         l_ptr->state = RESET_UNKNOWN;
 434
 435         if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
 436                 return;
 437
 438         tipc_node_link_down(l_ptr->owner, l_ptr);
 439         tipc_bearer_remove_dest(l_ptr->bearer_id, l_ptr->addr);
 440
 441         if (was_active_link && tipc_node_active_links(l_ptr->owner)) {
 442                 l_ptr->reset_checkpoint = checkpoint;
 443                 l_ptr->exp_msg_count = START_CHANGEOVER;
 444         }
 445
 446         /* Clean up all queues: */
 447         link_release_outqueue(l_ptr);
 448         kfree_skb(l_ptr->proto_msg_queue);
 449         l_ptr->proto_msg_queue = NULL;
 450         kfree_skb_list(l_ptr->oldest_deferred_in);
 451         if (!list_empty(&l_ptr->waiting_ports))
 452                 tipc_link_wakeup_ports(l_ptr, 1);
 453
 454         l_ptr->retransm_queue_head = 0;
 455         l_ptr->retransm_queue_size = 0;
 456         l_ptr->last_out = NULL;
 457         l_ptr->first_out = NULL;
 458         l_ptr->next_out = NULL;
 459         l_ptr->unacked_window = 0;
 460         l_ptr->checkpoint = 1;
 461         l_ptr->next_out_no = 1;
 462         l_ptr->deferred_inqueue_sz = 0;
 463         l_ptr->oldest_deferred_in = NULL;
 464         l_ptr->newest_deferred_in = NULL;
 465         l_ptr->fsm_msg_cnt = 0;
 466         l_ptr->stale_count = 0;
 467         link_reset_statistics(l_ptr);
 468 }
 469
 470 void tipc_link_reset_list(unsigned int bearer_id)
 471 {
 472         struct tipc_link *l_ptr;
 473         struct tipc_node *n_ptr;
 474
 475         rcu_read_lock();
 476         list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
 477                 tipc_node_lock(n_ptr);
 478                 l_ptr = n_ptr->links[bearer_id];
 479                 if (l_ptr)
 480                         tipc_link_reset(l_ptr);
 481                 tipc_node_unlock(n_ptr);
 482         }
 483         rcu_read_unlock();
 484 }
 485
 486 static void link_activate(struct tipc_link *l_ptr)
 487 {
 488         l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
 489         tipc_node_link_up(l_ptr->owner, l_ptr);
 490         tipc_bearer_add_dest(l_ptr->bearer_id, l_ptr->addr);
 491 }
 492
 493 /**
 494  * link_state_event - link finite state machine
 495  * @l_ptr: pointer to link
 496  * @event: state machine event to process
 497  */
 498 static void link_state_event(struct tipc_link *l_ptr, unsigned int event)
 499 {
 500         struct tipc_link *other;
 501         u32 cont_intv = l_ptr->continuity_interval;
 502
 503         if (l_ptr->flags & LINK_STOPPED)
 504                 return;
 505
 506         if (!(l_ptr->flags & LINK_STARTED) && (event != STARTING_EVT))
 507                 return;         /* Not yet. */
 508
 509         /* Check whether changeover is going on */
 510         if (l_ptr->exp_msg_count) {
 511                 if (event == TIMEOUT_EVT)
 512                         link_set_timer(l_ptr, cont_intv);
 513                 return;
 514         }
 515
 516         switch (l_ptr->state) {
 517         case WORKING_WORKING:
 518                 switch (event) {
 519                 case TRAFFIC_MSG_EVT:
 520                 case ACTIVATE_MSG:
 521                         break;
 522                 case TIMEOUT_EVT:
 523                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
 524                                 l_ptr->checkpoint = l_ptr->next_in_no;
 525                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
 526                                         tipc_link_proto_xmit(l_ptr, STATE_MSG,
 527                                                              0, 0, 0, 0, 0);
 528                                         l_ptr->fsm_msg_cnt++;
 529                                 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
 530                                         tipc_link_proto_xmit(l_ptr, STATE_MSG,
 531                                                              1, 0, 0, 0, 0);
 532                                         l_ptr->fsm_msg_cnt++;
 533                                 }
 534                                 link_set_timer(l_ptr, cont_intv);
 535                                 break;
 536                         }
 537                         l_ptr->state = WORKING_UNKNOWN;
 538                         l_ptr->fsm_msg_cnt = 0;
 539                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
 540                         l_ptr->fsm_msg_cnt++;
 541                         link_set_timer(l_ptr, cont_intv / 4);
 542                         break;
 543                 case RESET_MSG:
 544                         pr_info("%s<%s>, requested by peer\n", link_rst_msg,
 545                                 l_ptr->name);
 546                         tipc_link_reset(l_ptr);
 547                         l_ptr->state = RESET_RESET;
 548                         l_ptr->fsm_msg_cnt = 0;
 549                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
 550                                              0, 0, 0, 0, 0);
 551                         l_ptr->fsm_msg_cnt++;
 552                         link_set_timer(l_ptr, cont_intv);
 553                         break;
 554                 default:
 555                         pr_err("%s%u in WW state\n", link_unk_evt, event);
 556                 }
 557                 break;
 558         case WORKING_UNKNOWN:
 559                 switch (event) {
 560                 case TRAFFIC_MSG_EVT:
 561                 case ACTIVATE_MSG:
 562                         l_ptr->state = WORKING_WORKING;
 563                         l_ptr->fsm_msg_cnt = 0;
 564                         link_set_timer(l_ptr, cont_intv);
 565                         break;
 566                 case RESET_MSG:
 567                         pr_info("%s<%s>, requested by peer while probing\n",
 568                                 link_rst_msg, l_ptr->name);
 569                         tipc_link_reset(l_ptr);
 570                         l_ptr->state = RESET_RESET;
 571                         l_ptr->fsm_msg_cnt = 0;
 572                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
 573                                              0, 0, 0, 0, 0);
 574                         l_ptr->fsm_msg_cnt++;
 575                         link_set_timer(l_ptr, cont_intv);
 576                         break;
 577                 case TIMEOUT_EVT:
 578                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
 579                                 l_ptr->state = WORKING_WORKING;
 580                                 l_ptr->fsm_msg_cnt = 0;
 581                                 l_ptr->checkpoint = l_ptr->next_in_no;
 582                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
 583                                         tipc_link_proto_xmit(l_ptr, STATE_MSG,
 584                                                              0, 0, 0, 0, 0);
 585                                         l_ptr->fsm_msg_cnt++;
 586                                 }
 587                                 link_set_timer(l_ptr, cont_intv);
 588                         } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
 589                                 tipc_link_proto_xmit(l_ptr, STATE_MSG,
 590                                                      1, 0, 0, 0, 0);
 591                                 l_ptr->fsm_msg_cnt++;
 592                                 link_set_timer(l_ptr, cont_intv / 4);
 593                         } else {        /* Link has failed */
 594                                 pr_warn("%s<%s>, peer not responding\n",
 595                                         link_rst_msg, l_ptr->name);
 596                                 tipc_link_reset(l_ptr);
 597                                 l_ptr->state = RESET_UNKNOWN;
 598                                 l_ptr->fsm_msg_cnt = 0;
 599                                 tipc_link_proto_xmit(l_ptr, RESET_MSG,
 600                                                      0, 0, 0, 0, 0);
 601                                 l_ptr->fsm_msg_cnt++;
 602                                 link_set_timer(l_ptr, cont_intv);
 603                         }
 604                         break;
 605                 default:
 606                         pr_err("%s%u in WU state\n", link_unk_evt, event);
 607                 }
 608                 break;
 609         case RESET_UNKNOWN:
 610                 switch (event) {
 611                 case TRAFFIC_MSG_EVT:
 612                         break;
 613                 case ACTIVATE_MSG:
 614                         other = l_ptr->owner->active_links[0];
 615                         if (other && link_working_unknown(other))
 616                                 break;
 617                         l_ptr->state = WORKING_WORKING;
 618                         l_ptr->fsm_msg_cnt = 0;
 619                         link_activate(l_ptr);
 620                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
 621                         l_ptr->fsm_msg_cnt++;
 622                         if (l_ptr->owner->working_links == 1)
 623                                 tipc_link_sync_xmit(l_ptr);
 624                         link_set_timer(l_ptr, cont_intv);
 625                         break;
 626                 case RESET_MSG:
 627                         l_ptr->state = RESET_RESET;
 628                         l_ptr->fsm_msg_cnt = 0;
 629                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
 630                                              1, 0, 0, 0, 0);
 631                         l_ptr->fsm_msg_cnt++;
 632                         link_set_timer(l_ptr, cont_intv);
 633                         break;
 634                 case STARTING_EVT:
 635                         l_ptr->flags |= LINK_STARTED;
 636                         /* fall through */
 637                 case TIMEOUT_EVT:
 638                         tipc_link_proto_xmit(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
 639                         l_ptr->fsm_msg_cnt++;
 640                         link_set_timer(l_ptr, cont_intv);
 641                         break;
 642                 default:
 643                         pr_err("%s%u in RU state\n", link_unk_evt, event);
 644                 }
 645                 break;
 646         case RESET_RESET:
 647                 switch (event) {
 648                 case TRAFFIC_MSG_EVT:
 649                 case ACTIVATE_MSG:
 650                         other = l_ptr->owner->active_links[0];
 651                         if (other && link_working_unknown(other))
 652                                 break;
 653                         l_ptr->state = WORKING_WORKING;
 654                         l_ptr->fsm_msg_cnt = 0;
 655                         link_activate(l_ptr);
 656                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
 657                         l_ptr->fsm_msg_cnt++;
 658                         if (l_ptr->owner->working_links == 1)
 659                                 tipc_link_sync_xmit(l_ptr);
 660                         link_set_timer(l_ptr, cont_intv);
 661                         break;
 662                 case RESET_MSG:
 663                         break;
 664                 case TIMEOUT_EVT:
 665                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
 666                                              0, 0, 0, 0, 0);
 667                         l_ptr->fsm_msg_cnt++;
 668                         link_set_timer(l_ptr, cont_intv);
 669                         break;
 670                 default:
 671                         pr_err("%s%u in RR state\n", link_unk_evt, event);
 672                 }
 673                 break;
 674         default:
 675                 pr_err("Unknown link state %u/%u\n", l_ptr->state, event);
 676         }
 677 }
 678
 679 /*
 680  * link_bundle_buf(): Append contents of a buffer to
 681  * the tail of an existing one.
 682  */
 683 static int link_bundle_buf(struct tipc_link *l_ptr, struct sk_buff *bundler,
 684                            struct sk_buff *buf)
 685 {
 686         struct tipc_msg *bundler_msg = buf_msg(bundler);
 687         struct tipc_msg *msg = buf_msg(buf);
 688         u32 size = msg_size(msg);
 689         u32 bundle_size = msg_size(bundler_msg);
 690         u32 to_pos = align(bundle_size);
 691         u32 pad = to_pos - bundle_size;
 692
 693         if (msg_user(bundler_msg) != MSG_BUNDLER)
 694                 return 0;
 695         if (msg_type(bundler_msg) != OPEN_MSG)
 696                 return 0;
 697         if (skb_tailroom(bundler) < (pad + size))
 698                 return 0;
 699         if (l_ptr->max_pkt < (to_pos + size))
 700                 return 0;
 701
 702         skb_put(bundler, pad + size);
 703         skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
 704         msg_set_size(bundler_msg, to_pos + size);
 705         msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
 706         kfree_skb(buf);
 707         l_ptr->stats.sent_bundled++;
 708         return 1;
 709 }
 710
 711 static void link_add_to_outqueue(struct tipc_link *l_ptr,
 712                                  struct sk_buff *buf,
 713                                  struct tipc_msg *msg)
 714 {
 715         u32 ack = mod(l_ptr->next_in_no - 1);
 716         u32 seqno = mod(l_ptr->next_out_no++);
 717
 718         msg_set_word(msg, 2, ((ack << 16) | seqno));
 719         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
 720         buf->next = NULL;
 721         if (l_ptr->first_out) {
 722                 l_ptr->last_out->next = buf;
 723                 l_ptr->last_out = buf;
 724         } else
 725                 l_ptr->first_out = l_ptr->last_out = buf;
 726
 727         l_ptr->out_queue_size++;
 728         if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
 729                 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
 730 }
 731
 732 static void link_add_chain_to_outqueue(struct tipc_link *l_ptr,
 733                                        struct sk_buff *buf_chain,
 734                                        u32 long_msgno)
 735 {
 736         struct sk_buff *buf;
 737         struct tipc_msg *msg;
 738
 739         if (!l_ptr->next_out)
 740                 l_ptr->next_out = buf_chain;
 741         while (buf_chain) {
 742                 buf = buf_chain;
 743                 buf_chain = buf_chain->next;
 744
 745                 msg = buf_msg(buf);
 746                 msg_set_long_msgno(msg, long_msgno);
 747                 link_add_to_outqueue(l_ptr, buf, msg);
 748         }
 749 }
 750
 751 /*
 752  * tipc_link_xmit() is the 'full path' for messages, called from
 753  * inside TIPC when the 'fast path' in tipc_send_xmit
 754  * has failed, and from link_send()
 755  */
 756 int __tipc_link_xmit(struct tipc_link *l_ptr, struct sk_buff *buf)
 757 {
 758         struct tipc_msg *msg = buf_msg(buf);
 759         u32 size = msg_size(msg);
 760         u32 dsz = msg_data_sz(msg);
 761         u32 queue_size = l_ptr->out_queue_size;
 762         u32 imp = tipc_msg_tot_importance(msg);
 763         u32 queue_limit = l_ptr->queue_limit[imp];
 764         u32 max_packet = l_ptr->max_pkt;
 765
 766         /* Match msg importance against queue limits: */
 767         if (unlikely(queue_size >= queue_limit)) {
 768                 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
 769                         link_schedule_port(l_ptr, msg_origport(msg), size);
 770                         kfree_skb(buf);
 771                         return -ELINKCONG;
 772                 }
 773                 kfree_skb(buf);
 774                 if (imp > CONN_MANAGER) {
 775                         pr_warn("%s<%s>, send queue full", link_rst_msg,
 776                                 l_ptr->name);
 777                         tipc_link_reset(l_ptr);
 778                 }
 779                 return dsz;
 780         }
 781
 782         /* Fragmentation needed ? */
 783         if (size > max_packet)
 784                 return tipc_link_frag_xmit(l_ptr, buf);
 785
 786         /* Packet can be queued or sent. */
 787         if (likely(!link_congested(l_ptr))) {
 788                 link_add_to_outqueue(l_ptr, buf, msg);
 789
 790                 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
 791                 l_ptr->unacked_window = 0;
 792                 return dsz;
 793         }
 794         /* Congestion: can message be bundled ? */
 795         if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
 796             (msg_user(msg) != MSG_FRAGMENTER)) {
 797
 798                 /* Try adding message to an existing bundle */
 799                 if (l_ptr->next_out &&
 800                     link_bundle_buf(l_ptr, l_ptr->last_out, buf))
 801                         return dsz;
 802
 803                 /* Try creating a new bundle */
 804                 if (size <= max_packet * 2 / 3) {
 805                         struct sk_buff *bundler = tipc_buf_acquire(max_packet);
 806                         struct tipc_msg bundler_hdr;
 807
 808                         if (bundler) {
 809                                 tipc_msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
 810                                          INT_H_SIZE, l_ptr->addr);
 811                                 skb_copy_to_linear_data(bundler, &bundler_hdr,
 812                                                         INT_H_SIZE);
 813                                 skb_trim(bundler, INT_H_SIZE);
 814                                 link_bundle_buf(l_ptr, bundler, buf);
 815                                 buf = bundler;
 816                                 msg = buf_msg(buf);
 817                                 l_ptr->stats.sent_bundles++;
 818                         }
 819                 }
 820         }
 821         if (!l_ptr->next_out)
 822                 l_ptr->next_out = buf;
 823         link_add_to_outqueue(l_ptr, buf, msg);
 824         return dsz;
 825 }
 826
 827 /*
 828  * tipc_link_xmit(): same as __tipc_link_xmit(), but the link to use
 829  * has not been selected yet, and the the owner node is not locked
 830  * Called by TIPC internal users, e.g. the name distributor
 831  */
 832 int tipc_link_xmit(struct sk_buff *buf, u32 dest, u32 selector)
 833 {
 834         struct tipc_link *l_ptr;
 835         struct tipc_node *n_ptr;
 836         int res = -ELINKCONG;
 837
 838         n_ptr = tipc_node_find(dest);
 839         if (n_ptr) {
 840                 tipc_node_lock(n_ptr);
 841                 l_ptr = n_ptr->active_links[selector & 1];
 842                 if (l_ptr)
 843                         res = __tipc_link_xmit(l_ptr, buf);
 844                 else
 845                         kfree_skb(buf);
 846                 tipc_node_unlock(n_ptr);
 847         } else {
 848                 kfree_skb(buf);
 849         }
 850         return res;
 851 }
 852
 853 /*
 854  * tipc_link_sync_xmit - synchronize broadcast link endpoints.
 855  *
 856  * Give a newly added peer node the sequence number where it should
 857  * start receiving and acking broadcast packets.
 858  *
 859  * Called with node locked
 860  */
 861 static void tipc_link_sync_xmit(struct tipc_link *l)
 862 {
 863         struct sk_buff *buf;
 864         struct tipc_msg *msg;
 865
 866         buf = tipc_buf_acquire(INT_H_SIZE);
 867         if (!buf)
 868                 return;
 869
 870         msg = buf_msg(buf);
 871         tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE, l->addr);
 872         msg_set_last_bcast(msg, l->owner->bclink.acked);
 873         link_add_chain_to_outqueue(l, buf, 0);
 874         tipc_link_push_queue(l);
 875 }
 876
 877 /*
 878  * tipc_link_sync_rcv - synchronize broadcast link endpoints.
 879  * Receive the sequence number where we should start receiving and
 880  * acking broadcast packets from a newly added peer node, and open
 881  * up for reception of such packets.
 882  *
 883  * Called with node locked
 884  */
 885 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf)
 886 {
 887         struct tipc_msg *msg = buf_msg(buf);
 888
 889         n->bclink.last_sent = n->bclink.last_in = msg_last_bcast(msg);
 890         n->bclink.recv_permitted = true;
 891         kfree_skb(buf);
 892 }
 893
 894 /*
 895  * tipc_link_names_xmit - send name table entries to new neighbor
 896  *
 897  * Send routine for bulk delivery of name table messages when contact
 898  * with a new neighbor occurs. No link congestion checking is performed
 899  * because name table messages *must* be delivered. The messages must be
 900  * small enough not to require fragmentation.
 901  * Called without any locks held.
 902  */
 903 void tipc_link_names_xmit(struct list_head *message_list, u32 dest)
 904 {
 905         struct tipc_node *n_ptr;
 906         struct tipc_link *l_ptr;
 907         struct sk_buff *buf;
 908         struct sk_buff *temp_buf;
 909
 910         if (list_empty(message_list))
 911                 return;
 912
 913         n_ptr = tipc_node_find(dest);
 914         if (n_ptr) {
 915                 tipc_node_lock(n_ptr);
 916                 l_ptr = n_ptr->active_links[0];
 917                 if (l_ptr) {
 918                         /* convert circular list to linear list */
 919                         ((struct sk_buff *)message_list->prev)->next = NULL;
 920                         link_add_chain_to_outqueue(l_ptr,
 921                                 (struct sk_buff *)message_list->next, 0);
 922                         tipc_link_push_queue(l_ptr);
 923                         INIT_LIST_HEAD(message_list);
 924                 }
 925                 tipc_node_unlock(n_ptr);
 926         }
 927
 928         /* discard the messages if they couldn't be sent */
 929         list_for_each_safe(buf, temp_buf, ((struct sk_buff *)message_list)) {
 930                 list_del((struct list_head *)buf);
 931                 kfree_skb(buf);
 932         }
 933 }
 934
 935 /*
 936  * tipc_link_xmit_fast: Entry for data messages where the
 937  * destination link is known and the header is complete,
 938  * inclusive total message length. Very time critical.
 939  * Link is locked. Returns user data length.
 940  */
 941 static int tipc_link_xmit_fast(struct tipc_link *l_ptr, struct sk_buff *buf,
 942                                u32 *used_max_pkt)
 943 {
 944         struct tipc_msg *msg = buf_msg(buf);
 945         int res = msg_data_sz(msg);
 946
 947         if (likely(!link_congested(l_ptr))) {
 948                 if (likely(msg_size(msg) <= l_ptr->max_pkt)) {
 949                         link_add_to_outqueue(l_ptr, buf, msg);
 950                         tipc_bearer_send(l_ptr->bearer_id, buf,
 951                                          &l_ptr->media_addr);
 952                         l_ptr->unacked_window = 0;
 953                         return res;
 954                 }
 955                 else
 956                         *used_max_pkt = l_ptr->max_pkt;
 957         }
 958         return __tipc_link_xmit(l_ptr, buf);  /* All other cases */
 959 }
 960
 961 /*
 962  * tipc_link_iovec_xmit_fast: Entry for messages where the
 963  * destination processor is known and the header is complete,
 964  * except for total message length.
 965  * Returns user data length or errno.
 966  */
 967 int tipc_link_iovec_xmit_fast(struct tipc_port *sender,
 968                               struct iovec const *msg_sect,
 969                               unsigned int len, u32 destaddr)
 970 {
 971         struct tipc_msg *hdr = &sender->phdr;
 972         struct tipc_link *l_ptr;
 973         struct sk_buff *buf;
 974         struct tipc_node *node;
 975         int res;
 976         u32 selector = msg_origport(hdr) & 1;
 977
 978 again:
 979         /*
 980          * Try building message using port's max_pkt hint.
 981          * (Must not hold any locks while building message.)
 982          */
 983         res = tipc_msg_build(hdr, msg_sect, len, sender->max_pkt, &buf);
 984         /* Exit if build request was invalid */
 985         if (unlikely(res < 0))
 986                 return res;
 987
 988         node = tipc_node_find(destaddr);
 989         if (likely(node)) {
 990                 tipc_node_lock(node);
 991                 l_ptr = node->active_links[selector];
 992                 if (likely(l_ptr)) {
 993                         if (likely(buf)) {
 994                                 res = tipc_link_xmit_fast(l_ptr, buf,
 995                                                           &sender->max_pkt);
 996 exit:
 997                                 tipc_node_unlock(node);
 998                                 return res;
 999                         }
1000
1001                         /* Exit if link (or bearer) is congested */
1002                         if (link_congested(l_ptr)) {
1003                                 res = link_schedule_port(l_ptr,
1004                                                          sender->ref, res);
1005                                 goto exit;
1006                         }
1007
1008                         /*
1009                          * Message size exceeds max_pkt hint; update hint,
1010                          * then re-try fast path or fragment the message
1011                          */
1012                         sender->max_pkt = l_ptr->max_pkt;
1013                         tipc_node_unlock(node);
1014
1015
1016                         if ((msg_hdr_sz(hdr) + res) <= sender->max_pkt)
1017                                 goto again;
1018
1019                         return tipc_link_iovec_long_xmit(sender, msg_sect,
1020                                                          len, destaddr);
1021                 }
1022                 tipc_node_unlock(node);
1023         }
1024
1025         /* Couldn't find a link to the destination node */
1026         kfree_skb(buf);
1027         tipc_port_iovec_reject(sender, hdr, msg_sect, len, TIPC_ERR_NO_NODE);
1028         return -ENETUNREACH;
1029 }
1030
1031 /*
1032  * tipc_link_iovec_long_xmit(): Entry for long messages where the
1033  * destination node is known and the header is complete,
1034  * inclusive total message length.
1035  * Link and bearer congestion status have been checked to be ok,
1036  * and are ignored if they change.
1037  *
1038  * Note that fragments do not use the full link MTU so that they won't have
1039  * to undergo refragmentation if link changeover causes them to be sent
1040  * over another link with an additional tunnel header added as prefix.
1041  * (Refragmentation will still occur if the other link has a smaller MTU.)
1042  *
1043  * Returns user data length or errno.
1044  */
1045 static int tipc_link_iovec_long_xmit(struct tipc_port *sender,
1046                                      struct iovec const *msg_sect,
1047                                      unsigned int len, u32 destaddr)
1048 {
1049         struct tipc_link *l_ptr;
1050         struct tipc_node *node;
1051         struct tipc_msg *hdr = &sender->phdr;
1052         u32 dsz = len;
1053         u32 max_pkt, fragm_sz, rest;
1054         struct tipc_msg fragm_hdr;
1055         struct sk_buff *buf, *buf_chain, *prev;
1056         u32 fragm_crs, fragm_rest, hsz, sect_rest;
1057         const unchar __user *sect_crs;
1058         int curr_sect;
1059         u32 fragm_no;
1060         int res = 0;
1061
1062 again:
1063         fragm_no = 1;
1064         max_pkt = sender->max_pkt - INT_H_SIZE;
1065                 /* leave room for tunnel header in case of link changeover */
1066         fragm_sz = max_pkt - INT_H_SIZE;
1067                 /* leave room for fragmentation header in each fragment */
1068         rest = dsz;
1069         fragm_crs = 0;
1070         fragm_rest = 0;
1071         sect_rest = 0;
1072         sect_crs = NULL;
1073         curr_sect = -1;
1074
1075         /* Prepare reusable fragment header */
1076         tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1077                  INT_H_SIZE, msg_destnode(hdr));
1078         msg_set_size(&fragm_hdr, max_pkt);
1079         msg_set_fragm_no(&fragm_hdr, 1);
1080
1081         /* Prepare header of first fragment */
1082         buf_chain = buf = tipc_buf_acquire(max_pkt);
1083         if (!buf)
1084                 return -ENOMEM;
1085         buf->next = NULL;
1086         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1087         hsz = msg_hdr_sz(hdr);
1088         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1089
1090         /* Chop up message */
1091         fragm_crs = INT_H_SIZE + hsz;
1092         fragm_rest = fragm_sz - hsz;
1093
1094         do {            /* For all sections */
1095                 u32 sz;
1096
1097                 if (!sect_rest) {
1098                         sect_rest = msg_sect[++curr_sect].iov_len;
1099                         sect_crs = msg_sect[curr_sect].iov_base;
1100                 }
1101
1102                 if (sect_rest < fragm_rest)
1103                         sz = sect_rest;
1104                 else
1105                         sz = fragm_rest;
1106
1107                 if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1108                         res = -EFAULT;
1109 error:
1110                         kfree_skb_list(buf_chain);
1111                         return res;
1112                 }
1113                 sect_crs += sz;
1114                 sect_rest -= sz;
1115                 fragm_crs += sz;
1116                 fragm_rest -= sz;
1117                 rest -= sz;
1118
1119                 if (!fragm_rest && rest) {
1120
1121                         /* Initiate new fragment: */
1122                         if (rest <= fragm_sz) {
1123                                 fragm_sz = rest;
1124                                 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
1125                         } else {
1126                                 msg_set_type(&fragm_hdr, FRAGMENT);
1127                         }
1128                         msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1129                         msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1130                         prev = buf;
1131                         buf = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
1132                         if (!buf) {
1133                                 res = -ENOMEM;
1134                                 goto error;
1135                         }
1136
1137                         buf->next = NULL;
1138                         prev->next = buf;
1139                         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1140                         fragm_crs = INT_H_SIZE;
1141                         fragm_rest = fragm_sz;
1142                 }
1143         } while (rest > 0);
1144
1145         /*
1146          * Now we have a buffer chain. Select a link and check
1147          * that packet size is still OK
1148          */
1149         node = tipc_node_find(destaddr);
1150         if (likely(node)) {
1151                 tipc_node_lock(node);
1152                 l_ptr = node->active_links[sender->ref & 1];
1153                 if (!l_ptr) {
1154                         tipc_node_unlock(node);
1155                         goto reject;
1156                 }
1157                 if (l_ptr->max_pkt < max_pkt) {
1158                         sender->max_pkt = l_ptr->max_pkt;
1159                         tipc_node_unlock(node);
1160                         kfree_skb_list(buf_chain);
1161                         goto again;
1162                 }
1163         } else {
1164 reject:
1165                 kfree_skb_list(buf_chain);
1166                 tipc_port_iovec_reject(sender, hdr, msg_sect, len,
1167                                        TIPC_ERR_NO_NODE);
1168                 return -ENETUNREACH;
1169         }
1170
1171         /* Append chain of fragments to send queue & send them */
1172         l_ptr->long_msg_seq_no++;
1173         link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
1174         l_ptr->stats.sent_fragments += fragm_no;
1175         l_ptr->stats.sent_fragmented++;
1176         tipc_link_push_queue(l_ptr);
1177         tipc_node_unlock(node);
1178         return dsz;
1179 }
1180
1181 /*
1182  * tipc_link_push_packet: Push one unsent packet to the media
1183  */
1184 static u32 tipc_link_push_packet(struct tipc_link *l_ptr)
1185 {
1186         struct sk_buff *buf = l_ptr->first_out;
1187         u32 r_q_size = l_ptr->retransm_queue_size;
1188         u32 r_q_head = l_ptr->retransm_queue_head;
1189
1190         /* Step to position where retransmission failed, if any,    */
1191         /* consider that buffers may have been released in meantime */
1192         if (r_q_size && buf) {
1193                 u32 last = lesser(mod(r_q_head + r_q_size),
1194                                   link_last_sent(l_ptr));
1195                 u32 first = buf_seqno(buf);
1196
1197                 while (buf && less(first, r_q_head)) {
1198                         first = mod(first + 1);
1199                         buf = buf->next;
1200                 }
1201                 l_ptr->retransm_queue_head = r_q_head = first;
1202                 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1203         }
1204
1205         /* Continue retransmission now, if there is anything: */
1206         if (r_q_size && buf) {
1207                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1208                 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1209                 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
1210                 l_ptr->retransm_queue_head = mod(++r_q_head);
1211                 l_ptr->retransm_queue_size = --r_q_size;
1212                 l_ptr->stats.retransmitted++;
1213                 return 0;
1214         }
1215
1216         /* Send deferred protocol message, if any: */
1217         buf = l_ptr->proto_msg_queue;
1218         if (buf) {
1219                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1220                 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1221                 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
1222                 l_ptr->unacked_window = 0;
1223                 kfree_skb(buf);
1224                 l_ptr->proto_msg_queue = NULL;
1225                 return 0;
1226         }
1227
1228         /* Send one deferred data message, if send window not full: */
1229         buf = l_ptr->next_out;
1230         if (buf) {
1231                 struct tipc_msg *msg = buf_msg(buf);
1232                 u32 next = msg_seqno(msg);
1233                 u32 first = buf_seqno(l_ptr->first_out);
1234
1235                 if (mod(next - first) < l_ptr->queue_limit[0]) {
1236                         msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1237                         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1238                         tipc_bearer_send(l_ptr->bearer_id, buf,
1239                                          &l_ptr->media_addr);
1240                         if (msg_user(msg) == MSG_BUNDLER)
1241                                 msg_set_type(msg, CLOSED_MSG);
1242                         l_ptr->next_out = buf->next;
1243                         return 0;
1244                 }
1245         }
1246         return 1;
1247 }
1248
1249 /*
1250  * push_queue(): push out the unsent messages of a link where
1251  *               congestion has abated. Node is locked
1252  */
1253 void tipc_link_push_queue(struct tipc_link *l_ptr)
1254 {
1255         u32 res;
1256
1257         do {
1258                 res = tipc_link_push_packet(l_ptr);
1259         } while (!res);
1260 }
1261
1262 void tipc_link_reset_all(struct tipc_node *node)
1263 {
1264         char addr_string[16];
1265         u32 i;
1266
1267         tipc_node_lock(node);
1268
1269         pr_warn("Resetting all links to %s\n",
1270                 tipc_addr_string_fill(addr_string, node->addr));
1271
1272         for (i = 0; i < MAX_BEARERS; i++) {
1273                 if (node->links[i]) {
1274                         link_print(node->links[i], "Resetting link\n");
1275                         tipc_link_reset(node->links[i]);
1276                 }
1277         }
1278
1279         tipc_node_unlock(node);
1280 }
1281
1282 static void link_retransmit_failure(struct tipc_link *l_ptr,
1283                                     struct sk_buff *buf)
1284 {
1285         struct tipc_msg *msg = buf_msg(buf);
1286
1287         pr_warn("Retransmission failure on link <%s>\n", l_ptr->name);
1288
1289         if (l_ptr->addr) {
1290                 /* Handle failure on standard link */
1291                 link_print(l_ptr, "Resetting link\n");
1292                 tipc_link_reset(l_ptr);
1293
1294         } else {
1295                 /* Handle failure on broadcast link */
1296                 struct tipc_node *n_ptr;
1297                 char addr_string[16];
1298
1299                 pr_info("Msg seq number: %u,  ", msg_seqno(msg));
1300                 pr_cont("Outstanding acks: %lu\n",
1301                         (unsigned long) TIPC_SKB_CB(buf)->handle);
1302
1303                 n_ptr = tipc_bclink_retransmit_to();
1304                 tipc_node_lock(n_ptr);
1305
1306                 tipc_addr_string_fill(addr_string, n_ptr->addr);
1307                 pr_info("Broadcast link info for %s\n", addr_string);
1308                 pr_info("Reception permitted: %d,  Acked: %u\n",
1309                         n_ptr->bclink.recv_permitted,
1310                         n_ptr->bclink.acked);
1311                 pr_info("Last in: %u,  Oos state: %u,  Last sent: %u\n",
1312                         n_ptr->bclink.last_in,
1313                         n_ptr->bclink.oos_state,
1314                         n_ptr->bclink.last_sent);
1315
1316                 tipc_node_unlock(n_ptr);
1317
1318                 tipc_bclink_set_flags(TIPC_BCLINK_RESET);
1319                 l_ptr->stale_count = 0;
1320         }
1321 }
1322
1323 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *buf,
1324                           u32 retransmits)
1325 {
1326         struct tipc_msg *msg;
1327
1328         if (!buf)
1329                 return;
1330
1331         msg = buf_msg(buf);
1332
1333         /* Detect repeated retransmit failures */
1334         if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1335                 if (++l_ptr->stale_count > 100) {
1336                         link_retransmit_failure(l_ptr, buf);
1337                         return;
1338                 }
1339         } else {
1340                 l_ptr->last_retransmitted = msg_seqno(msg);
1341                 l_ptr->stale_count = 1;
1342         }
1343
1344         while (retransmits && (buf != l_ptr->next_out) && buf) {
1345                 msg = buf_msg(buf);
1346                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1347                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1348                 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
1349                 buf = buf->next;
1350                 retransmits--;
1351                 l_ptr->stats.retransmitted++;
1352         }
1353
1354         l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1355 }
1356
1357 /**
1358  * link_insert_deferred_queue - insert deferred messages back into receive chain
1359  */
1360 static struct sk_buff *link_insert_deferred_queue(struct tipc_link *l_ptr,
1361                                                   struct sk_buff *buf)
1362 {
1363         u32 seq_no;
1364
1365         if (l_ptr->oldest_deferred_in == NULL)
1366                 return buf;
1367
1368         seq_no = buf_seqno(l_ptr->oldest_deferred_in);
1369         if (seq_no == mod(l_ptr->next_in_no)) {
1370                 l_ptr->newest_deferred_in->next = buf;
1371                 buf = l_ptr->oldest_deferred_in;
1372                 l_ptr->oldest_deferred_in = NULL;
1373                 l_ptr->deferred_inqueue_sz = 0;
1374         }
1375         return buf;
1376 }
1377
1378 /**
1379  * link_recv_buf_validate - validate basic format of received message
1380  *
1381  * This routine ensures a TIPC message has an acceptable header, and at least
1382  * as much data as the header indicates it should.  The routine also ensures
1383  * that the entire message header is stored in the main fragment of the message
1384  * buffer, to simplify future access to message header fields.
1385  *
1386  * Note: Having extra info present in the message header or data areas is OK.
1387  * TIPC will ignore the excess, under the assumption that it is optional info
1388  * introduced by a later release of the protocol.
1389  */
1390 static int link_recv_buf_validate(struct sk_buff *buf)
1391 {
1392         static u32 min_data_hdr_size[8] = {
1393                 SHORT_H_SIZE, MCAST_H_SIZE, NAMED_H_SIZE, BASIC_H_SIZE,
1394                 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1395                 };
1396
1397         struct tipc_msg *msg;
1398         u32 tipc_hdr[2];
1399         u32 size;
1400         u32 hdr_size;
1401         u32 min_hdr_size;
1402
1403         /* If this packet comes from the defer queue, the skb has already
1404          * been validated
1405          */
1406         if (unlikely(TIPC_SKB_CB(buf)->deferred))
1407                 return 1;
1408
1409         if (unlikely(buf->len < MIN_H_SIZE))
1410                 return 0;
1411
1412         msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1413         if (msg == NULL)
1414                 return 0;
1415
1416         if (unlikely(msg_version(msg) != TIPC_VERSION))
1417                 return 0;
1418
1419         size = msg_size(msg);
1420         hdr_size = msg_hdr_sz(msg);
1421         min_hdr_size = msg_isdata(msg) ?
1422                 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1423
1424         if (unlikely((hdr_size < min_hdr_size) ||
1425                      (size < hdr_size) ||
1426                      (buf->len < size) ||
1427                      (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1428                 return 0;
1429
1430         return pskb_may_pull(buf, hdr_size);
1431 }
1432
1433 /**
1434  * tipc_rcv - process TIPC packets/messages arriving from off-node
1435  * @head: pointer to message buffer chain
1436  * @b_ptr: pointer to bearer message arrived on
1437  *
1438  * Invoked with no locks held.  Bearer pointer must point to a valid bearer
1439  * structure (i.e. cannot be NULL), but bearer can be inactive.
1440  */
1441 void tipc_rcv(struct sk_buff *head, struct tipc_bearer *b_ptr)
1442 {
1443         while (head) {
1444                 struct tipc_node *n_ptr;
1445                 struct tipc_link *l_ptr;
1446                 struct sk_buff *crs;
1447                 struct sk_buff *buf = head;
1448                 struct tipc_msg *msg;
1449                 u32 seq_no;
1450                 u32 ackd;
1451                 u32 released = 0;
1452
1453                 head = head->next;
1454                 buf->next = NULL;
1455
1456                 /* Ensure message is well-formed */
1457                 if (unlikely(!link_recv_buf_validate(buf)))
1458                         goto discard;
1459
1460                 /* Ensure message data is a single contiguous unit */
1461                 if (unlikely(skb_linearize(buf)))
1462                         goto discard;
1463
1464                 /* Handle arrival of a non-unicast link message */
1465                 msg = buf_msg(buf);
1466
1467                 if (unlikely(msg_non_seq(msg))) {
1468                         if (msg_user(msg) ==  LINK_CONFIG)
1469                                 tipc_disc_rcv(buf, b_ptr);
1470                         else
1471                                 tipc_bclink_rcv(buf);
1472                         continue;
1473                 }
1474
1475                 /* Discard unicast link messages destined for another node */
1476                 if (unlikely(!msg_short(msg) &&
1477                              (msg_destnode(msg) != tipc_own_addr)))
1478                         goto discard;
1479
1480                 /* Locate neighboring node that sent message */
1481                 n_ptr = tipc_node_find(msg_prevnode(msg));
1482                 if (unlikely(!n_ptr))
1483                         goto discard;
1484                 tipc_node_lock(n_ptr);
1485
1486                 /* Locate unicast link endpoint that should handle message */
1487                 l_ptr = n_ptr->links[b_ptr->identity];
1488                 if (unlikely(!l_ptr))
1489                         goto unlock_discard;
1490
1491                 /* Verify that communication with node is currently allowed */
1492                 if ((n_ptr->action_flags & TIPC_WAIT_PEER_LINKS_DOWN) &&
1493                     msg_user(msg) == LINK_PROTOCOL &&
1494                     (msg_type(msg) == RESET_MSG ||
1495                     msg_type(msg) == ACTIVATE_MSG) &&
1496                     !msg_redundant_link(msg))
1497                         n_ptr->action_flags &= ~TIPC_WAIT_PEER_LINKS_DOWN;
1498
1499                 if (tipc_node_blocked(n_ptr))
1500                         goto unlock_discard;
1501
1502                 /* Validate message sequence number info */
1503                 seq_no = msg_seqno(msg);
1504                 ackd = msg_ack(msg);
1505
1506                 /* Release acked messages */
1507                 if (n_ptr->bclink.recv_permitted)
1508                         tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1509
1510                 crs = l_ptr->first_out;
1511                 while ((crs != l_ptr->next_out) &&
1512                        less_eq(buf_seqno(crs), ackd)) {
1513                         struct sk_buff *next = crs->next;
1514                         kfree_skb(crs);
1515                         crs = next;
1516                         released++;
1517                 }
1518                 if (released) {
1519                         l_ptr->first_out = crs;
1520                         l_ptr->out_queue_size -= released;
1521                 }
1522
1523                 /* Try sending any messages link endpoint has pending */
1524                 if (unlikely(l_ptr->next_out))
1525                         tipc_link_push_queue(l_ptr);
1526
1527                 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1528                         tipc_link_wakeup_ports(l_ptr, 0);
1529
1530                 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1531                         l_ptr->stats.sent_acks++;
1532                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1533                 }
1534
1535                 /* Process the incoming packet */
1536                 if (unlikely(!link_working_working(l_ptr))) {
1537                         if (msg_user(msg) == LINK_PROTOCOL) {
1538                                 tipc_link_proto_rcv(l_ptr, buf);
1539                                 head = link_insert_deferred_queue(l_ptr, head);
1540                                 tipc_node_unlock(n_ptr);
1541                                 continue;
1542                         }
1543
1544                         /* Traffic message. Conditionally activate link */
1545                         link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1546
1547                         if (link_working_working(l_ptr)) {
1548                                 /* Re-insert buffer in front of queue */
1549                                 buf->next = head;
1550                                 head = buf;
1551                                 tipc_node_unlock(n_ptr);
1552                                 continue;
1553                         }
1554                         goto unlock_discard;
1555                 }
1556
1557                 /* Link is now in state WORKING_WORKING */
1558                 if (unlikely(seq_no != mod(l_ptr->next_in_no))) {
1559                         link_handle_out_of_seq_msg(l_ptr, buf);
1560                         head = link_insert_deferred_queue(l_ptr, head);
1561                         tipc_node_unlock(n_ptr);
1562                         continue;
1563                 }
1564                 l_ptr->next_in_no++;
1565                 if (unlikely(l_ptr->oldest_deferred_in))
1566                         head = link_insert_deferred_queue(l_ptr, head);
1567
1568                 /* Deliver packet/message to correct user: */
1569                 if (unlikely(msg_user(msg) ==  CHANGEOVER_PROTOCOL)) {
1570                         if (!tipc_link_tunnel_rcv(n_ptr, &buf)) {
1571                                 tipc_node_unlock(n_ptr);
1572                                 continue;
1573                         }
1574                         msg = buf_msg(buf);
1575                 } else if (msg_user(msg) == MSG_FRAGMENTER) {
1576                         l_ptr->stats.recv_fragments++;
1577                         if (tipc_buf_append(&l_ptr->reasm_buf, &buf)) {
1578                                 l_ptr->stats.recv_fragmented++;
1579                                 msg = buf_msg(buf);
1580                         } else {
1581                                 if (!l_ptr->reasm_buf)
1582                                         tipc_link_reset(l_ptr);
1583                                 tipc_node_unlock(n_ptr);
1584                                 continue;
1585                         }
1586                 }
1587
1588                 switch (msg_user(msg)) {
1589                 case TIPC_LOW_IMPORTANCE:
1590                 case TIPC_MEDIUM_IMPORTANCE:
1591                 case TIPC_HIGH_IMPORTANCE:
1592                 case TIPC_CRITICAL_IMPORTANCE:
1593                         tipc_node_unlock(n_ptr);
1594                         tipc_sk_rcv(buf);
1595                         continue;
1596                 case MSG_BUNDLER:
1597                         l_ptr->stats.recv_bundles++;
1598                         l_ptr->stats.recv_bundled += msg_msgcnt(msg);
1599                         tipc_node_unlock(n_ptr);
1600                         tipc_link_bundle_rcv(buf);
1601                         continue;
1602                 case NAME_DISTRIBUTOR:
1603                         n_ptr->bclink.recv_permitted = true;
1604                         tipc_node_unlock(n_ptr);
1605                         tipc_named_rcv(buf);
1606                         continue;
1607                 case CONN_MANAGER:
1608                         tipc_node_unlock(n_ptr);
1609                         tipc_port_proto_rcv(buf);
1610                         continue;
1611                 case BCAST_PROTOCOL:
1612                         tipc_link_sync_rcv(n_ptr, buf);
1613                         break;
1614                 default:
1615                         kfree_skb(buf);
1616                         break;
1617                 }
1618                 tipc_node_unlock(n_ptr);
1619                 continue;
1620 unlock_discard:
1621                 tipc_node_unlock(n_ptr);
1622 discard:
1623                 kfree_skb(buf);
1624         }
1625 }
1626
1627 /**
1628  * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1629  *
1630  * Returns increase in queue length (i.e. 0 or 1)
1631  */
1632 u32 tipc_link_defer_pkt(struct sk_buff **head, struct sk_buff **tail,
1633                         struct sk_buff *buf)
1634 {
1635         struct sk_buff *queue_buf;
1636         struct sk_buff **prev;
1637         u32 seq_no = buf_seqno(buf);
1638
1639         buf->next = NULL;
1640
1641         /* Empty queue ? */
1642         if (*head == NULL) {
1643                 *head = *tail = buf;
1644                 return 1;
1645         }
1646
1647         /* Last ? */
1648         if (less(buf_seqno(*tail), seq_no)) {
1649                 (*tail)->next = buf;
1650                 *tail = buf;
1651                 return 1;
1652         }
1653
1654         /* Locate insertion point in queue, then insert; discard if duplicate */
1655         prev = head;
1656         queue_buf = *head;
1657         for (;;) {
1658                 u32 curr_seqno = buf_seqno(queue_buf);
1659
1660                 if (seq_no == curr_seqno) {
1661                         kfree_skb(buf);
1662                         return 0;
1663                 }
1664
1665                 if (less(seq_no, curr_seqno))
1666                         break;
1667
1668                 prev = &queue_buf->next;
1669                 queue_buf = queue_buf->next;
1670         }
1671
1672         buf->next = queue_buf;
1673         *prev = buf;
1674         return 1;
1675 }
1676
1677 /*
1678  * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
1679  */
1680 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
1681                                        struct sk_buff *buf)
1682 {
1683         u32 seq_no = buf_seqno(buf);
1684
1685         if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
1686                 tipc_link_proto_rcv(l_ptr, buf);
1687                 return;
1688         }
1689
1690         /* Record OOS packet arrival (force mismatch on next timeout) */
1691         l_ptr->checkpoint--;
1692
1693         /*
1694          * Discard packet if a duplicate; otherwise add it to deferred queue
1695          * and notify peer of gap as per protocol specification
1696          */
1697         if (less(seq_no, mod(l_ptr->next_in_no))) {
1698                 l_ptr->stats.duplicates++;
1699                 kfree_skb(buf);
1700                 return;
1701         }
1702
1703         if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
1704                                 &l_ptr->newest_deferred_in, buf)) {
1705                 l_ptr->deferred_inqueue_sz++;
1706                 l_ptr->stats.deferred_recv++;
1707                 TIPC_SKB_CB(buf)->deferred = true;
1708                 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
1709                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1710         } else
1711                 l_ptr->stats.duplicates++;
1712 }
1713
1714 /*
1715  * Send protocol message to the other endpoint.
1716  */
1717 void tipc_link_proto_xmit(struct tipc_link *l_ptr, u32 msg_typ, int probe_msg,
1718                           u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
1719 {
1720         struct sk_buff *buf = NULL;
1721         struct tipc_msg *msg = l_ptr->pmsg;
1722         u32 msg_size = sizeof(l_ptr->proto_msg);
1723         int r_flag;
1724
1725         /* Discard any previous message that was deferred due to congestion */
1726         if (l_ptr->proto_msg_queue) {
1727                 kfree_skb(l_ptr->proto_msg_queue);
1728                 l_ptr->proto_msg_queue = NULL;
1729         }
1730
1731         /* Don't send protocol message during link changeover */
1732         if (l_ptr->exp_msg_count)
1733                 return;
1734
1735         /* Abort non-RESET send if communication with node is prohibited */
1736         if ((tipc_node_blocked(l_ptr->owner)) && (msg_typ != RESET_MSG))
1737                 return;
1738
1739         /* Create protocol message with "out-of-sequence" sequence number */
1740         msg_set_type(msg, msg_typ);
1741         msg_set_net_plane(msg, l_ptr->net_plane);
1742         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1743         msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
1744
1745         if (msg_typ == STATE_MSG) {
1746                 u32 next_sent = mod(l_ptr->next_out_no);
1747
1748                 if (!tipc_link_is_up(l_ptr))
1749                         return;
1750                 if (l_ptr->next_out)
1751                         next_sent = buf_seqno(l_ptr->next_out);
1752                 msg_set_next_sent(msg, next_sent);
1753                 if (l_ptr->oldest_deferred_in) {
1754                         u32 rec = buf_seqno(l_ptr->oldest_deferred_in);
1755                         gap = mod(rec - mod(l_ptr->next_in_no));
1756                 }
1757                 msg_set_seq_gap(msg, gap);
1758                 if (gap)
1759                         l_ptr->stats.sent_nacks++;
1760                 msg_set_link_tolerance(msg, tolerance);
1761                 msg_set_linkprio(msg, priority);
1762                 msg_set_max_pkt(msg, ack_mtu);
1763                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1764                 msg_set_probe(msg, probe_msg != 0);
1765                 if (probe_msg) {
1766                         u32 mtu = l_ptr->max_pkt;
1767
1768                         if ((mtu < l_ptr->max_pkt_target) &&
1769                             link_working_working(l_ptr) &&
1770                             l_ptr->fsm_msg_cnt) {
1771                                 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1772                                 if (l_ptr->max_pkt_probes == 10) {
1773                                         l_ptr->max_pkt_target = (msg_size - 4);
1774                                         l_ptr->max_pkt_probes = 0;
1775                                         msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1776                                 }
1777                                 l_ptr->max_pkt_probes++;
1778                         }
1779
1780                         l_ptr->stats.sent_probes++;
1781                 }
1782                 l_ptr->stats.sent_states++;
1783         } else {                /* RESET_MSG or ACTIVATE_MSG */
1784                 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
1785                 msg_set_seq_gap(msg, 0);
1786                 msg_set_next_sent(msg, 1);
1787                 msg_set_probe(msg, 0);
1788                 msg_set_link_tolerance(msg, l_ptr->tolerance);
1789                 msg_set_linkprio(msg, l_ptr->priority);
1790                 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
1791         }
1792
1793         r_flag = (l_ptr->owner->working_links > tipc_link_is_up(l_ptr));
1794         msg_set_redundant_link(msg, r_flag);
1795         msg_set_linkprio(msg, l_ptr->priority);
1796         msg_set_size(msg, msg_size);
1797
1798         msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
1799
1800         buf = tipc_buf_acquire(msg_size);
1801         if (!buf)
1802                 return;
1803
1804         skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1805         buf->priority = TC_PRIO_CONTROL;
1806
1807         tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
1808         l_ptr->unacked_window = 0;
1809         kfree_skb(buf);
1810 }
1811
1812 /*
1813  * Receive protocol message :
1814  * Note that network plane id propagates through the network, and may
1815  * change at any time. The node with lowest address rules
1816  */
1817 static void tipc_link_proto_rcv(struct tipc_link *l_ptr, struct sk_buff *buf)
1818 {
1819         u32 rec_gap = 0;
1820         u32 max_pkt_info;
1821         u32 max_pkt_ack;
1822         u32 msg_tol;
1823         struct tipc_msg *msg = buf_msg(buf);
1824
1825         /* Discard protocol message during link changeover */
1826         if (l_ptr->exp_msg_count)
1827                 goto exit;
1828
1829         if (l_ptr->net_plane != msg_net_plane(msg))
1830                 if (tipc_own_addr > msg_prevnode(msg))
1831                         l_ptr->net_plane = msg_net_plane(msg);
1832
1833         switch (msg_type(msg)) {
1834
1835         case RESET_MSG:
1836                 if (!link_working_unknown(l_ptr) &&
1837                     (l_ptr->peer_session != INVALID_SESSION)) {
1838                         if (less_eq(msg_session(msg), l_ptr->peer_session))
1839                                 break; /* duplicate or old reset: ignore */
1840                 }
1841
1842                 if (!msg_redundant_link(msg) && (link_working_working(l_ptr) ||
1843                                 link_working_unknown(l_ptr))) {
1844                         /*
1845                          * peer has lost contact -- don't allow peer's links
1846                          * to reactivate before we recognize loss & clean up
1847                          */
1848                         l_ptr->owner->action_flags |= TIPC_WAIT_OWN_LINKS_DOWN;
1849                 }
1850
1851                 link_state_event(l_ptr, RESET_MSG);
1852
1853                 /* fall thru' */
1854         case ACTIVATE_MSG:
1855                 /* Update link settings according other endpoint's values */
1856                 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
1857
1858                 msg_tol = msg_link_tolerance(msg);
1859                 if (msg_tol > l_ptr->tolerance)
1860                         link_set_supervision_props(l_ptr, msg_tol);
1861
1862                 if (msg_linkprio(msg) > l_ptr->priority)
1863                         l_ptr->priority = msg_linkprio(msg);
1864
1865                 max_pkt_info = msg_max_pkt(msg);
1866                 if (max_pkt_info) {
1867                         if (max_pkt_info < l_ptr->max_pkt_target)
1868                                 l_ptr->max_pkt_target = max_pkt_info;
1869                         if (l_ptr->max_pkt > l_ptr->max_pkt_target)
1870                                 l_ptr->max_pkt = l_ptr->max_pkt_target;
1871                 } else {
1872                         l_ptr->max_pkt = l_ptr->max_pkt_target;
1873                 }
1874
1875                 /* Synchronize broadcast link info, if not done previously */
1876                 if (!tipc_node_is_up(l_ptr->owner)) {
1877                         l_ptr->owner->bclink.last_sent =
1878                                 l_ptr->owner->bclink.last_in =
1879                                 msg_last_bcast(msg);
1880                         l_ptr->owner->bclink.oos_state = 0;
1881                 }
1882
1883                 l_ptr->peer_session = msg_session(msg);
1884                 l_ptr->peer_bearer_id = msg_bearer_id(msg);
1885
1886                 if (msg_type(msg) == ACTIVATE_MSG)
1887                         link_state_event(l_ptr, ACTIVATE_MSG);
1888                 break;
1889         case STATE_MSG:
1890
1891                 msg_tol = msg_link_tolerance(msg);
1892                 if (msg_tol)
1893                         link_set_supervision_props(l_ptr, msg_tol);
1894
1895                 if (msg_linkprio(msg) &&
1896                     (msg_linkprio(msg) != l_ptr->priority)) {
1897                         pr_warn("%s<%s>, priority change %u->%u\n",
1898                                 link_rst_msg, l_ptr->name, l_ptr->priority,
1899                                 msg_linkprio(msg));
1900                         l_ptr->priority = msg_linkprio(msg);
1901                         tipc_link_reset(l_ptr); /* Enforce change to take effect */
1902                         break;
1903                 }
1904
1905                 /* Record reception; force mismatch at next timeout: */
1906                 l_ptr->checkpoint--;
1907
1908                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1909                 l_ptr->stats.recv_states++;
1910                 if (link_reset_unknown(l_ptr))
1911                         break;
1912
1913                 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
1914                         rec_gap = mod(msg_next_sent(msg) -
1915                                       mod(l_ptr->next_in_no));
1916                 }
1917
1918                 max_pkt_ack = msg_max_pkt(msg);
1919                 if (max_pkt_ack > l_ptr->max_pkt) {
1920                         l_ptr->max_pkt = max_pkt_ack;
1921                         l_ptr->max_pkt_probes = 0;
1922                 }
1923
1924                 max_pkt_ack = 0;
1925                 if (msg_probe(msg)) {
1926                         l_ptr->stats.recv_probes++;
1927                         if (msg_size(msg) > sizeof(l_ptr->proto_msg))
1928                                 max_pkt_ack = msg_size(msg);
1929                 }
1930
1931                 /* Protocol message before retransmits, reduce loss risk */
1932                 if (l_ptr->owner->bclink.recv_permitted)
1933                         tipc_bclink_update_link_state(l_ptr->owner,
1934                                                       msg_last_bcast(msg));
1935
1936                 if (rec_gap || (msg_probe(msg))) {
1937                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, rec_gap, 0,
1938                                              0, max_pkt_ack);
1939                 }
1940                 if (msg_seq_gap(msg)) {
1941                         l_ptr->stats.recv_nacks++;
1942                         tipc_link_retransmit(l_ptr, l_ptr->first_out,
1943                                              msg_seq_gap(msg));
1944                 }
1945                 break;
1946         }
1947 exit:
1948         kfree_skb(buf);
1949 }
1950
1951
1952 /* tipc_link_tunnel_xmit(): Tunnel one packet via a link belonging to
1953  * a different bearer. Owner node is locked.
1954  */
1955 static void tipc_link_tunnel_xmit(struct tipc_link *l_ptr,
1956                                   struct tipc_msg *tunnel_hdr,
1957                                   struct tipc_msg *msg,
1958                                   u32 selector)
1959 {
1960         struct tipc_link *tunnel;
1961         struct sk_buff *buf;
1962         u32 length = msg_size(msg);
1963
1964         tunnel = l_ptr->owner->active_links[selector & 1];
1965         if (!tipc_link_is_up(tunnel)) {
1966                 pr_warn("%stunnel link no longer available\n", link_co_err);
1967                 return;
1968         }
1969         msg_set_size(tunnel_hdr, length + INT_H_SIZE);
1970         buf = tipc_buf_acquire(length + INT_H_SIZE);
1971         if (!buf) {
1972                 pr_warn("%sunable to send tunnel msg\n", link_co_err);
1973                 return;
1974         }
1975         skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
1976         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
1977         __tipc_link_xmit(tunnel, buf);
1978 }
1979
1980
1981 /* tipc_link_failover_send_queue(): A link has gone down, but a second
1982  * link is still active. We can do failover. Tunnel the failing link's
1983  * whole send queue via the remaining link. This way, we don't lose
1984  * any packets, and sequence order is preserved for subsequent traffic
1985  * sent over the remaining link. Owner node is locked.
1986  */
1987 void tipc_link_failover_send_queue(struct tipc_link *l_ptr)
1988 {
1989         u32 msgcount = l_ptr->out_queue_size;
1990         struct sk_buff *crs = l_ptr->first_out;
1991         struct tipc_link *tunnel = l_ptr->owner->active_links[0];
1992         struct tipc_msg tunnel_hdr;
1993         int split_bundles;
1994
1995         if (!tunnel)
1996                 return;
1997
1998         tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
1999                  ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
2000         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2001         msg_set_msgcnt(&tunnel_hdr, msgcount);
2002
2003         if (!l_ptr->first_out) {
2004                 struct sk_buff *buf;
2005
2006                 buf = tipc_buf_acquire(INT_H_SIZE);
2007                 if (buf) {
2008                         skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2009                         msg_set_size(&tunnel_hdr, INT_H_SIZE);
2010                         __tipc_link_xmit(tunnel, buf);
2011                 } else {
2012                         pr_warn("%sunable to send changeover msg\n",
2013                                 link_co_err);
2014                 }
2015                 return;
2016         }
2017
2018         split_bundles = (l_ptr->owner->active_links[0] !=
2019                          l_ptr->owner->active_links[1]);
2020
2021         while (crs) {
2022                 struct tipc_msg *msg = buf_msg(crs);
2023
2024                 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2025                         struct tipc_msg *m = msg_get_wrapped(msg);
2026                         unchar *pos = (unchar *)m;
2027
2028                         msgcount = msg_msgcnt(msg);
2029                         while (msgcount--) {
2030                                 msg_set_seqno(m, msg_seqno(msg));
2031                                 tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, m,
2032                                                       msg_link_selector(m));
2033                                 pos += align(msg_size(m));
2034                                 m = (struct tipc_msg *)pos;
2035                         }
2036                 } else {
2037                         tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, msg,
2038                                               msg_link_selector(msg));
2039                 }
2040                 crs = crs->next;
2041         }
2042 }
2043
2044 /* tipc_link_dup_queue_xmit(): A second link has become active. Tunnel a
2045  * duplicate of the first link's send queue via the new link. This way, we
2046  * are guaranteed that currently queued packets from a socket are delivered
2047  * before future traffic from the same socket, even if this is using the
2048  * new link. The last arriving copy of each duplicate packet is dropped at
2049  * the receiving end by the regular protocol check, so packet cardinality
2050  * and sequence order is preserved per sender/receiver socket pair.
2051  * Owner node is locked.
2052  */
2053 void tipc_link_dup_queue_xmit(struct tipc_link *l_ptr,
2054                               struct tipc_link *tunnel)
2055 {
2056         struct sk_buff *iter;
2057         struct tipc_msg tunnel_hdr;
2058
2059         tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2060                  DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
2061         msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2062         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2063         iter = l_ptr->first_out;
2064         while (iter) {
2065                 struct sk_buff *outbuf;
2066                 struct tipc_msg *msg = buf_msg(iter);
2067                 u32 length = msg_size(msg);
2068
2069                 if (msg_user(msg) == MSG_BUNDLER)
2070                         msg_set_type(msg, CLOSED_MSG);
2071                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));   /* Update */
2072                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2073                 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2074                 outbuf = tipc_buf_acquire(length + INT_H_SIZE);
2075                 if (outbuf == NULL) {
2076                         pr_warn("%sunable to send duplicate msg\n",
2077                                 link_co_err);
2078                         return;
2079                 }
2080                 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2081                 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2082                                                length);
2083                 __tipc_link_xmit(tunnel, outbuf);
2084                 if (!tipc_link_is_up(l_ptr))
2085                         return;
2086                 iter = iter->next;
2087         }
2088 }
2089
2090 /**
2091  * buf_extract - extracts embedded TIPC message from another message
2092  * @skb: encapsulating message buffer
2093  * @from_pos: offset to extract from
2094  *
2095  * Returns a new message buffer containing an embedded message.  The
2096  * encapsulating message itself is left unchanged.
2097  */
2098 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2099 {
2100         struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2101         u32 size = msg_size(msg);
2102         struct sk_buff *eb;
2103
2104         eb = tipc_buf_acquire(size);
2105         if (eb)
2106                 skb_copy_to_linear_data(eb, msg, size);
2107         return eb;
2108 }
2109
2110
2111
2112 /* tipc_link_dup_rcv(): Receive a tunnelled DUPLICATE_MSG packet.
2113  * Owner node is locked.
2114  */
2115 static void tipc_link_dup_rcv(struct tipc_link *l_ptr,
2116                               struct sk_buff *t_buf)
2117 {
2118         struct sk_buff *buf;
2119
2120         if (!tipc_link_is_up(l_ptr))
2121                 return;
2122
2123         buf = buf_extract(t_buf, INT_H_SIZE);
2124         if (buf == NULL) {
2125                 pr_warn("%sfailed to extract inner dup pkt\n", link_co_err);
2126                 return;
2127         }
2128
2129         /* Add buffer to deferred queue, if applicable: */
2130         link_handle_out_of_seq_msg(l_ptr, buf);
2131 }
2132
2133 /*  tipc_link_failover_rcv(): Receive a tunnelled ORIGINAL_MSG packet
2134  *  Owner node is locked.
2135  */
2136 static struct sk_buff *tipc_link_failover_rcv(struct tipc_link *l_ptr,
2137                                               struct sk_buff *t_buf)
2138 {
2139         struct tipc_msg *t_msg = buf_msg(t_buf);
2140         struct sk_buff *buf = NULL;
2141         struct tipc_msg *msg;
2142
2143         if (tipc_link_is_up(l_ptr))
2144                 tipc_link_reset(l_ptr);
2145
2146         /* First failover packet? */
2147         if (l_ptr->exp_msg_count == START_CHANGEOVER)
2148                 l_ptr->exp_msg_count = msg_msgcnt(t_msg);
2149
2150         /* Should there be an inner packet? */
2151         if (l_ptr->exp_msg_count) {
2152                 l_ptr->exp_msg_count--;
2153                 buf = buf_extract(t_buf, INT_H_SIZE);
2154                 if (buf == NULL) {
2155                         pr_warn("%sno inner failover pkt\n", link_co_err);
2156                         goto exit;
2157                 }
2158                 msg = buf_msg(buf);
2159
2160                 if (less(msg_seqno(msg), l_ptr->reset_checkpoint)) {
2161                         kfree_skb(buf);
2162                         buf = NULL;
2163                         goto exit;
2164                 }
2165                 if (msg_user(msg) == MSG_FRAGMENTER) {
2166                         l_ptr->stats.recv_fragments++;
2167                         tipc_buf_append(&l_ptr->reasm_buf, &buf);
2168                 }
2169         }
2170 exit:
2171         if ((l_ptr->exp_msg_count == 0) && (l_ptr->flags & LINK_STOPPED)) {
2172                 tipc_node_detach_link(l_ptr->owner, l_ptr);
2173                 kfree(l_ptr);
2174         }
2175         return buf;
2176 }
2177
2178 /*  tipc_link_tunnel_rcv(): Receive a tunnelled packet, sent
2179  *  via other link as result of a failover (ORIGINAL_MSG) or
2180  *  a new active link (DUPLICATE_MSG). Failover packets are
2181  *  returned to the active link for delivery upwards.
2182  *  Owner node is locked.
2183  */
2184 static int tipc_link_tunnel_rcv(struct tipc_node *n_ptr,
2185                                 struct sk_buff **buf)
2186 {
2187         struct sk_buff *t_buf = *buf;
2188         struct tipc_link *l_ptr;
2189         struct tipc_msg *t_msg = buf_msg(t_buf);
2190         u32 bearer_id = msg_bearer_id(t_msg);
2191
2192         *buf = NULL;
2193
2194         if (bearer_id >= MAX_BEARERS)
2195                 goto exit;
2196
2197         l_ptr = n_ptr->links[bearer_id];
2198         if (!l_ptr)
2199                 goto exit;
2200
2201         if (msg_type(t_msg) == DUPLICATE_MSG)
2202                 tipc_link_dup_rcv(l_ptr, t_buf);
2203         else if (msg_type(t_msg) == ORIGINAL_MSG)
2204                 *buf = tipc_link_failover_rcv(l_ptr, t_buf);
2205         else
2206                 pr_warn("%sunknown tunnel pkt received\n", link_co_err);
2207 exit:
2208         kfree_skb(t_buf);
2209         return *buf != NULL;
2210 }
2211
2212 /*
2213  *  Bundler functionality:
2214  */
2215 void tipc_link_bundle_rcv(struct sk_buff *buf)
2216 {
2217         u32 msgcount = msg_msgcnt(buf_msg(buf));
2218         u32 pos = INT_H_SIZE;
2219         struct sk_buff *obuf;
2220
2221         while (msgcount--) {
2222                 obuf = buf_extract(buf, pos);
2223                 if (obuf == NULL) {
2224                         pr_warn("Link unable to unbundle message(s)\n");
2225                         break;
2226                 }
2227                 pos += align(msg_size(buf_msg(obuf)));
2228                 tipc_net_route_msg(obuf);
2229         }
2230         kfree_skb(buf);
2231 }
2232
2233 /*
2234  *  Fragmentation/defragmentation:
2235  */
2236
2237 /*
2238  * tipc_link_frag_xmit: Entry for buffers needing fragmentation.
2239  * The buffer is complete, inclusive total message length.
2240  * Returns user data length.
2241  */
2242 static int tipc_link_frag_xmit(struct tipc_link *l_ptr, struct sk_buff *buf)
2243 {
2244         struct sk_buff *buf_chain = NULL;
2245         struct sk_buff *buf_chain_tail = (struct sk_buff *)&buf_chain;
2246         struct tipc_msg *inmsg = buf_msg(buf);
2247         struct tipc_msg fragm_hdr;
2248         u32 insize = msg_size(inmsg);
2249         u32 dsz = msg_data_sz(inmsg);
2250         unchar *crs = buf->data;
2251         u32 rest = insize;
2252         u32 pack_sz = l_ptr->max_pkt;
2253         u32 fragm_sz = pack_sz - INT_H_SIZE;
2254         u32 fragm_no = 0;
2255         u32 destaddr;
2256
2257         if (msg_short(inmsg))
2258                 destaddr = l_ptr->addr;
2259         else
2260                 destaddr = msg_destnode(inmsg);
2261
2262         /* Prepare reusable fragment header: */
2263         tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2264                  INT_H_SIZE, destaddr);
2265
2266         /* Chop up message: */
2267         while (rest > 0) {
2268                 struct sk_buff *fragm;
2269
2270                 if (rest <= fragm_sz) {
2271                         fragm_sz = rest;
2272                         msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2273                 }
2274                 fragm = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
2275                 if (fragm == NULL) {
2276                         kfree_skb(buf);
2277                         kfree_skb_list(buf_chain);
2278                         return -ENOMEM;
2279                 }
2280                 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2281                 fragm_no++;
2282                 msg_set_fragm_no(&fragm_hdr, fragm_no);
2283                 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2284                 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2285                                                fragm_sz);
2286                 buf_chain_tail->next = fragm;
2287                 buf_chain_tail = fragm;
2288
2289                 rest -= fragm_sz;
2290                 crs += fragm_sz;
2291                 msg_set_type(&fragm_hdr, FRAGMENT);
2292         }
2293         kfree_skb(buf);
2294
2295         /* Append chain of fragments to send queue & send them */
2296         l_ptr->long_msg_seq_no++;
2297         link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
2298         l_ptr->stats.sent_fragments += fragm_no;
2299         l_ptr->stats.sent_fragmented++;
2300         tipc_link_push_queue(l_ptr);
2301
2302         return dsz;
2303 }
2304
2305 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance)
2306 {
2307         if ((tolerance < TIPC_MIN_LINK_TOL) || (tolerance > TIPC_MAX_LINK_TOL))
2308                 return;
2309
2310         l_ptr->tolerance = tolerance;
2311         l_ptr->continuity_interval =
2312                 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2313         l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2314 }
2315
2316 void tipc_link_set_queue_limits(struct tipc_link *l_ptr, u32 window)
2317 {
2318         /* Data messages from this node, inclusive FIRST_FRAGM */
2319         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2320         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2321         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2322         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2323         /* Transiting data messages,inclusive FIRST_FRAGM */
2324         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2325         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2326         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2327         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2328         l_ptr->queue_limit[CONN_MANAGER] = 1200;
2329         l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2330         l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2331         /* FRAGMENT and LAST_FRAGMENT packets */
2332         l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2333 }
2334
2335 /* tipc_link_find_owner - locate owner node of link by link's name
2336  * @name: pointer to link name string
2337  * @bearer_id: pointer to index in 'node->links' array where the link was found.
2338  *
2339  * Returns pointer to node owning the link, or 0 if no matching link is found.
2340  */
2341 static struct tipc_node *tipc_link_find_owner(const char *link_name,
2342                                               unsigned int *bearer_id)
2343 {
2344         struct tipc_link *l_ptr;
2345         struct tipc_node *n_ptr;
2346         struct tipc_node *found_node = 0;
2347         int i;
2348
2349         *bearer_id = 0;
2350         rcu_read_lock();
2351         list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
2352                 tipc_node_lock(n_ptr);
2353                 for (i = 0; i < MAX_BEARERS; i++) {
2354                         l_ptr = n_ptr->links[i];
2355                         if (l_ptr && !strcmp(l_ptr->name, link_name)) {
2356                                 *bearer_id = i;
2357                                 found_node = n_ptr;
2358                                 break;
2359                         }
2360                 }
2361                 tipc_node_unlock(n_ptr);
2362                 if (found_node)
2363                         break;
2364         }
2365         rcu_read_unlock();
2366
2367         return found_node;
2368 }
2369
2370 /**
2371  * link_value_is_valid -- validate proposed link tolerance/priority/window
2372  *
2373  * @cmd: value type (TIPC_CMD_SET_LINK_*)
2374  * @new_value: the new value
2375  *
2376  * Returns 1 if value is within range, 0 if not.
2377  */
2378 static int link_value_is_valid(u16 cmd, u32 new_value)
2379 {
2380         switch (cmd) {
2381         case TIPC_CMD_SET_LINK_TOL:
2382                 return (new_value >= TIPC_MIN_LINK_TOL) &&
2383                         (new_value <= TIPC_MAX_LINK_TOL);
2384         case TIPC_CMD_SET_LINK_PRI:
2385                 return (new_value <= TIPC_MAX_LINK_PRI);
2386         case TIPC_CMD_SET_LINK_WINDOW:
2387                 return (new_value >= TIPC_MIN_LINK_WIN) &&
2388                         (new_value <= TIPC_MAX_LINK_WIN);
2389         }
2390         return 0;
2391 }
2392
2393 /**
2394  * link_cmd_set_value - change priority/tolerance/window for link/bearer/media
2395  * @name: ptr to link, bearer, or media name
2396  * @new_value: new value of link, bearer, or media setting
2397  * @cmd: which link, bearer, or media attribute to set (TIPC_CMD_SET_LINK_*)
2398  *
2399  * Caller must hold RTNL lock to ensure link/bearer/media is not deleted.
2400  *
2401  * Returns 0 if value updated and negative value on error.
2402  */
2403 static int link_cmd_set_value(const char *name, u32 new_value, u16 cmd)
2404 {
2405         struct tipc_node *node;
2406         struct tipc_link *l_ptr;
2407         struct tipc_bearer *b_ptr;
2408         struct tipc_media *m_ptr;
2409         int bearer_id;
2410         int res = 0;
2411
2412         node = tipc_link_find_owner(name, &bearer_id);
2413         if (node) {
2414                 tipc_node_lock(node);
2415                 l_ptr = node->links[bearer_id];
2416
2417                 if (l_ptr) {
2418                         switch (cmd) {
2419                         case TIPC_CMD_SET_LINK_TOL:
2420                                 link_set_supervision_props(l_ptr, new_value);
2421                                 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0,
2422                                                      new_value, 0, 0);
2423                                 break;
2424                         case TIPC_CMD_SET_LINK_PRI:
2425                                 l_ptr->priority = new_value;
2426                                 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0,
2427                                                      0, new_value, 0);
2428                                 break;
2429                         case TIPC_CMD_SET_LINK_WINDOW:
2430                                 tipc_link_set_queue_limits(l_ptr, new_value);
2431                                 break;
2432                         default:
2433                                 res = -EINVAL;
2434                                 break;
2435                         }
2436                 }
2437                 tipc_node_unlock(node);
2438                 return res;
2439         }
2440
2441         b_ptr = tipc_bearer_find(name);
2442         if (b_ptr) {
2443                 switch (cmd) {
2444                 case TIPC_CMD_SET_LINK_TOL:
2445                         b_ptr->tolerance = new_value;
2446                         break;
2447                 case TIPC_CMD_SET_LINK_PRI:
2448                         b_ptr->priority = new_value;
2449                         break;
2450                 case TIPC_CMD_SET_LINK_WINDOW:
2451                         b_ptr->window = new_value;
2452                         break;
2453                 default:
2454                         res = -EINVAL;
2455                         break;
2456                 }
2457                 return res;
2458         }
2459
2460         m_ptr = tipc_media_find(name);
2461         if (!m_ptr)
2462                 return -ENODEV;
2463         switch (cmd) {
2464         case TIPC_CMD_SET_LINK_TOL:
2465                 m_ptr->tolerance = new_value;
2466                 break;
2467         case TIPC_CMD_SET_LINK_PRI:
2468                 m_ptr->priority = new_value;
2469                 break;
2470         case TIPC_CMD_SET_LINK_WINDOW:
2471                 m_ptr->window = new_value;
2472                 break;
2473         default:
2474                 res = -EINVAL;
2475                 break;
2476         }
2477         return res;
2478 }
2479
2480 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2481                                      u16 cmd)
2482 {
2483         struct tipc_link_config *args;
2484         u32 new_value;
2485         int res;
2486
2487         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2488                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2489
2490         args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2491         new_value = ntohl(args->value);
2492
2493         if (!link_value_is_valid(cmd, new_value))
2494                 return tipc_cfg_reply_error_string(
2495                         "cannot change, value invalid");
2496
2497         if (!strcmp(args->name, tipc_bclink_name)) {
2498                 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2499                     (tipc_bclink_set_queue_limits(new_value) == 0))
2500                         return tipc_cfg_reply_none();
2501                 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2502                                                    " (cannot change setting on broadcast link)");
2503         }
2504
2505         res = link_cmd_set_value(args->name, new_value, cmd);
2506         if (res)
2507                 return tipc_cfg_reply_error_string("cannot change link setting");
2508
2509         return tipc_cfg_reply_none();
2510 }
2511
2512 /**
2513  * link_reset_statistics - reset link statistics
2514  * @l_ptr: pointer to link
2515  */
2516 static void link_reset_statistics(struct tipc_link *l_ptr)
2517 {
2518         memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2519         l_ptr->stats.sent_info = l_ptr->next_out_no;
2520         l_ptr->stats.recv_info = l_ptr->next_in_no;
2521 }
2522
2523 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2524 {
2525         char *link_name;
2526         struct tipc_link *l_ptr;
2527         struct tipc_node *node;
2528         unsigned int bearer_id;
2529
2530         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2531                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2532
2533         link_name = (char *)TLV_DATA(req_tlv_area);
2534         if (!strcmp(link_name, tipc_bclink_name)) {
2535                 if (tipc_bclink_reset_stats())
2536                         return tipc_cfg_reply_error_string("link not found");
2537                 return tipc_cfg_reply_none();
2538         }
2539         node = tipc_link_find_owner(link_name, &bearer_id);
2540         if (!node)
2541                 return tipc_cfg_reply_error_string("link not found");
2542
2543         tipc_node_lock(node);
2544         l_ptr = node->links[bearer_id];
2545         if (!l_ptr) {
2546                 tipc_node_unlock(node);
2547                 return tipc_cfg_reply_error_string("link not found");
2548         }
2549         link_reset_statistics(l_ptr);
2550         tipc_node_unlock(node);
2551         return tipc_cfg_reply_none();
2552 }
2553
2554 /**
2555  * percent - convert count to a percentage of total (rounding up or down)
2556  */
2557 static u32 percent(u32 count, u32 total)
2558 {
2559         return (count * 100 + (total / 2)) / total;
2560 }
2561
2562 /**
2563  * tipc_link_stats - print link statistics
2564  * @name: link name
2565  * @buf: print buffer area
2566  * @buf_size: size of print buffer area
2567  *
2568  * Returns length of print buffer data string (or 0 if error)
2569  */
2570 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
2571 {
2572         struct tipc_link *l;
2573         struct tipc_stats *s;
2574         struct tipc_node *node;
2575         char *status;
2576         u32 profile_total = 0;
2577         unsigned int bearer_id;
2578         int ret;
2579
2580         if (!strcmp(name, tipc_bclink_name))
2581                 return tipc_bclink_stats(buf, buf_size);
2582
2583         node = tipc_link_find_owner(name, &bearer_id);
2584         if (!node)
2585                 return 0;
2586
2587         tipc_node_lock(node);
2588
2589         l = node->links[bearer_id];
2590         if (!l) {
2591                 tipc_node_unlock(node);
2592                 return 0;
2593         }
2594
2595         s = &l->stats;
2596
2597         if (tipc_link_is_active(l))
2598                 status = "ACTIVE";
2599         else if (tipc_link_is_up(l))
2600                 status = "STANDBY";
2601         else
2602                 status = "DEFUNCT";
2603
2604         ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
2605                             "  %s  MTU:%u  Priority:%u  Tolerance:%u ms"
2606                             "  Window:%u packets\n",
2607                             l->name, status, l->max_pkt, l->priority,
2608                             l->tolerance, l->queue_limit[0]);
2609
2610         ret += tipc_snprintf(buf + ret, buf_size - ret,
2611                              "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
2612                              l->next_in_no - s->recv_info, s->recv_fragments,
2613                              s->recv_fragmented, s->recv_bundles,
2614                              s->recv_bundled);
2615
2616         ret += tipc_snprintf(buf + ret, buf_size - ret,
2617                              "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
2618                              l->next_out_no - s->sent_info, s->sent_fragments,
2619                              s->sent_fragmented, s->sent_bundles,
2620                              s->sent_bundled);
2621
2622         profile_total = s->msg_length_counts;
2623         if (!profile_total)
2624                 profile_total = 1;
2625
2626         ret += tipc_snprintf(buf + ret, buf_size - ret,
2627                              "  TX profile sample:%u packets  average:%u octets\n"
2628                              "  0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
2629                              "-16384:%u%% -32768:%u%% -66000:%u%%\n",
2630                              s->msg_length_counts,
2631                              s->msg_lengths_total / profile_total,
2632                              percent(s->msg_length_profile[0], profile_total),
2633                              percent(s->msg_length_profile[1], profile_total),
2634                              percent(s->msg_length_profile[2], profile_total),
2635                              percent(s->msg_length_profile[3], profile_total),
2636                              percent(s->msg_length_profile[4], profile_total),
2637                              percent(s->msg_length_profile[5], profile_total),
2638                              percent(s->msg_length_profile[6], profile_total));
2639
2640         ret += tipc_snprintf(buf + ret, buf_size - ret,
2641                              "  RX states:%u probes:%u naks:%u defs:%u"
2642                              " dups:%u\n", s->recv_states, s->recv_probes,
2643                              s->recv_nacks, s->deferred_recv, s->duplicates);
2644
2645         ret += tipc_snprintf(buf + ret, buf_size - ret,
2646                              "  TX states:%u probes:%u naks:%u acks:%u"
2647                              " dups:%u\n", s->sent_states, s->sent_probes,
2648                              s->sent_nacks, s->sent_acks, s->retransmitted);
2649
2650         ret += tipc_snprintf(buf + ret, buf_size - ret,
2651                              "  Congestion link:%u  Send queue"
2652                              " max:%u avg:%u\n", s->link_congs,
2653                              s->max_queue_sz, s->queue_sz_counts ?
2654                              (s->accu_queue_sz / s->queue_sz_counts) : 0);
2655
2656         tipc_node_unlock(node);
2657         return ret;
2658 }
2659
2660 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
2661 {
2662         struct sk_buff *buf;
2663         struct tlv_desc *rep_tlv;
2664         int str_len;
2665         int pb_len;
2666         char *pb;
2667
2668         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2669                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2670
2671         buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2672         if (!buf)
2673                 return NULL;
2674
2675         rep_tlv = (struct tlv_desc *)buf->data;
2676         pb = TLV_DATA(rep_tlv);
2677         pb_len = ULTRA_STRING_MAX_LEN;
2678         str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
2679                                   pb, pb_len);
2680         if (!str_len) {
2681                 kfree_skb(buf);
2682                 return tipc_cfg_reply_error_string("link not found");
2683         }
2684         str_len += 1;   /* for "\0" */
2685         skb_put(buf, TLV_SPACE(str_len));
2686         TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2687
2688         return buf;
2689 }
2690
2691 /**
2692  * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
2693  * @dest: network address of destination node
2694  * @selector: used to select from set of active links
2695  *
2696  * If no active link can be found, uses default maximum packet size.
2697  */
2698 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
2699 {
2700         struct tipc_node *n_ptr;
2701         struct tipc_link *l_ptr;
2702         u32 res = MAX_PKT_DEFAULT;
2703
2704         if (dest == tipc_own_addr)
2705                 return MAX_MSG_SIZE;
2706
2707         n_ptr = tipc_node_find(dest);
2708         if (n_ptr) {
2709                 tipc_node_lock(n_ptr);
2710                 l_ptr = n_ptr->active_links[selector & 1];
2711                 if (l_ptr)
2712                         res = l_ptr->max_pkt;
2713                 tipc_node_unlock(n_ptr);
2714         }
2715         return res;
2716 }
2717
2718 static void link_print(struct tipc_link *l_ptr, const char *str)
2719 {
2720         struct tipc_bearer *b_ptr;
2721
2722         rcu_read_lock();
2723         b_ptr = rcu_dereference_rtnl(bearer_list[l_ptr->bearer_id]);
2724         if (b_ptr)
2725                 pr_info("%s Link %x<%s>:", str, l_ptr->addr, b_ptr->name);
2726         rcu_read_unlock();
2727
2728         if (link_working_unknown(l_ptr))
2729                 pr_cont(":WU\n");
2730         else if (link_reset_reset(l_ptr))
2731                 pr_cont(":RR\n");
2732         else if (link_reset_unknown(l_ptr))
2733                 pr_cont(":RU\n");
2734         else if (link_working_working(l_ptr))
2735                 pr_cont(":WW\n");
2736         else
2737                 pr_cont("\n");
2738 }