net/tipc/link.c

   1 /*
   2  * net/tipc/link.c: TIPC link code
   3  *
   4  * Copyright (c) 1996-2007, Ericsson AB
   5  * Copyright (c) 2004-2007, 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 "dbg.h"
  39 #include "link.h"
  40 #include "net.h"
  41 #include "node.h"
  42 #include "port.h"
  43 #include "addr.h"
  44 #include "node_subscr.h"
  45 #include "name_distr.h"
  46 #include "bearer.h"
  47 #include "name_table.h"
  48 #include "discover.h"
  49 #include "config.h"
  50 #include "bcast.h"
  51
  52
  53 /*
  54  * Limit for deferred reception queue:
  55  */
  56
  57 #define DEF_QUEUE_LIMIT 256u
  58
  59 /*
  60  * Link state events:
  61  */
  62
  63 #define  STARTING_EVT    856384768      /* link processing trigger */
  64 #define  TRAFFIC_MSG_EVT 560815u        /* rx'd ??? */
  65 #define  TIMEOUT_EVT     560817u        /* link timer expired */
  66
  67 /*
  68  * The following two 'message types' is really just implementation
  69  * data conveniently stored in the message header.
  70  * They must not be considered part of the protocol
  71  */
  72 #define OPEN_MSG   0
  73 #define CLOSED_MSG 1
  74
  75 /*
  76  * State value stored in 'exp_msg_count'
  77  */
  78
  79 #define START_CHANGEOVER 100000u
  80
  81 /**
  82  * struct link_name - deconstructed link name
  83  * @addr_local: network address of node at this end
  84  * @if_local: name of interface at this end
  85  * @addr_peer: network address of node at far end
  86  * @if_peer: name of interface at far end
  87  */
  88
  89 struct link_name {
  90         u32 addr_local;
  91         char if_local[TIPC_MAX_IF_NAME];
  92         u32 addr_peer;
  93         char if_peer[TIPC_MAX_IF_NAME];
  94 };
  95
  96 #if 0
  97
  98 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
  99
 100 /**
 101  * struct link_event - link up/down event notification
 102  */
 103
 104 struct link_event {
 105         u32 addr;
 106         int up;
 107         void (*fcn)(u32, char *, int);
 108         char name[TIPC_MAX_LINK_NAME];
 109 };
 110
 111 #endif
 112
 113 static void link_handle_out_of_seq_msg(struct link *l_ptr,
 114                                        struct sk_buff *buf);
 115 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf);
 116 static int  link_recv_changeover_msg(struct link **l_ptr, struct sk_buff **buf);
 117 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance);
 118 static int  link_send_sections_long(struct port *sender,
 119                                     struct iovec const *msg_sect,
 120                                     u32 num_sect, u32 destnode);
 121 static void link_check_defragm_bufs(struct link *l_ptr);
 122 static void link_state_event(struct link *l_ptr, u32 event);
 123 static void link_reset_statistics(struct link *l_ptr);
 124 static void link_print(struct link *l_ptr, struct print_buf *buf,
 125                        const char *str);
 126
 127 /*
 128  * Debugging code used by link routines only
 129  *
 130  * When debugging link problems on a system that has multiple links,
 131  * the standard TIPC debugging routines may not be useful since they
 132  * allow the output from multiple links to be intermixed.  For this reason
 133  * routines of the form "dbg_link_XXX()" have been created that will capture
 134  * debug info into a link's personal print buffer, which can then be dumped
 135  * into the TIPC system log (TIPC_LOG) upon request.
 136  *
 137  * To enable per-link debugging, use LINK_LOG_BUF_SIZE to specify the size
 138  * of the print buffer used by each link.  If LINK_LOG_BUF_SIZE is set to 0,
 139  * the dbg_link_XXX() routines simply send their output to the standard
 140  * debug print buffer (DBG_OUTPUT), if it has been defined; this can be useful
 141  * when there is only a single link in the system being debugged.
 142  *
 143  * Notes:
 144  * - When enabled, LINK_LOG_BUF_SIZE should be set to at least TIPC_PB_MIN_SIZE
 145  * - "l_ptr" must be valid when using dbg_link_XXX() macros
 146  */
 147
 148 #define LINK_LOG_BUF_SIZE 0
 149
 150 #define dbg_link(fmt, arg...)  do {if (LINK_LOG_BUF_SIZE) tipc_printf(&l_ptr->print_buf, fmt, ## arg); } while(0)
 151 #define dbg_link_msg(msg, txt) do {if (LINK_LOG_BUF_SIZE) tipc_msg_print(&l_ptr->print_buf, msg, txt); } while(0)
 152 #define dbg_link_state(txt) do {if (LINK_LOG_BUF_SIZE) link_print(l_ptr, &l_ptr->print_buf, txt); } while(0)
 153 #define dbg_link_dump() do { \
 154         if (LINK_LOG_BUF_SIZE) { \
 155                 tipc_printf(LOG, "\n\nDumping link <%s>:\n", l_ptr->name); \
 156                 tipc_printbuf_move(LOG, &l_ptr->print_buf); \
 157         } \
 158 } while (0)
 159
 160 static void dbg_print_link(struct link *l_ptr, const char *str)
 161 {
 162         if (DBG_OUTPUT != TIPC_NULL)
 163                 link_print(l_ptr, DBG_OUTPUT, str);
 164 }
 165
 166 static void dbg_print_buf_chain(struct sk_buff *root_buf)
 167 {
 168         if (DBG_OUTPUT != TIPC_NULL) {
 169                 struct sk_buff *buf = root_buf;
 170
 171                 while (buf) {
 172                         msg_dbg(buf_msg(buf), "In chain: ");
 173                         buf = buf->next;
 174                 }
 175         }
 176 }
 177
 178 /*
 179  *  Simple link routines
 180  */
 181
 182 static unsigned int align(unsigned int i)
 183 {
 184         return (i + 3) & ~3u;
 185 }
 186
 187 static int link_working_working(struct link *l_ptr)
 188 {
 189         return (l_ptr->state == WORKING_WORKING);
 190 }
 191
 192 static int link_working_unknown(struct link *l_ptr)
 193 {
 194         return (l_ptr->state == WORKING_UNKNOWN);
 195 }
 196
 197 static int link_reset_unknown(struct link *l_ptr)
 198 {
 199         return (l_ptr->state == RESET_UNKNOWN);
 200 }
 201
 202 static int link_reset_reset(struct link *l_ptr)
 203 {
 204         return (l_ptr->state == RESET_RESET);
 205 }
 206
 207 static int link_blocked(struct link *l_ptr)
 208 {
 209         return (l_ptr->exp_msg_count || l_ptr->blocked);
 210 }
 211
 212 static int link_congested(struct link *l_ptr)
 213 {
 214         return (l_ptr->out_queue_size >= l_ptr->queue_limit[0]);
 215 }
 216
 217 static u32 link_max_pkt(struct link *l_ptr)
 218 {
 219         return l_ptr->max_pkt;
 220 }
 221
 222 static void link_init_max_pkt(struct link *l_ptr)
 223 {
 224         u32 max_pkt;
 225
 226         max_pkt = (l_ptr->b_ptr->publ.mtu & ~3);
 227         if (max_pkt > MAX_MSG_SIZE)
 228                 max_pkt = MAX_MSG_SIZE;
 229
 230         l_ptr->max_pkt_target = max_pkt;
 231         if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
 232                 l_ptr->max_pkt = l_ptr->max_pkt_target;
 233         else
 234                 l_ptr->max_pkt = MAX_PKT_DEFAULT;
 235
 236         l_ptr->max_pkt_probes = 0;
 237 }
 238
 239 static u32 link_next_sent(struct link *l_ptr)
 240 {
 241         if (l_ptr->next_out)
 242                 return msg_seqno(buf_msg(l_ptr->next_out));
 243         return mod(l_ptr->next_out_no);
 244 }
 245
 246 static u32 link_last_sent(struct link *l_ptr)
 247 {
 248         return mod(link_next_sent(l_ptr) - 1);
 249 }
 250
 251 /*
 252  *  Simple non-static link routines (i.e. referenced outside this file)
 253  */
 254
 255 int tipc_link_is_up(struct link *l_ptr)
 256 {
 257         if (!l_ptr)
 258                 return 0;
 259         return (link_working_working(l_ptr) || link_working_unknown(l_ptr));
 260 }
 261
 262 int tipc_link_is_active(struct link *l_ptr)
 263 {
 264         return ((l_ptr->owner->active_links[0] == l_ptr) ||
 265                 (l_ptr->owner->active_links[1] == l_ptr));
 266 }
 267
 268 /**
 269  * link_name_validate - validate & (optionally) deconstruct link name
 270  * @name - ptr to link name string
 271  * @name_parts - ptr to area for link name components (or NULL if not needed)
 272  *
 273  * Returns 1 if link name is valid, otherwise 0.
 274  */
 275
 276 static int link_name_validate(const char *name, struct link_name *name_parts)
 277 {
 278         char name_copy[TIPC_MAX_LINK_NAME];
 279         char *addr_local;
 280         char *if_local;
 281         char *addr_peer;
 282         char *if_peer;
 283         char dummy;
 284         u32 z_local, c_local, n_local;
 285         u32 z_peer, c_peer, n_peer;
 286         u32 if_local_len;
 287         u32 if_peer_len;
 288
 289         /* copy link name & ensure length is OK */
 290
 291         name_copy[TIPC_MAX_LINK_NAME - 1] = 0;
 292         /* need above in case non-Posix strncpy() doesn't pad with nulls */
 293         strncpy(name_copy, name, TIPC_MAX_LINK_NAME);
 294         if (name_copy[TIPC_MAX_LINK_NAME - 1] != 0)
 295                 return 0;
 296
 297         /* ensure all component parts of link name are present */
 298
 299         addr_local = name_copy;
 300         if ((if_local = strchr(addr_local, ':')) == NULL)
 301                 return 0;
 302         *(if_local++) = 0;
 303         if ((addr_peer = strchr(if_local, '-')) == NULL)
 304                 return 0;
 305         *(addr_peer++) = 0;
 306         if_local_len = addr_peer - if_local;
 307         if ((if_peer = strchr(addr_peer, ':')) == NULL)
 308                 return 0;
 309         *(if_peer++) = 0;
 310         if_peer_len = strlen(if_peer) + 1;
 311
 312         /* validate component parts of link name */
 313
 314         if ((sscanf(addr_local, "%u.%u.%u%c",
 315                     &z_local, &c_local, &n_local, &dummy) != 3) ||
 316             (sscanf(addr_peer, "%u.%u.%u%c",
 317                     &z_peer, &c_peer, &n_peer, &dummy) != 3) ||
 318             (z_local > 255) || (c_local > 4095) || (n_local > 4095) ||
 319             (z_peer  > 255) || (c_peer  > 4095) || (n_peer  > 4095) ||
 320             (if_local_len <= 1) || (if_local_len > TIPC_MAX_IF_NAME) ||
 321             (if_peer_len  <= 1) || (if_peer_len  > TIPC_MAX_IF_NAME) ||
 322             (strspn(if_local, tipc_alphabet) != (if_local_len - 1)) ||
 323             (strspn(if_peer, tipc_alphabet) != (if_peer_len - 1)))
 324                 return 0;
 325
 326         /* return link name components, if necessary */
 327
 328         if (name_parts) {
 329                 name_parts->addr_local = tipc_addr(z_local, c_local, n_local);
 330                 strcpy(name_parts->if_local, if_local);
 331                 name_parts->addr_peer = tipc_addr(z_peer, c_peer, n_peer);
 332                 strcpy(name_parts->if_peer, if_peer);
 333         }
 334         return 1;
 335 }
 336
 337 /**
 338  * link_timeout - handle expiration of link timer
 339  * @l_ptr: pointer to link
 340  *
 341  * This routine must not grab "tipc_net_lock" to avoid a potential deadlock conflict
 342  * with tipc_link_delete().  (There is no risk that the node will be deleted by
 343  * another thread because tipc_link_delete() always cancels the link timer before
 344  * tipc_node_delete() is called.)
 345  */
 346
 347 static void link_timeout(struct link *l_ptr)
 348 {
 349         tipc_node_lock(l_ptr->owner);
 350
 351         /* update counters used in statistical profiling of send traffic */
 352
 353         l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
 354         l_ptr->stats.queue_sz_counts++;
 355
 356         if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
 357                 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
 358
 359         if (l_ptr->first_out) {
 360                 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
 361                 u32 length = msg_size(msg);
 362
 363                 if ((msg_user(msg) == MSG_FRAGMENTER)
 364                     && (msg_type(msg) == FIRST_FRAGMENT)) {
 365                         length = msg_size(msg_get_wrapped(msg));
 366                 }
 367                 if (length) {
 368                         l_ptr->stats.msg_lengths_total += length;
 369                         l_ptr->stats.msg_length_counts++;
 370                         if (length <= 64)
 371                                 l_ptr->stats.msg_length_profile[0]++;
 372                         else if (length <= 256)
 373                                 l_ptr->stats.msg_length_profile[1]++;
 374                         else if (length <= 1024)
 375                                 l_ptr->stats.msg_length_profile[2]++;
 376                         else if (length <= 4096)
 377                                 l_ptr->stats.msg_length_profile[3]++;
 378                         else if (length <= 16384)
 379                                 l_ptr->stats.msg_length_profile[4]++;
 380                         else if (length <= 32768)
 381                                 l_ptr->stats.msg_length_profile[5]++;
 382                         else
 383                                 l_ptr->stats.msg_length_profile[6]++;
 384                 }
 385         }
 386
 387         /* do all other link processing performed on a periodic basis */
 388
 389         link_check_defragm_bufs(l_ptr);
 390
 391         link_state_event(l_ptr, TIMEOUT_EVT);
 392
 393         if (l_ptr->next_out)
 394                 tipc_link_push_queue(l_ptr);
 395
 396         tipc_node_unlock(l_ptr->owner);
 397 }
 398
 399 static void link_set_timer(struct link *l_ptr, u32 time)
 400 {
 401         k_start_timer(&l_ptr->timer, time);
 402 }
 403
 404 /**
 405  * tipc_link_create - create a new link
 406  * @b_ptr: pointer to associated bearer
 407  * @peer: network address of node at other end of link
 408  * @media_addr: media address to use when sending messages over link
 409  *
 410  * Returns pointer to link.
 411  */
 412
 413 struct link *tipc_link_create(struct bearer *b_ptr, const u32 peer,
 414                               const struct tipc_media_addr *media_addr)
 415 {
 416         struct link *l_ptr;
 417         struct tipc_msg *msg;
 418         char *if_name;
 419
 420         l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
 421         if (!l_ptr) {
 422                 warn("Link creation failed, no memory\n");
 423                 return NULL;
 424         }
 425
 426         if (LINK_LOG_BUF_SIZE) {
 427                 char *pb = kmalloc(LINK_LOG_BUF_SIZE, GFP_ATOMIC);
 428
 429                 if (!pb) {
 430                         kfree(l_ptr);
 431                         warn("Link creation failed, no memory for print buffer\n");
 432                         return NULL;
 433                 }
 434                 tipc_printbuf_init(&l_ptr->print_buf, pb, LINK_LOG_BUF_SIZE);
 435         }
 436
 437         l_ptr->addr = peer;
 438         if_name = strchr(b_ptr->publ.name, ':') + 1;
 439         sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:",
 440                 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
 441                 tipc_node(tipc_own_addr),
 442                 if_name,
 443                 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
 444                 /* note: peer i/f is appended to link name by reset/activate */
 445         memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
 446         l_ptr->checkpoint = 1;
 447         l_ptr->b_ptr = b_ptr;
 448         link_set_supervision_props(l_ptr, b_ptr->media->tolerance);
 449         l_ptr->state = RESET_UNKNOWN;
 450
 451         l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
 452         msg = l_ptr->pmsg;
 453         msg_init(msg, LINK_PROTOCOL, RESET_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
 454         msg_set_size(msg, sizeof(l_ptr->proto_msg));
 455         msg_set_session(msg, tipc_random);
 456         msg_set_bearer_id(msg, b_ptr->identity);
 457         strcpy((char *)msg_data(msg), if_name);
 458
 459         l_ptr->priority = b_ptr->priority;
 460         tipc_link_set_queue_limits(l_ptr, b_ptr->media->window);
 461
 462         link_init_max_pkt(l_ptr);
 463
 464         l_ptr->next_out_no = 1;
 465         INIT_LIST_HEAD(&l_ptr->waiting_ports);
 466
 467         link_reset_statistics(l_ptr);
 468
 469         l_ptr->owner = tipc_node_attach_link(l_ptr);
 470         if (!l_ptr->owner) {
 471                 if (LINK_LOG_BUF_SIZE)
 472                         kfree(l_ptr->print_buf.buf);
 473                 kfree(l_ptr);
 474                 return NULL;
 475         }
 476
 477         k_init_timer(&l_ptr->timer, (Handler)link_timeout, (unsigned long)l_ptr);
 478         list_add_tail(&l_ptr->link_list, &b_ptr->links);
 479         tipc_k_signal((Handler)tipc_link_start, (unsigned long)l_ptr);
 480
 481         dbg("tipc_link_create(): tolerance = %u,cont intv = %u, abort_limit = %u\n",
 482             l_ptr->tolerance, l_ptr->continuity_interval, l_ptr->abort_limit);
 483
 484         return l_ptr;
 485 }
 486
 487 /**
 488  * tipc_link_delete - delete a link
 489  * @l_ptr: pointer to link
 490  *
 491  * Note: 'tipc_net_lock' is write_locked, bearer is locked.
 492  * This routine must not grab the node lock until after link timer cancellation
 493  * to avoid a potential deadlock situation.
 494  */
 495
 496 void tipc_link_delete(struct link *l_ptr)
 497 {
 498         if (!l_ptr) {
 499                 err("Attempt to delete non-existent link\n");
 500                 return;
 501         }
 502
 503         dbg("tipc_link_delete()\n");
 504
 505         k_cancel_timer(&l_ptr->timer);
 506
 507         tipc_node_lock(l_ptr->owner);
 508         tipc_link_reset(l_ptr);
 509         tipc_node_detach_link(l_ptr->owner, l_ptr);
 510         tipc_link_stop(l_ptr);
 511         list_del_init(&l_ptr->link_list);
 512         if (LINK_LOG_BUF_SIZE)
 513                 kfree(l_ptr->print_buf.buf);
 514         tipc_node_unlock(l_ptr->owner);
 515         k_term_timer(&l_ptr->timer);
 516         kfree(l_ptr);
 517 }
 518
 519 void tipc_link_start(struct link *l_ptr)
 520 {
 521         dbg("tipc_link_start %x\n", l_ptr);
 522         link_state_event(l_ptr, STARTING_EVT);
 523 }
 524
 525 /**
 526  * link_schedule_port - schedule port for deferred sending
 527  * @l_ptr: pointer to link
 528  * @origport: reference to sending port
 529  * @sz: amount of data to be sent
 530  *
 531  * Schedules port for renewed sending of messages after link congestion
 532  * has abated.
 533  */
 534
 535 static int link_schedule_port(struct link *l_ptr, u32 origport, u32 sz)
 536 {
 537         struct port *p_ptr;
 538
 539         spin_lock_bh(&tipc_port_list_lock);
 540         p_ptr = tipc_port_lock(origport);
 541         if (p_ptr) {
 542                 if (!p_ptr->wakeup)
 543                         goto exit;
 544                 if (!list_empty(&p_ptr->wait_list))
 545                         goto exit;
 546                 p_ptr->congested_link = l_ptr;
 547                 p_ptr->publ.congested = 1;
 548                 p_ptr->waiting_pkts = 1 + ((sz - 1) / link_max_pkt(l_ptr));
 549                 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
 550                 l_ptr->stats.link_congs++;
 551 exit:
 552                 tipc_port_unlock(p_ptr);
 553         }
 554         spin_unlock_bh(&tipc_port_list_lock);
 555         return -ELINKCONG;
 556 }
 557
 558 void tipc_link_wakeup_ports(struct link *l_ptr, int all)
 559 {
 560         struct port *p_ptr;
 561         struct port *temp_p_ptr;
 562         int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
 563
 564         if (all)
 565                 win = 100000;
 566         if (win <= 0)
 567                 return;
 568         if (!spin_trylock_bh(&tipc_port_list_lock))
 569                 return;
 570         if (link_congested(l_ptr))
 571                 goto exit;
 572         list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
 573                                  wait_list) {
 574                 if (win <= 0)
 575                         break;
 576                 list_del_init(&p_ptr->wait_list);
 577                 p_ptr->congested_link = NULL;
 578                 spin_lock_bh(p_ptr->publ.lock);
 579                 p_ptr->publ.congested = 0;
 580                 p_ptr->wakeup(&p_ptr->publ);
 581                 win -= p_ptr->waiting_pkts;
 582                 spin_unlock_bh(p_ptr->publ.lock);
 583         }
 584
 585 exit:
 586         spin_unlock_bh(&tipc_port_list_lock);
 587 }
 588
 589 /**
 590  * link_release_outqueue - purge link's outbound message queue
 591  * @l_ptr: pointer to link
 592  */
 593
 594 static void link_release_outqueue(struct link *l_ptr)
 595 {
 596         struct sk_buff *buf = l_ptr->first_out;
 597         struct sk_buff *next;
 598
 599         while (buf) {
 600                 next = buf->next;
 601                 buf_discard(buf);
 602                 buf = next;
 603         }
 604         l_ptr->first_out = NULL;
 605         l_ptr->out_queue_size = 0;
 606 }
 607
 608 /**
 609  * tipc_link_reset_fragments - purge link's inbound message fragments queue
 610  * @l_ptr: pointer to link
 611  */
 612
 613 void tipc_link_reset_fragments(struct link *l_ptr)
 614 {
 615         struct sk_buff *buf = l_ptr->defragm_buf;
 616         struct sk_buff *next;
 617
 618         while (buf) {
 619                 next = buf->next;
 620                 buf_discard(buf);
 621                 buf = next;
 622         }
 623         l_ptr->defragm_buf = NULL;
 624 }
 625
 626 /**
 627  * tipc_link_stop - purge all inbound and outbound messages associated with link
 628  * @l_ptr: pointer to link
 629  */
 630
 631 void tipc_link_stop(struct link *l_ptr)
 632 {
 633         struct sk_buff *buf;
 634         struct sk_buff *next;
 635
 636         buf = l_ptr->oldest_deferred_in;
 637         while (buf) {
 638                 next = buf->next;
 639                 buf_discard(buf);
 640                 buf = next;
 641         }
 642
 643         buf = l_ptr->first_out;
 644         while (buf) {
 645                 next = buf->next;
 646                 buf_discard(buf);
 647                 buf = next;
 648         }
 649
 650         tipc_link_reset_fragments(l_ptr);
 651
 652         buf_discard(l_ptr->proto_msg_queue);
 653         l_ptr->proto_msg_queue = NULL;
 654 }
 655
 656 #if 0
 657
 658 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
 659
 660 static void link_recv_event(struct link_event *ev)
 661 {
 662         ev->fcn(ev->addr, ev->name, ev->up);
 663         kfree(ev);
 664 }
 665
 666 static void link_send_event(void (*fcn)(u32 a, char *n, int up),
 667                             struct link *l_ptr, int up)
 668 {
 669         struct link_event *ev;
 670
 671         ev = kmalloc(sizeof(*ev), GFP_ATOMIC);
 672         if (!ev) {
 673                 warn("Link event allocation failure\n");
 674                 return;
 675         }
 676         ev->addr = l_ptr->addr;
 677         ev->up = up;
 678         ev->fcn = fcn;
 679         memcpy(ev->name, l_ptr->name, TIPC_MAX_LINK_NAME);
 680         tipc_k_signal((Handler)link_recv_event, (unsigned long)ev);
 681 }
 682
 683 #else
 684
 685 #define link_send_event(fcn, l_ptr, up) do { } while (0)
 686
 687 #endif
 688
 689 void tipc_link_reset(struct link *l_ptr)
 690 {
 691         struct sk_buff *buf;
 692         u32 prev_state = l_ptr->state;
 693         u32 checkpoint = l_ptr->next_in_no;
 694         int was_active_link = tipc_link_is_active(l_ptr);
 695
 696         msg_set_session(l_ptr->pmsg, msg_session(l_ptr->pmsg) + 1);
 697
 698         /* Link is down, accept any session: */
 699         l_ptr->peer_session = 0;
 700
 701         /* Prepare for max packet size negotiation */
 702         link_init_max_pkt(l_ptr);
 703
 704         l_ptr->state = RESET_UNKNOWN;
 705         dbg_link_state("Resetting Link\n");
 706
 707         if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
 708                 return;
 709
 710         tipc_node_link_down(l_ptr->owner, l_ptr);
 711         tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
 712 #if 0
 713         tipc_printf(TIPC_CONS, "\nReset link <%s>\n", l_ptr->name);
 714         dbg_link_dump();
 715 #endif
 716         if (was_active_link && tipc_node_has_active_links(l_ptr->owner) &&
 717             l_ptr->owner->permit_changeover) {
 718                 l_ptr->reset_checkpoint = checkpoint;
 719                 l_ptr->exp_msg_count = START_CHANGEOVER;
 720         }
 721
 722         /* Clean up all queues: */
 723
 724         link_release_outqueue(l_ptr);
 725         buf_discard(l_ptr->proto_msg_queue);
 726         l_ptr->proto_msg_queue = NULL;
 727         buf = l_ptr->oldest_deferred_in;
 728         while (buf) {
 729                 struct sk_buff *next = buf->next;
 730                 buf_discard(buf);
 731                 buf = next;
 732         }
 733         if (!list_empty(&l_ptr->waiting_ports))
 734                 tipc_link_wakeup_ports(l_ptr, 1);
 735
 736         l_ptr->retransm_queue_head = 0;
 737         l_ptr->retransm_queue_size = 0;
 738         l_ptr->last_out = NULL;
 739         l_ptr->first_out = NULL;
 740         l_ptr->next_out = NULL;
 741         l_ptr->unacked_window = 0;
 742         l_ptr->checkpoint = 1;
 743         l_ptr->next_out_no = 1;
 744         l_ptr->deferred_inqueue_sz = 0;
 745         l_ptr->oldest_deferred_in = NULL;
 746         l_ptr->newest_deferred_in = NULL;
 747         l_ptr->fsm_msg_cnt = 0;
 748         l_ptr->stale_count = 0;
 749         link_reset_statistics(l_ptr);
 750
 751         link_send_event(tipc_cfg_link_event, l_ptr, 0);
 752         if (!in_own_cluster(l_ptr->addr))
 753                 link_send_event(tipc_disc_link_event, l_ptr, 0);
 754 }
 755
 756
 757 static void link_activate(struct link *l_ptr)
 758 {
 759         l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
 760         tipc_node_link_up(l_ptr->owner, l_ptr);
 761         tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
 762         link_send_event(tipc_cfg_link_event, l_ptr, 1);
 763         if (!in_own_cluster(l_ptr->addr))
 764                 link_send_event(tipc_disc_link_event, l_ptr, 1);
 765 }
 766
 767 /**
 768  * link_state_event - link finite state machine
 769  * @l_ptr: pointer to link
 770  * @event: state machine event to process
 771  */
 772
 773 static void link_state_event(struct link *l_ptr, unsigned event)
 774 {
 775         struct link *other;
 776         u32 cont_intv = l_ptr->continuity_interval;
 777
 778         if (!l_ptr->started && (event != STARTING_EVT))
 779                 return;         /* Not yet. */
 780
 781         if (link_blocked(l_ptr)) {
 782                 if (event == TIMEOUT_EVT) {
 783                         link_set_timer(l_ptr, cont_intv);
 784                 }
 785                 return;   /* Changeover going on */
 786         }
 787         dbg_link("STATE_EV: <%s> ", l_ptr->name);
 788
 789         switch (l_ptr->state) {
 790         case WORKING_WORKING:
 791                 dbg_link("WW/");
 792                 switch (event) {
 793                 case TRAFFIC_MSG_EVT:
 794                         dbg_link("TRF-");
 795                         /* fall through */
 796                 case ACTIVATE_MSG:
 797                         dbg_link("ACT\n");
 798                         break;
 799                 case TIMEOUT_EVT:
 800                         dbg_link("TIM ");
 801                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
 802                                 l_ptr->checkpoint = l_ptr->next_in_no;
 803                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
 804                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
 805                                                                  0, 0, 0, 0, 0);
 806                                         l_ptr->fsm_msg_cnt++;
 807                                 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
 808                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
 809                                                                  1, 0, 0, 0, 0);
 810                                         l_ptr->fsm_msg_cnt++;
 811                                 }
 812                                 link_set_timer(l_ptr, cont_intv);
 813                                 break;
 814                         }
 815                         dbg_link(" -> WU\n");
 816                         l_ptr->state = WORKING_UNKNOWN;
 817                         l_ptr->fsm_msg_cnt = 0;
 818                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
 819                         l_ptr->fsm_msg_cnt++;
 820                         link_set_timer(l_ptr, cont_intv / 4);
 821                         break;
 822                 case RESET_MSG:
 823                         dbg_link("RES -> RR\n");
 824                         info("Resetting link <%s>, requested by peer\n",
 825                              l_ptr->name);
 826                         tipc_link_reset(l_ptr);
 827                         l_ptr->state = RESET_RESET;
 828                         l_ptr->fsm_msg_cnt = 0;
 829                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
 830                         l_ptr->fsm_msg_cnt++;
 831                         link_set_timer(l_ptr, cont_intv);
 832                         break;
 833                 default:
 834                         err("Unknown link event %u in WW state\n", event);
 835                 }
 836                 break;
 837         case WORKING_UNKNOWN:
 838                 dbg_link("WU/");
 839                 switch (event) {
 840                 case TRAFFIC_MSG_EVT:
 841                         dbg_link("TRF-");
 842                 case ACTIVATE_MSG:
 843                         dbg_link("ACT -> WW\n");
 844                         l_ptr->state = WORKING_WORKING;
 845                         l_ptr->fsm_msg_cnt = 0;
 846                         link_set_timer(l_ptr, cont_intv);
 847                         break;
 848                 case RESET_MSG:
 849                         dbg_link("RES -> RR\n");
 850                         info("Resetting link <%s>, requested by peer "
 851                              "while probing\n", l_ptr->name);
 852                         tipc_link_reset(l_ptr);
 853                         l_ptr->state = RESET_RESET;
 854                         l_ptr->fsm_msg_cnt = 0;
 855                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
 856                         l_ptr->fsm_msg_cnt++;
 857                         link_set_timer(l_ptr, cont_intv);
 858                         break;
 859                 case TIMEOUT_EVT:
 860                         dbg_link("TIM ");
 861                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
 862                                 dbg_link("-> WW \n");
 863                                 l_ptr->state = WORKING_WORKING;
 864                                 l_ptr->fsm_msg_cnt = 0;
 865                                 l_ptr->checkpoint = l_ptr->next_in_no;
 866                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
 867                                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
 868                                                                  0, 0, 0, 0, 0);
 869                                         l_ptr->fsm_msg_cnt++;
 870                                 }
 871                                 link_set_timer(l_ptr, cont_intv);
 872                         } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
 873                                 dbg_link("Probing %u/%u,timer = %u ms)\n",
 874                                          l_ptr->fsm_msg_cnt, l_ptr->abort_limit,
 875                                          cont_intv / 4);
 876                                 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
 877                                                          1, 0, 0, 0, 0);
 878                                 l_ptr->fsm_msg_cnt++;
 879                                 link_set_timer(l_ptr, cont_intv / 4);
 880                         } else {        /* Link has failed */
 881                                 dbg_link("-> RU (%u probes unanswered)\n",
 882                                          l_ptr->fsm_msg_cnt);
 883                                 warn("Resetting link <%s>, peer not responding\n",
 884                                      l_ptr->name);
 885                                 tipc_link_reset(l_ptr);
 886                                 l_ptr->state = RESET_UNKNOWN;
 887                                 l_ptr->fsm_msg_cnt = 0;
 888                                 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
 889                                                          0, 0, 0, 0, 0);
 890                                 l_ptr->fsm_msg_cnt++;
 891                                 link_set_timer(l_ptr, cont_intv);
 892                         }
 893                         break;
 894                 default:
 895                         err("Unknown link event %u in WU state\n", event);
 896                 }
 897                 break;
 898         case RESET_UNKNOWN:
 899                 dbg_link("RU/");
 900                 switch (event) {
 901                 case TRAFFIC_MSG_EVT:
 902                         dbg_link("TRF-\n");
 903                         break;
 904                 case ACTIVATE_MSG:
 905                         other = l_ptr->owner->active_links[0];
 906                         if (other && link_working_unknown(other)) {
 907                                 dbg_link("ACT\n");
 908                                 break;
 909                         }
 910                         dbg_link("ACT -> WW\n");
 911                         l_ptr->state = WORKING_WORKING;
 912                         l_ptr->fsm_msg_cnt = 0;
 913                         link_activate(l_ptr);
 914                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
 915                         l_ptr->fsm_msg_cnt++;
 916                         link_set_timer(l_ptr, cont_intv);
 917                         break;
 918                 case RESET_MSG:
 919                         dbg_link("RES \n");
 920                         dbg_link(" -> RR\n");
 921                         l_ptr->state = RESET_RESET;
 922                         l_ptr->fsm_msg_cnt = 0;
 923                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
 924                         l_ptr->fsm_msg_cnt++;
 925                         link_set_timer(l_ptr, cont_intv);
 926                         break;
 927                 case STARTING_EVT:
 928                         dbg_link("START-");
 929                         l_ptr->started = 1;
 930                         /* fall through */
 931                 case TIMEOUT_EVT:
 932                         dbg_link("TIM \n");
 933                         tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
 934                         l_ptr->fsm_msg_cnt++;
 935                         link_set_timer(l_ptr, cont_intv);
 936                         break;
 937                 default:
 938                         err("Unknown link event %u in RU state\n", event);
 939                 }
 940                 break;
 941         case RESET_RESET:
 942                 dbg_link("RR/ ");
 943                 switch (event) {
 944                 case TRAFFIC_MSG_EVT:
 945                         dbg_link("TRF-");
 946                         /* fall through */
 947                 case ACTIVATE_MSG:
 948                         other = l_ptr->owner->active_links[0];
 949                         if (other && link_working_unknown(other)) {
 950                                 dbg_link("ACT\n");
 951                                 break;
 952                         }
 953                         dbg_link("ACT -> WW\n");
 954                         l_ptr->state = WORKING_WORKING;
 955                         l_ptr->fsm_msg_cnt = 0;
 956                         link_activate(l_ptr);
 957                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
 958                         l_ptr->fsm_msg_cnt++;
 959                         link_set_timer(l_ptr, cont_intv);
 960                         break;
 961                 case RESET_MSG:
 962                         dbg_link("RES\n");
 963                         break;
 964                 case TIMEOUT_EVT:
 965                         dbg_link("TIM\n");
 966                         tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
 967                         l_ptr->fsm_msg_cnt++;
 968                         link_set_timer(l_ptr, cont_intv);
 969                         dbg_link("fsm_msg_cnt %u\n", l_ptr->fsm_msg_cnt);
 970                         break;
 971                 default:
 972                         err("Unknown link event %u in RR state\n", event);
 973                 }
 974                 break;
 975         default:
 976                 err("Unknown link state %u/%u\n", l_ptr->state, event);
 977         }
 978 }
 979
 980 /*
 981  * link_bundle_buf(): Append contents of a buffer to
 982  * the tail of an existing one.
 983  */
 984
 985 static int link_bundle_buf(struct link *l_ptr,
 986                            struct sk_buff *bundler,
 987                            struct sk_buff *buf)
 988 {
 989         struct tipc_msg *bundler_msg = buf_msg(bundler);
 990         struct tipc_msg *msg = buf_msg(buf);
 991         u32 size = msg_size(msg);
 992         u32 bundle_size = msg_size(bundler_msg);
 993         u32 to_pos = align(bundle_size);
 994         u32 pad = to_pos - bundle_size;
 995
 996         if (msg_user(bundler_msg) != MSG_BUNDLER)
 997                 return 0;
 998         if (msg_type(bundler_msg) != OPEN_MSG)
 999                 return 0;
1000         if (skb_tailroom(bundler) < (pad + size))
1001                 return 0;
1002         if (link_max_pkt(l_ptr) < (to_pos + size))
1003                 return 0;
1004
1005         skb_put(bundler, pad + size);
1006         skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
1007         msg_set_size(bundler_msg, to_pos + size);
1008         msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
1009         dbg("Packed msg # %u(%u octets) into pos %u in buf(#%u)\n",
1010             msg_msgcnt(bundler_msg), size, to_pos, msg_seqno(bundler_msg));
1011         msg_dbg(msg, "PACKD:");
1012         buf_discard(buf);
1013         l_ptr->stats.sent_bundled++;
1014         return 1;
1015 }
1016
1017 static void link_add_to_outqueue(struct link *l_ptr,
1018                                  struct sk_buff *buf,
1019                                  struct tipc_msg *msg)
1020 {
1021         u32 ack = mod(l_ptr->next_in_no - 1);
1022         u32 seqno = mod(l_ptr->next_out_no++);
1023
1024         msg_set_word(msg, 2, ((ack << 16) | seqno));
1025         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1026         buf->next = NULL;
1027         if (l_ptr->first_out) {
1028                 l_ptr->last_out->next = buf;
1029                 l_ptr->last_out = buf;
1030         } else
1031                 l_ptr->first_out = l_ptr->last_out = buf;
1032         l_ptr->out_queue_size++;
1033 }
1034
1035 /*
1036  * tipc_link_send_buf() is the 'full path' for messages, called from
1037  * inside TIPC when the 'fast path' in tipc_send_buf
1038  * has failed, and from link_send()
1039  */
1040
1041 int tipc_link_send_buf(struct link *l_ptr, struct sk_buff *buf)
1042 {
1043         struct tipc_msg *msg = buf_msg(buf);
1044         u32 size = msg_size(msg);
1045         u32 dsz = msg_data_sz(msg);
1046         u32 queue_size = l_ptr->out_queue_size;
1047         u32 imp = msg_tot_importance(msg);
1048         u32 queue_limit = l_ptr->queue_limit[imp];
1049         u32 max_packet = link_max_pkt(l_ptr);
1050
1051         msg_set_prevnode(msg, tipc_own_addr);   /* If routed message */
1052
1053         /* Match msg importance against queue limits: */
1054
1055         if (unlikely(queue_size >= queue_limit)) {
1056                 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
1057                         return link_schedule_port(l_ptr, msg_origport(msg),
1058                                                   size);
1059                 }
1060                 msg_dbg(msg, "TIPC: Congestion, throwing away\n");
1061                 buf_discard(buf);
1062                 if (imp > CONN_MANAGER) {
1063                         warn("Resetting link <%s>, send queue full", l_ptr->name);
1064                         tipc_link_reset(l_ptr);
1065                 }
1066                 return dsz;
1067         }
1068
1069         /* Fragmentation needed ? */
1070
1071         if (size > max_packet)
1072                 return tipc_link_send_long_buf(l_ptr, buf);
1073
1074         /* Packet can be queued or sent: */
1075
1076         if (queue_size > l_ptr->stats.max_queue_sz)
1077                 l_ptr->stats.max_queue_sz = queue_size;
1078
1079         if (likely(!tipc_bearer_congested(l_ptr->b_ptr, l_ptr) &&
1080                    !link_congested(l_ptr))) {
1081                 link_add_to_outqueue(l_ptr, buf, msg);
1082
1083                 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr))) {
1084                         l_ptr->unacked_window = 0;
1085                 } else {
1086                         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1087                         l_ptr->stats.bearer_congs++;
1088                         l_ptr->next_out = buf;
1089                 }
1090                 return dsz;
1091         }
1092         /* Congestion: can message be bundled ?: */
1093
1094         if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
1095             (msg_user(msg) != MSG_FRAGMENTER)) {
1096
1097                 /* Try adding message to an existing bundle */
1098
1099                 if (l_ptr->next_out &&
1100                     link_bundle_buf(l_ptr, l_ptr->last_out, buf)) {
1101                         tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1102                         return dsz;
1103                 }
1104
1105                 /* Try creating a new bundle */
1106
1107                 if (size <= max_packet * 2 / 3) {
1108                         struct sk_buff *bundler = buf_acquire(max_packet);
1109                         struct tipc_msg bundler_hdr;
1110
1111                         if (bundler) {
1112                                 msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
1113                                          TIPC_OK, INT_H_SIZE, l_ptr->addr);
1114                                 skb_copy_to_linear_data(bundler, &bundler_hdr,
1115                                                         INT_H_SIZE);
1116                                 skb_trim(bundler, INT_H_SIZE);
1117                                 link_bundle_buf(l_ptr, bundler, buf);
1118                                 buf = bundler;
1119                                 msg = buf_msg(buf);
1120                                 l_ptr->stats.sent_bundles++;
1121                         }
1122                 }
1123         }
1124         if (!l_ptr->next_out)
1125                 l_ptr->next_out = buf;
1126         link_add_to_outqueue(l_ptr, buf, msg);
1127         tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1128         return dsz;
1129 }
1130
1131 /*
1132  * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
1133  * not been selected yet, and the the owner node is not locked
1134  * Called by TIPC internal users, e.g. the name distributor
1135  */
1136
1137 int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
1138 {
1139         struct link *l_ptr;
1140         struct node *n_ptr;
1141         int res = -ELINKCONG;
1142
1143         read_lock_bh(&tipc_net_lock);
1144         n_ptr = tipc_node_select(dest, selector);
1145         if (n_ptr) {
1146                 tipc_node_lock(n_ptr);
1147                 l_ptr = n_ptr->active_links[selector & 1];
1148                 if (l_ptr) {
1149                         dbg("tipc_link_send: found link %x for dest %x\n", l_ptr, dest);
1150                         res = tipc_link_send_buf(l_ptr, buf);
1151                 } else {
1152                         dbg("Attempt to send msg to unreachable node:\n");
1153                         msg_dbg(buf_msg(buf),">>>");
1154                         buf_discard(buf);
1155                 }
1156                 tipc_node_unlock(n_ptr);
1157         } else {
1158                 dbg("Attempt to send msg to unknown node:\n");
1159                 msg_dbg(buf_msg(buf),">>>");
1160                 buf_discard(buf);
1161         }
1162         read_unlock_bh(&tipc_net_lock);
1163         return res;
1164 }
1165
1166 /*
1167  * link_send_buf_fast: Entry for data messages where the
1168  * destination link is known and the header is complete,
1169  * inclusive total message length. Very time critical.
1170  * Link is locked. Returns user data length.
1171  */
1172
1173 static int link_send_buf_fast(struct link *l_ptr, struct sk_buff *buf,
1174                               u32 *used_max_pkt)
1175 {
1176         struct tipc_msg *msg = buf_msg(buf);
1177         int res = msg_data_sz(msg);
1178
1179         if (likely(!link_congested(l_ptr))) {
1180                 if (likely(msg_size(msg) <= link_max_pkt(l_ptr))) {
1181                         if (likely(list_empty(&l_ptr->b_ptr->cong_links))) {
1182                                 link_add_to_outqueue(l_ptr, buf, msg);
1183                                 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf,
1184                                                             &l_ptr->media_addr))) {
1185                                         l_ptr->unacked_window = 0;
1186                                         msg_dbg(msg,"SENT_FAST:");
1187                                         return res;
1188                                 }
1189                                 dbg("failed sent fast...\n");
1190                                 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1191                                 l_ptr->stats.bearer_congs++;
1192                                 l_ptr->next_out = buf;
1193                                 return res;
1194                         }
1195                 }
1196                 else
1197                         *used_max_pkt = link_max_pkt(l_ptr);
1198         }
1199         return tipc_link_send_buf(l_ptr, buf);  /* All other cases */
1200 }
1201
1202 /*
1203  * tipc_send_buf_fast: Entry for data messages where the
1204  * destination node is known and the header is complete,
1205  * inclusive total message length.
1206  * Returns user data length.
1207  */
1208 int tipc_send_buf_fast(struct sk_buff *buf, u32 destnode)
1209 {
1210         struct link *l_ptr;
1211         struct node *n_ptr;
1212         int res;
1213         u32 selector = msg_origport(buf_msg(buf)) & 1;
1214         u32 dummy;
1215
1216         if (destnode == tipc_own_addr)
1217                 return tipc_port_recv_msg(buf);
1218
1219         read_lock_bh(&tipc_net_lock);
1220         n_ptr = tipc_node_select(destnode, selector);
1221         if (likely(n_ptr)) {
1222                 tipc_node_lock(n_ptr);
1223                 l_ptr = n_ptr->active_links[selector];
1224                 dbg("send_fast: buf %x selected %x, destnode = %x\n",
1225                     buf, l_ptr, destnode);
1226                 if (likely(l_ptr)) {
1227                         res = link_send_buf_fast(l_ptr, buf, &dummy);
1228                         tipc_node_unlock(n_ptr);
1229                         read_unlock_bh(&tipc_net_lock);
1230                         return res;
1231                 }
1232                 tipc_node_unlock(n_ptr);
1233         }
1234         read_unlock_bh(&tipc_net_lock);
1235         res = msg_data_sz(buf_msg(buf));
1236         tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1237         return res;
1238 }
1239
1240
1241 /*
1242  * tipc_link_send_sections_fast: Entry for messages where the
1243  * destination processor is known and the header is complete,
1244  * except for total message length.
1245  * Returns user data length or errno.
1246  */
1247 int tipc_link_send_sections_fast(struct port *sender,
1248                                  struct iovec const *msg_sect,
1249                                  const u32 num_sect,
1250                                  u32 destaddr)
1251 {
1252         struct tipc_msg *hdr = &sender->publ.phdr;
1253         struct link *l_ptr;
1254         struct sk_buff *buf;
1255         struct node *node;
1256         int res;
1257         u32 selector = msg_origport(hdr) & 1;
1258
1259 again:
1260         /*
1261          * Try building message using port's max_pkt hint.
1262          * (Must not hold any locks while building message.)
1263          */
1264
1265         res = msg_build(hdr, msg_sect, num_sect, sender->publ.max_pkt,
1266                         !sender->user_port, &buf);
1267
1268         read_lock_bh(&tipc_net_lock);
1269         node = tipc_node_select(destaddr, selector);
1270         if (likely(node)) {
1271                 tipc_node_lock(node);
1272                 l_ptr = node->active_links[selector];
1273                 if (likely(l_ptr)) {
1274                         if (likely(buf)) {
1275                                 res = link_send_buf_fast(l_ptr, buf,
1276                                                          &sender->publ.max_pkt);
1277                                 if (unlikely(res < 0))
1278                                         buf_discard(buf);
1279 exit:
1280                                 tipc_node_unlock(node);
1281                                 read_unlock_bh(&tipc_net_lock);
1282                                 return res;
1283                         }
1284
1285                         /* Exit if build request was invalid */
1286
1287                         if (unlikely(res < 0))
1288                                 goto exit;
1289
1290                         /* Exit if link (or bearer) is congested */
1291
1292                         if (link_congested(l_ptr) ||
1293                             !list_empty(&l_ptr->b_ptr->cong_links)) {
1294                                 res = link_schedule_port(l_ptr,
1295                                                          sender->publ.ref, res);
1296                                 goto exit;
1297                         }
1298
1299                         /*
1300                          * Message size exceeds max_pkt hint; update hint,
1301                          * then re-try fast path or fragment the message
1302                          */
1303
1304                         sender->publ.max_pkt = link_max_pkt(l_ptr);
1305                         tipc_node_unlock(node);
1306                         read_unlock_bh(&tipc_net_lock);
1307
1308
1309                         if ((msg_hdr_sz(hdr) + res) <= sender->publ.max_pkt)
1310                                 goto again;
1311
1312                         return link_send_sections_long(sender, msg_sect,
1313                                                        num_sect, destaddr);
1314                 }
1315                 tipc_node_unlock(node);
1316         }
1317         read_unlock_bh(&tipc_net_lock);
1318
1319         /* Couldn't find a link to the destination node */
1320
1321         if (buf)
1322                 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1323         if (res >= 0)
1324                 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1325                                                  TIPC_ERR_NO_NODE);
1326         return res;
1327 }
1328
1329 /*
1330  * link_send_sections_long(): Entry for long messages where the
1331  * destination node is known and the header is complete,
1332  * inclusive total message length.
1333  * Link and bearer congestion status have been checked to be ok,
1334  * and are ignored if they change.
1335  *
1336  * Note that fragments do not use the full link MTU so that they won't have
1337  * to undergo refragmentation if link changeover causes them to be sent
1338  * over another link with an additional tunnel header added as prefix.
1339  * (Refragmentation will still occur if the other link has a smaller MTU.)
1340  *
1341  * Returns user data length or errno.
1342  */
1343 static int link_send_sections_long(struct port *sender,
1344                                    struct iovec const *msg_sect,
1345                                    u32 num_sect,
1346                                    u32 destaddr)
1347 {
1348         struct link *l_ptr;
1349         struct node *node;
1350         struct tipc_msg *hdr = &sender->publ.phdr;
1351         u32 dsz = msg_data_sz(hdr);
1352         u32 max_pkt,fragm_sz,rest;
1353         struct tipc_msg fragm_hdr;
1354         struct sk_buff *buf,*buf_chain,*prev;
1355         u32 fragm_crs,fragm_rest,hsz,sect_rest;
1356         const unchar *sect_crs;
1357         int curr_sect;
1358         u32 fragm_no;
1359
1360 again:
1361         fragm_no = 1;
1362         max_pkt = sender->publ.max_pkt - INT_H_SIZE;
1363                 /* leave room for tunnel header in case of link changeover */
1364         fragm_sz = max_pkt - INT_H_SIZE;
1365                 /* leave room for fragmentation header in each fragment */
1366         rest = dsz;
1367         fragm_crs = 0;
1368         fragm_rest = 0;
1369         sect_rest = 0;
1370         sect_crs = NULL;
1371         curr_sect = -1;
1372
1373         /* Prepare reusable fragment header: */
1374
1375         msg_dbg(hdr, ">FRAGMENTING>");
1376         msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1377                  TIPC_OK, INT_H_SIZE, msg_destnode(hdr));
1378         msg_set_link_selector(&fragm_hdr, sender->publ.ref);
1379         msg_set_size(&fragm_hdr, max_pkt);
1380         msg_set_fragm_no(&fragm_hdr, 1);
1381
1382         /* Prepare header of first fragment: */
1383
1384         buf_chain = buf = buf_acquire(max_pkt);
1385         if (!buf)
1386                 return -ENOMEM;
1387         buf->next = NULL;
1388         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1389         hsz = msg_hdr_sz(hdr);
1390         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1391         msg_dbg(buf_msg(buf), ">BUILD>");
1392
1393         /* Chop up message: */
1394
1395         fragm_crs = INT_H_SIZE + hsz;
1396         fragm_rest = fragm_sz - hsz;
1397
1398         do {            /* For all sections */
1399                 u32 sz;
1400
1401                 if (!sect_rest) {
1402                         sect_rest = msg_sect[++curr_sect].iov_len;
1403                         sect_crs = (const unchar *)msg_sect[curr_sect].iov_base;
1404                 }
1405
1406                 if (sect_rest < fragm_rest)
1407                         sz = sect_rest;
1408                 else
1409                         sz = fragm_rest;
1410
1411                 if (likely(!sender->user_port)) {
1412                         if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1413 error:
1414                                 for (; buf_chain; buf_chain = buf) {
1415                                         buf = buf_chain->next;
1416                                         buf_discard(buf_chain);
1417                                 }
1418                                 return -EFAULT;
1419                         }
1420                 } else
1421                         skb_copy_to_linear_data_offset(buf, fragm_crs,
1422                                                        sect_crs, sz);
1423                 sect_crs += sz;
1424                 sect_rest -= sz;
1425                 fragm_crs += sz;
1426                 fragm_rest -= sz;
1427                 rest -= sz;
1428
1429                 if (!fragm_rest && rest) {
1430
1431                         /* Initiate new fragment: */
1432                         if (rest <= fragm_sz) {
1433                                 fragm_sz = rest;
1434                                 msg_set_type(&fragm_hdr,LAST_FRAGMENT);
1435                         } else {
1436                                 msg_set_type(&fragm_hdr, FRAGMENT);
1437                         }
1438                         msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1439                         msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1440                         prev = buf;
1441                         buf = buf_acquire(fragm_sz + INT_H_SIZE);
1442                         if (!buf)
1443                                 goto error;
1444
1445                         buf->next = NULL;
1446                         prev->next = buf;
1447                         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1448                         fragm_crs = INT_H_SIZE;
1449                         fragm_rest = fragm_sz;
1450                         msg_dbg(buf_msg(buf),"  >BUILD>");
1451                 }
1452         }
1453         while (rest > 0);
1454
1455         /*
1456          * Now we have a buffer chain. Select a link and check
1457          * that packet size is still OK
1458          */
1459         node = tipc_node_select(destaddr, sender->publ.ref & 1);
1460         if (likely(node)) {
1461                 tipc_node_lock(node);
1462                 l_ptr = node->active_links[sender->publ.ref & 1];
1463                 if (!l_ptr) {
1464                         tipc_node_unlock(node);
1465                         goto reject;
1466                 }
1467                 if (link_max_pkt(l_ptr) < max_pkt) {
1468                         sender->publ.max_pkt = link_max_pkt(l_ptr);
1469                         tipc_node_unlock(node);
1470                         for (; buf_chain; buf_chain = buf) {
1471                                 buf = buf_chain->next;
1472                                 buf_discard(buf_chain);
1473                         }
1474                         goto again;
1475                 }
1476         } else {
1477 reject:
1478                 for (; buf_chain; buf_chain = buf) {
1479                         buf = buf_chain->next;
1480                         buf_discard(buf_chain);
1481                 }
1482                 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1483                                                  TIPC_ERR_NO_NODE);
1484         }
1485
1486         /* Append whole chain to send queue: */
1487
1488         buf = buf_chain;
1489         l_ptr->long_msg_seq_no = mod(l_ptr->long_msg_seq_no + 1);
1490         if (!l_ptr->next_out)
1491                 l_ptr->next_out = buf_chain;
1492         l_ptr->stats.sent_fragmented++;
1493         while (buf) {
1494                 struct sk_buff *next = buf->next;
1495                 struct tipc_msg *msg = buf_msg(buf);
1496
1497                 l_ptr->stats.sent_fragments++;
1498                 msg_set_long_msgno(msg, l_ptr->long_msg_seq_no);
1499                 link_add_to_outqueue(l_ptr, buf, msg);
1500                 msg_dbg(msg, ">ADD>");
1501                 buf = next;
1502         }
1503
1504         /* Send it, if possible: */
1505
1506         tipc_link_push_queue(l_ptr);
1507         tipc_node_unlock(node);
1508         return dsz;
1509 }
1510
1511 /*
1512  * tipc_link_push_packet: Push one unsent packet to the media
1513  */
1514 u32 tipc_link_push_packet(struct link *l_ptr)
1515 {
1516         struct sk_buff *buf = l_ptr->first_out;
1517         u32 r_q_size = l_ptr->retransm_queue_size;
1518         u32 r_q_head = l_ptr->retransm_queue_head;
1519
1520         /* Step to position where retransmission failed, if any,    */
1521         /* consider that buffers may have been released in meantime */
1522
1523         if (r_q_size && buf) {
1524                 u32 last = lesser(mod(r_q_head + r_q_size),
1525                                   link_last_sent(l_ptr));
1526                 u32 first = msg_seqno(buf_msg(buf));
1527
1528                 while (buf && less(first, r_q_head)) {
1529                         first = mod(first + 1);
1530                         buf = buf->next;
1531                 }
1532                 l_ptr->retransm_queue_head = r_q_head = first;
1533                 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1534         }
1535
1536         /* Continue retransmission now, if there is anything: */
1537
1538         if (r_q_size && buf && !skb_cloned(buf)) {
1539                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1540                 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1541                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1542                         msg_dbg(buf_msg(buf), ">DEF-RETR>");
1543                         l_ptr->retransm_queue_head = mod(++r_q_head);
1544                         l_ptr->retransm_queue_size = --r_q_size;
1545                         l_ptr->stats.retransmitted++;
1546                         return TIPC_OK;
1547                 } else {
1548                         l_ptr->stats.bearer_congs++;
1549                         msg_dbg(buf_msg(buf), "|>DEF-RETR>");
1550                         return PUSH_FAILED;
1551                 }
1552         }
1553
1554         /* Send deferred protocol message, if any: */
1555
1556         buf = l_ptr->proto_msg_queue;
1557         if (buf) {
1558                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1559                 msg_set_bcast_ack(buf_msg(buf),l_ptr->owner->bclink.last_in);
1560                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1561                         msg_dbg(buf_msg(buf), ">DEF-PROT>");
1562                         l_ptr->unacked_window = 0;
1563                         buf_discard(buf);
1564                         l_ptr->proto_msg_queue = NULL;
1565                         return TIPC_OK;
1566                 } else {
1567                         msg_dbg(buf_msg(buf), "|>DEF-PROT>");
1568                         l_ptr->stats.bearer_congs++;
1569                         return PUSH_FAILED;
1570                 }
1571         }
1572
1573         /* Send one deferred data message, if send window not full: */
1574
1575         buf = l_ptr->next_out;
1576         if (buf) {
1577                 struct tipc_msg *msg = buf_msg(buf);
1578                 u32 next = msg_seqno(msg);
1579                 u32 first = msg_seqno(buf_msg(l_ptr->first_out));
1580
1581                 if (mod(next - first) < l_ptr->queue_limit[0]) {
1582                         msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1583                         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1584                         if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1585                                 if (msg_user(msg) == MSG_BUNDLER)
1586                                         msg_set_type(msg, CLOSED_MSG);
1587                                 msg_dbg(msg, ">PUSH-DATA>");
1588                                 l_ptr->next_out = buf->next;
1589                                 return TIPC_OK;
1590                         } else {
1591                                 msg_dbg(msg, "|PUSH-DATA|");
1592                                 l_ptr->stats.bearer_congs++;
1593                                 return PUSH_FAILED;
1594                         }
1595                 }
1596         }
1597         return PUSH_FINISHED;
1598 }
1599
1600 /*
1601  * push_queue(): push out the unsent messages of a link where
1602  *               congestion has abated. Node is locked
1603  */
1604 void tipc_link_push_queue(struct link *l_ptr)
1605 {
1606         u32 res;
1607
1608         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr))
1609                 return;
1610
1611         do {
1612                 res = tipc_link_push_packet(l_ptr);
1613         }
1614         while (res == TIPC_OK);
1615         if (res == PUSH_FAILED)
1616                 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1617 }
1618
1619 static void link_reset_all(unsigned long addr)
1620 {
1621         struct node *n_ptr;
1622         char addr_string[16];
1623         u32 i;
1624
1625         read_lock_bh(&tipc_net_lock);
1626         n_ptr = tipc_node_find((u32)addr);
1627         if (!n_ptr) {
1628                 read_unlock_bh(&tipc_net_lock);
1629                 return; /* node no longer exists */
1630         }
1631
1632         tipc_node_lock(n_ptr);
1633
1634         warn("Resetting all links to %s\n",
1635              addr_string_fill(addr_string, n_ptr->addr));
1636
1637         for (i = 0; i < MAX_BEARERS; i++) {
1638                 if (n_ptr->links[i]) {
1639                         link_print(n_ptr->links[i], TIPC_OUTPUT,
1640                                    "Resetting link\n");
1641                         tipc_link_reset(n_ptr->links[i]);
1642                 }
1643         }
1644
1645         tipc_node_unlock(n_ptr);
1646         read_unlock_bh(&tipc_net_lock);
1647 }
1648
1649 static void link_retransmit_failure(struct link *l_ptr, struct sk_buff *buf)
1650 {
1651         struct tipc_msg *msg = buf_msg(buf);
1652
1653         warn("Retransmission failure on link <%s>\n", l_ptr->name);
1654         tipc_msg_print(TIPC_OUTPUT, msg, ">RETR-FAIL>");
1655
1656         if (l_ptr->addr) {
1657
1658                 /* Handle failure on standard link */
1659
1660                 link_print(l_ptr, TIPC_OUTPUT, "Resetting link\n");
1661                 tipc_link_reset(l_ptr);
1662
1663         } else {
1664
1665                 /* Handle failure on broadcast link */
1666
1667                 struct node *n_ptr;
1668                 char addr_string[16];
1669
1670                 tipc_printf(TIPC_OUTPUT, "Msg seq number: %u,  ", msg_seqno(msg));
1671                 tipc_printf(TIPC_OUTPUT, "Outstanding acks: %lu\n",
1672                                      (unsigned long) TIPC_SKB_CB(buf)->handle);
1673
1674                 n_ptr = l_ptr->owner->next;
1675                 tipc_node_lock(n_ptr);
1676
1677                 addr_string_fill(addr_string, n_ptr->addr);
1678                 tipc_printf(TIPC_OUTPUT, "Multicast link info for %s\n", addr_string);
1679                 tipc_printf(TIPC_OUTPUT, "Supported: %d,  ", n_ptr->bclink.supported);
1680                 tipc_printf(TIPC_OUTPUT, "Acked: %u\n", n_ptr->bclink.acked);
1681                 tipc_printf(TIPC_OUTPUT, "Last in: %u,  ", n_ptr->bclink.last_in);
1682                 tipc_printf(TIPC_OUTPUT, "Gap after: %u,  ", n_ptr->bclink.gap_after);
1683                 tipc_printf(TIPC_OUTPUT, "Gap to: %u\n", n_ptr->bclink.gap_to);
1684                 tipc_printf(TIPC_OUTPUT, "Nack sync: %u\n\n", n_ptr->bclink.nack_sync);
1685
1686                 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1687
1688                 tipc_node_unlock(n_ptr);
1689
1690                 l_ptr->stale_count = 0;
1691         }
1692 }
1693
1694 void tipc_link_retransmit(struct link *l_ptr, struct sk_buff *buf,
1695                           u32 retransmits)
1696 {
1697         struct tipc_msg *msg;
1698
1699         if (!buf)
1700                 return;
1701
1702         msg = buf_msg(buf);
1703
1704         dbg("Retransmitting %u in link %x\n", retransmits, l_ptr);
1705
1706         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1707                 if (!skb_cloned(buf)) {
1708                         msg_dbg(msg, ">NO_RETR->BCONG>");
1709                         dbg_print_link(l_ptr, "   ");
1710                         l_ptr->retransm_queue_head = msg_seqno(msg);
1711                         l_ptr->retransm_queue_size = retransmits;
1712                         return;
1713                 } else {
1714                         /* Don't retransmit if driver already has the buffer */
1715                 }
1716         } else {
1717                 /* Detect repeated retransmit failures on uncongested bearer */
1718
1719                 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1720                         if (++l_ptr->stale_count > 100) {
1721                                 link_retransmit_failure(l_ptr, buf);
1722                                 return;
1723                         }
1724                 } else {
1725                         l_ptr->last_retransmitted = msg_seqno(msg);
1726                         l_ptr->stale_count = 1;
1727                 }
1728         }
1729
1730         while (retransmits && (buf != l_ptr->next_out) && buf && !skb_cloned(buf)) {
1731                 msg = buf_msg(buf);
1732                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1733                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1734                 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1735                         msg_dbg(buf_msg(buf), ">RETR>");
1736                         buf = buf->next;
1737                         retransmits--;
1738                         l_ptr->stats.retransmitted++;
1739                 } else {
1740                         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1741                         l_ptr->stats.bearer_congs++;
1742                         l_ptr->retransm_queue_head = msg_seqno(buf_msg(buf));
1743                         l_ptr->retransm_queue_size = retransmits;
1744                         return;
1745                 }
1746         }
1747
1748         l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1749 }
1750
1751 /*
1752  * link_recv_non_seq: Receive packets which are outside
1753  *                    the link sequence flow
1754  */
1755
1756 static void link_recv_non_seq(struct sk_buff *buf)
1757 {
1758         struct tipc_msg *msg = buf_msg(buf);
1759
1760         if (msg_user(msg) ==  LINK_CONFIG)
1761                 tipc_disc_recv_msg(buf);
1762         else
1763                 tipc_bclink_recv_pkt(buf);
1764 }
1765
1766 /**
1767  * link_insert_deferred_queue - insert deferred messages back into receive chain
1768  */
1769
1770 static struct sk_buff *link_insert_deferred_queue(struct link *l_ptr,
1771                                                   struct sk_buff *buf)
1772 {
1773         u32 seq_no;
1774
1775         if (l_ptr->oldest_deferred_in == NULL)
1776                 return buf;
1777
1778         seq_no = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1779         if (seq_no == mod(l_ptr->next_in_no)) {
1780                 l_ptr->newest_deferred_in->next = buf;
1781                 buf = l_ptr->oldest_deferred_in;
1782                 l_ptr->oldest_deferred_in = NULL;
1783                 l_ptr->deferred_inqueue_sz = 0;
1784         }
1785         return buf;
1786 }
1787
1788 /**
1789  * link_recv_buf_validate - validate basic format of received message
1790  *
1791  * This routine ensures a TIPC message has an acceptable header, and at least
1792  * as much data as the header indicates it should.  The routine also ensures
1793  * that the entire message header is stored in the main fragment of the message
1794  * buffer, to simplify future access to message header fields.
1795  *
1796  * Note: Having extra info present in the message header or data areas is OK.
1797  * TIPC will ignore the excess, under the assumption that it is optional info
1798  * introduced by a later release of the protocol.
1799  */
1800
1801 static int link_recv_buf_validate(struct sk_buff *buf)
1802 {
1803         static u32 min_data_hdr_size[8] = {
1804                 SHORT_H_SIZE, MCAST_H_SIZE, LONG_H_SIZE, DIR_MSG_H_SIZE,
1805                 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1806                 };
1807
1808         struct tipc_msg *msg;
1809         u32 tipc_hdr[2];
1810         u32 size;
1811         u32 hdr_size;
1812         u32 min_hdr_size;
1813
1814         if (unlikely(buf->len < MIN_H_SIZE))
1815                 return 0;
1816
1817         msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1818         if (msg == NULL)
1819                 return 0;
1820
1821         if (unlikely(msg_version(msg) != TIPC_VERSION))
1822                 return 0;
1823
1824         size = msg_size(msg);
1825         hdr_size = msg_hdr_sz(msg);
1826         min_hdr_size = msg_isdata(msg) ?
1827                 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1828
1829         if (unlikely((hdr_size < min_hdr_size) ||
1830                      (size < hdr_size) ||
1831                      (buf->len < size) ||
1832                      (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1833                 return 0;
1834
1835         return pskb_may_pull(buf, hdr_size);
1836 }
1837
1838 void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *tb_ptr)
1839 {
1840         read_lock_bh(&tipc_net_lock);
1841         while (head) {
1842                 struct bearer *b_ptr;
1843                 struct node *n_ptr;
1844                 struct link *l_ptr;
1845                 struct sk_buff *crs;
1846                 struct sk_buff *buf = head;
1847                 struct tipc_msg *msg;
1848                 u32 seq_no;
1849                 u32 ackd;
1850                 u32 released = 0;
1851                 int type;
1852
1853                 b_ptr = (struct bearer *)tb_ptr;
1854                 TIPC_SKB_CB(buf)->handle = b_ptr;
1855
1856                 head = head->next;
1857
1858                 /* Ensure message is well-formed */
1859
1860                 if (unlikely(!link_recv_buf_validate(buf)))
1861                         goto cont;
1862
1863                 /* Ensure message data is a single contiguous unit */
1864
1865                 if (unlikely(buf_linearize(buf))) {
1866                         goto cont;
1867                 }
1868
1869                 /* Handle arrival of a non-unicast link message */
1870
1871                 msg = buf_msg(buf);
1872
1873                 if (unlikely(msg_non_seq(msg))) {
1874                         link_recv_non_seq(buf);
1875                         continue;
1876                 }
1877
1878                 if (unlikely(!msg_short(msg) &&
1879                              (msg_destnode(msg) != tipc_own_addr)))
1880                         goto cont;
1881
1882                 /* Locate unicast link endpoint that should handle message */
1883
1884                 n_ptr = tipc_node_find(msg_prevnode(msg));
1885                 if (unlikely(!n_ptr))
1886                         goto cont;
1887                 tipc_node_lock(n_ptr);
1888
1889                 l_ptr = n_ptr->links[b_ptr->identity];
1890                 if (unlikely(!l_ptr)) {
1891                         tipc_node_unlock(n_ptr);
1892                         goto cont;
1893                 }
1894
1895                 /* Validate message sequence number info */
1896
1897                 seq_no = msg_seqno(msg);
1898                 ackd = msg_ack(msg);
1899
1900                 /* Release acked messages */
1901
1902                 if (less(n_ptr->bclink.acked, msg_bcast_ack(msg))) {
1903                         if (tipc_node_is_up(n_ptr) && n_ptr->bclink.supported)
1904                                 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1905                 }
1906
1907                 crs = l_ptr->first_out;
1908                 while ((crs != l_ptr->next_out) &&
1909                        less_eq(msg_seqno(buf_msg(crs)), ackd)) {
1910                         struct sk_buff *next = crs->next;
1911
1912                         buf_discard(crs);
1913                         crs = next;
1914                         released++;
1915                 }
1916                 if (released) {
1917                         l_ptr->first_out = crs;
1918                         l_ptr->out_queue_size -= released;
1919                 }
1920
1921                 /* Try sending any messages link endpoint has pending */
1922
1923                 if (unlikely(l_ptr->next_out))
1924                         tipc_link_push_queue(l_ptr);
1925                 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1926                         tipc_link_wakeup_ports(l_ptr, 0);
1927                 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1928                         l_ptr->stats.sent_acks++;
1929                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1930                 }
1931
1932                 /* Now (finally!) process the incoming message */
1933
1934 protocol_check:
1935                 if (likely(link_working_working(l_ptr))) {
1936                         if (likely(seq_no == mod(l_ptr->next_in_no))) {
1937                                 l_ptr->next_in_no++;
1938                                 if (unlikely(l_ptr->oldest_deferred_in))
1939                                         head = link_insert_deferred_queue(l_ptr,
1940                                                                           head);
1941                                 if (likely(msg_is_dest(msg, tipc_own_addr))) {
1942 deliver:
1943                                         if (likely(msg_isdata(msg))) {
1944                                                 tipc_node_unlock(n_ptr);
1945                                                 tipc_port_recv_msg(buf);
1946                                                 continue;
1947                                         }
1948                                         switch (msg_user(msg)) {
1949                                         case MSG_BUNDLER:
1950                                                 l_ptr->stats.recv_bundles++;
1951                                                 l_ptr->stats.recv_bundled +=
1952                                                         msg_msgcnt(msg);
1953                                                 tipc_node_unlock(n_ptr);
1954                                                 tipc_link_recv_bundle(buf);
1955                                                 continue;
1956                                         case ROUTE_DISTRIBUTOR:
1957                                                 tipc_node_unlock(n_ptr);
1958                                                 tipc_cltr_recv_routing_table(buf);
1959                                                 continue;
1960                                         case NAME_DISTRIBUTOR:
1961                                                 tipc_node_unlock(n_ptr);
1962                                                 tipc_named_recv(buf);
1963                                                 continue;
1964                                         case CONN_MANAGER:
1965                                                 tipc_node_unlock(n_ptr);
1966                                                 tipc_port_recv_proto_msg(buf);
1967                                                 continue;
1968                                         case MSG_FRAGMENTER:
1969                                                 l_ptr->stats.recv_fragments++;
1970                                                 if (tipc_link_recv_fragment(&l_ptr->defragm_buf,
1971                                                                             &buf, &msg)) {
1972                                                         l_ptr->stats.recv_fragmented++;
1973                                                         goto deliver;
1974                                                 }
1975                                                 break;
1976                                         case CHANGEOVER_PROTOCOL:
1977                                                 type = msg_type(msg);
1978                                                 if (link_recv_changeover_msg(&l_ptr, &buf)) {
1979                                                         msg = buf_msg(buf);
1980                                                         seq_no = msg_seqno(msg);
1981                                                         TIPC_SKB_CB(buf)->handle
1982                                                                 = b_ptr;
1983                                                         if (type == ORIGINAL_MSG)
1984                                                                 goto deliver;
1985                                                         goto protocol_check;
1986                                                 }
1987                                                 break;
1988                                         }
1989                                 }
1990                                 tipc_node_unlock(n_ptr);
1991                                 tipc_net_route_msg(buf);
1992                                 continue;
1993                         }
1994                         link_handle_out_of_seq_msg(l_ptr, buf);
1995                         head = link_insert_deferred_queue(l_ptr, head);
1996                         tipc_node_unlock(n_ptr);
1997                         continue;
1998                 }
1999
2000                 if (msg_user(msg) == LINK_PROTOCOL) {
2001                         link_recv_proto_msg(l_ptr, buf);
2002                         head = link_insert_deferred_queue(l_ptr, head);
2003                         tipc_node_unlock(n_ptr);
2004                         continue;
2005                 }
2006                 msg_dbg(msg,"NSEQ<REC<");
2007                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2008
2009                 if (link_working_working(l_ptr)) {
2010                         /* Re-insert in front of queue */
2011                         msg_dbg(msg,"RECV-REINS:");
2012                         buf->next = head;
2013                         head = buf;
2014                         tipc_node_unlock(n_ptr);
2015                         continue;
2016                 }
2017                 tipc_node_unlock(n_ptr);
2018 cont:
2019                 buf_discard(buf);
2020         }
2021         read_unlock_bh(&tipc_net_lock);
2022 }
2023
2024 /*
2025  * link_defer_buf(): Sort a received out-of-sequence packet
2026  *                   into the deferred reception queue.
2027  * Returns the increase of the queue length,i.e. 0 or 1
2028  */
2029
2030 u32 tipc_link_defer_pkt(struct sk_buff **head,
2031                         struct sk_buff **tail,
2032                         struct sk_buff *buf)
2033 {
2034         struct sk_buff *prev = NULL;
2035         struct sk_buff *crs = *head;
2036         u32 seq_no = msg_seqno(buf_msg(buf));
2037
2038         buf->next = NULL;
2039
2040         /* Empty queue ? */
2041         if (*head == NULL) {
2042                 *head = *tail = buf;
2043                 return 1;
2044         }
2045
2046         /* Last ? */
2047         if (less(msg_seqno(buf_msg(*tail)), seq_no)) {
2048                 (*tail)->next = buf;
2049                 *tail = buf;
2050                 return 1;
2051         }
2052
2053         /* Scan through queue and sort it in */
2054         do {
2055                 struct tipc_msg *msg = buf_msg(crs);
2056
2057                 if (less(seq_no, msg_seqno(msg))) {
2058                         buf->next = crs;
2059                         if (prev)
2060                                 prev->next = buf;
2061                         else
2062                                 *head = buf;
2063                         return 1;
2064                 }
2065                 if (seq_no == msg_seqno(msg)) {
2066                         break;
2067                 }
2068                 prev = crs;
2069                 crs = crs->next;
2070         }
2071         while (crs);
2072
2073         /* Message is a duplicate of an existing message */
2074
2075         buf_discard(buf);
2076         return 0;
2077 }
2078
2079 /**
2080  * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
2081  */
2082
2083 static void link_handle_out_of_seq_msg(struct link *l_ptr,
2084                                        struct sk_buff *buf)
2085 {
2086         u32 seq_no = msg_seqno(buf_msg(buf));
2087
2088         if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
2089                 link_recv_proto_msg(l_ptr, buf);
2090                 return;
2091         }
2092
2093         dbg("rx OOS msg: seq_no %u, expecting %u (%u)\n",
2094             seq_no, mod(l_ptr->next_in_no), l_ptr->next_in_no);
2095
2096         /* Record OOS packet arrival (force mismatch on next timeout) */
2097
2098         l_ptr->checkpoint--;
2099
2100         /*
2101          * Discard packet if a duplicate; otherwise add it to deferred queue
2102          * and notify peer of gap as per protocol specification
2103          */
2104
2105         if (less(seq_no, mod(l_ptr->next_in_no))) {
2106                 l_ptr->stats.duplicates++;
2107                 buf_discard(buf);
2108                 return;
2109         }
2110
2111         if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
2112                                 &l_ptr->newest_deferred_in, buf)) {
2113                 l_ptr->deferred_inqueue_sz++;
2114                 l_ptr->stats.deferred_recv++;
2115                 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
2116                         tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
2117         } else
2118                 l_ptr->stats.duplicates++;
2119 }
2120
2121 /*
2122  * Send protocol message to the other endpoint.
2123  */
2124 void tipc_link_send_proto_msg(struct link *l_ptr, u32 msg_typ, int probe_msg,
2125                               u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
2126 {
2127         struct sk_buff *buf = NULL;
2128         struct tipc_msg *msg = l_ptr->pmsg;
2129         u32 msg_size = sizeof(l_ptr->proto_msg);
2130
2131         if (link_blocked(l_ptr))
2132                 return;
2133         msg_set_type(msg, msg_typ);
2134         msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
2135         msg_set_bcast_ack(msg, mod(l_ptr->owner->bclink.last_in));
2136         msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
2137
2138         if (msg_typ == STATE_MSG) {
2139                 u32 next_sent = mod(l_ptr->next_out_no);
2140
2141                 if (!tipc_link_is_up(l_ptr))
2142                         return;
2143                 if (l_ptr->next_out)
2144                         next_sent = msg_seqno(buf_msg(l_ptr->next_out));
2145                 msg_set_next_sent(msg, next_sent);
2146                 if (l_ptr->oldest_deferred_in) {
2147                         u32 rec = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
2148                         gap = mod(rec - mod(l_ptr->next_in_no));
2149                 }
2150                 msg_set_seq_gap(msg, gap);
2151                 if (gap)
2152                         l_ptr->stats.sent_nacks++;
2153                 msg_set_link_tolerance(msg, tolerance);
2154                 msg_set_linkprio(msg, priority);
2155                 msg_set_max_pkt(msg, ack_mtu);
2156                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
2157                 msg_set_probe(msg, probe_msg != 0);
2158                 if (probe_msg) {
2159                         u32 mtu = l_ptr->max_pkt;
2160
2161                         if ((mtu < l_ptr->max_pkt_target) &&
2162                             link_working_working(l_ptr) &&
2163                             l_ptr->fsm_msg_cnt) {
2164                                 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2165                                 if (l_ptr->max_pkt_probes == 10) {
2166                                         l_ptr->max_pkt_target = (msg_size - 4);
2167                                         l_ptr->max_pkt_probes = 0;
2168                                         msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2169                                 }
2170                                 l_ptr->max_pkt_probes++;
2171                         }
2172
2173                         l_ptr->stats.sent_probes++;
2174                 }
2175                 l_ptr->stats.sent_states++;
2176         } else {                /* RESET_MSG or ACTIVATE_MSG */
2177                 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
2178                 msg_set_seq_gap(msg, 0);
2179                 msg_set_next_sent(msg, 1);
2180                 msg_set_link_tolerance(msg, l_ptr->tolerance);
2181                 msg_set_linkprio(msg, l_ptr->priority);
2182                 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
2183         }
2184
2185         if (tipc_node_has_redundant_links(l_ptr->owner)) {
2186                 msg_set_redundant_link(msg);
2187         } else {
2188                 msg_clear_redundant_link(msg);
2189         }
2190         msg_set_linkprio(msg, l_ptr->priority);
2191
2192         /* Ensure sequence number will not fit : */
2193
2194         msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
2195
2196         /* Congestion? */
2197
2198         if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
2199                 if (!l_ptr->proto_msg_queue) {
2200                         l_ptr->proto_msg_queue =
2201                                 buf_acquire(sizeof(l_ptr->proto_msg));
2202                 }
2203                 buf = l_ptr->proto_msg_queue;
2204                 if (!buf)
2205                         return;
2206                 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2207                 return;
2208         }
2209         msg_set_timestamp(msg, jiffies_to_msecs(jiffies));
2210
2211         /* Message can be sent */
2212
2213         msg_dbg(msg, ">>");
2214
2215         buf = buf_acquire(msg_size);
2216         if (!buf)
2217                 return;
2218
2219         skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2220         msg_set_size(buf_msg(buf), msg_size);
2221
2222         if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
2223                 l_ptr->unacked_window = 0;
2224                 buf_discard(buf);
2225                 return;
2226         }
2227
2228         /* New congestion */
2229         tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
2230         l_ptr->proto_msg_queue = buf;
2231         l_ptr->stats.bearer_congs++;
2232 }
2233
2234 /*
2235  * Receive protocol message :
2236  * Note that network plane id propagates through the network, and may
2237  * change at any time. The node with lowest address rules
2238  */
2239
2240 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf)
2241 {
2242         u32 rec_gap = 0;
2243         u32 max_pkt_info;
2244         u32 max_pkt_ack;
2245         u32 msg_tol;
2246         struct tipc_msg *msg = buf_msg(buf);
2247
2248         dbg("AT(%u):", jiffies_to_msecs(jiffies));
2249         msg_dbg(msg, "<<");
2250         if (link_blocked(l_ptr))
2251                 goto exit;
2252
2253         /* record unnumbered packet arrival (force mismatch on next timeout) */
2254
2255         l_ptr->checkpoint--;
2256
2257         if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
2258                 if (tipc_own_addr > msg_prevnode(msg))
2259                         l_ptr->b_ptr->net_plane = msg_net_plane(msg);
2260
2261         l_ptr->owner->permit_changeover = msg_redundant_link(msg);
2262
2263         switch (msg_type(msg)) {
2264
2265         case RESET_MSG:
2266                 if (!link_working_unknown(l_ptr) && l_ptr->peer_session) {
2267                         if (msg_session(msg) == l_ptr->peer_session) {
2268                                 dbg("Duplicate RESET: %u<->%u\n",
2269                                     msg_session(msg), l_ptr->peer_session);
2270                                 break; /* duplicate: ignore */
2271                         }
2272                 }
2273                 /* fall thru' */
2274         case ACTIVATE_MSG:
2275                 /* Update link settings according other endpoint's values */
2276
2277                 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2278
2279                 if ((msg_tol = msg_link_tolerance(msg)) &&
2280                     (msg_tol > l_ptr->tolerance))
2281                         link_set_supervision_props(l_ptr, msg_tol);
2282
2283                 if (msg_linkprio(msg) > l_ptr->priority)
2284                         l_ptr->priority = msg_linkprio(msg);
2285
2286                 max_pkt_info = msg_max_pkt(msg);
2287                 if (max_pkt_info) {
2288                         if (max_pkt_info < l_ptr->max_pkt_target)
2289                                 l_ptr->max_pkt_target = max_pkt_info;
2290                         if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2291                                 l_ptr->max_pkt = l_ptr->max_pkt_target;
2292                 } else {
2293                         l_ptr->max_pkt = l_ptr->max_pkt_target;
2294                 }
2295                 l_ptr->owner->bclink.supported = (max_pkt_info != 0);
2296
2297                 link_state_event(l_ptr, msg_type(msg));
2298
2299                 l_ptr->peer_session = msg_session(msg);
2300                 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2301
2302                 /* Synchronize broadcast sequence numbers */
2303                 if (!tipc_node_has_redundant_links(l_ptr->owner)) {
2304                         l_ptr->owner->bclink.last_in = mod(msg_last_bcast(msg));
2305                 }
2306                 break;
2307         case STATE_MSG:
2308
2309                 if ((msg_tol = msg_link_tolerance(msg)))
2310                         link_set_supervision_props(l_ptr, msg_tol);
2311
2312                 if (msg_linkprio(msg) &&
2313                     (msg_linkprio(msg) != l_ptr->priority)) {
2314                         warn("Resetting link <%s>, priority change %u->%u\n",
2315                              l_ptr->name, l_ptr->priority, msg_linkprio(msg));
2316                         l_ptr->priority = msg_linkprio(msg);
2317                         tipc_link_reset(l_ptr); /* Enforce change to take effect */
2318                         break;
2319                 }
2320                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2321                 l_ptr->stats.recv_states++;
2322                 if (link_reset_unknown(l_ptr))
2323                         break;
2324
2325                 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2326                         rec_gap = mod(msg_next_sent(msg) -
2327                                       mod(l_ptr->next_in_no));
2328                 }
2329
2330                 max_pkt_ack = msg_max_pkt(msg);
2331                 if (max_pkt_ack > l_ptr->max_pkt) {
2332                         dbg("Link <%s> updated MTU %u -> %u\n",
2333                             l_ptr->name, l_ptr->max_pkt, max_pkt_ack);
2334                         l_ptr->max_pkt = max_pkt_ack;
2335                         l_ptr->max_pkt_probes = 0;
2336                 }
2337
2338                 max_pkt_ack = 0;
2339                 if (msg_probe(msg)) {
2340                         l_ptr->stats.recv_probes++;
2341                         if (msg_size(msg) > sizeof(l_ptr->proto_msg)) {
2342                                 max_pkt_ack = msg_size(msg);
2343                         }
2344                 }
2345
2346                 /* Protocol message before retransmits, reduce loss risk */
2347
2348                 tipc_bclink_check_gap(l_ptr->owner, msg_last_bcast(msg));
2349
2350                 if (rec_gap || (msg_probe(msg))) {
2351                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2352                                                  0, rec_gap, 0, 0, max_pkt_ack);
2353                 }
2354                 if (msg_seq_gap(msg)) {
2355                         msg_dbg(msg, "With Gap:");
2356                         l_ptr->stats.recv_nacks++;
2357                         tipc_link_retransmit(l_ptr, l_ptr->first_out,
2358                                              msg_seq_gap(msg));
2359                 }
2360                 break;
2361         default:
2362                 msg_dbg(buf_msg(buf), "<DISCARDING UNKNOWN<");
2363         }
2364 exit:
2365         buf_discard(buf);
2366 }
2367
2368
2369 /*
2370  * tipc_link_tunnel(): Send one message via a link belonging to
2371  * another bearer. Owner node is locked.
2372  */
2373 void tipc_link_tunnel(struct link *l_ptr,
2374                       struct tipc_msg *tunnel_hdr,
2375                       struct tipc_msg  *msg,
2376                       u32 selector)
2377 {
2378         struct link *tunnel;
2379         struct sk_buff *buf;
2380         u32 length = msg_size(msg);
2381
2382         tunnel = l_ptr->owner->active_links[selector & 1];
2383         if (!tipc_link_is_up(tunnel)) {
2384                 warn("Link changeover error, "
2385                      "tunnel link no longer available\n");
2386                 return;
2387         }
2388         msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2389         buf = buf_acquire(length + INT_H_SIZE);
2390         if (!buf) {
2391                 warn("Link changeover error, "
2392                      "unable to send tunnel msg\n");
2393                 return;
2394         }
2395         skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
2396         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
2397         dbg("%c->%c:", l_ptr->b_ptr->net_plane, tunnel->b_ptr->net_plane);
2398         msg_dbg(buf_msg(buf), ">SEND>");
2399         tipc_link_send_buf(tunnel, buf);
2400 }
2401
2402
2403
2404 /*
2405  * changeover(): Send whole message queue via the remaining link
2406  *               Owner node is locked.
2407  */
2408
2409 void tipc_link_changeover(struct link *l_ptr)
2410 {
2411         u32 msgcount = l_ptr->out_queue_size;
2412         struct sk_buff *crs = l_ptr->first_out;
2413         struct link *tunnel = l_ptr->owner->active_links[0];
2414         struct tipc_msg tunnel_hdr;
2415         int split_bundles;
2416
2417         if (!tunnel)
2418                 return;
2419
2420         if (!l_ptr->owner->permit_changeover) {
2421                 warn("Link changeover error, "
2422                      "peer did not permit changeover\n");
2423                 return;
2424         }
2425
2426         msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2427                  ORIGINAL_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
2428         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2429         msg_set_msgcnt(&tunnel_hdr, msgcount);
2430         dbg("Link changeover requires %u tunnel messages\n", msgcount);
2431
2432         if (!l_ptr->first_out) {
2433                 struct sk_buff *buf;
2434
2435                 buf = buf_acquire(INT_H_SIZE);
2436                 if (buf) {
2437                         skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2438                         msg_set_size(&tunnel_hdr, INT_H_SIZE);
2439                         dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2440                             tunnel->b_ptr->net_plane);
2441                         msg_dbg(&tunnel_hdr, "EMPTY>SEND>");
2442                         tipc_link_send_buf(tunnel, buf);
2443                 } else {
2444                         warn("Link changeover error, "
2445                              "unable to send changeover msg\n");
2446                 }
2447                 return;
2448         }
2449
2450         split_bundles = (l_ptr->owner->active_links[0] !=
2451                          l_ptr->owner->active_links[1]);
2452
2453         while (crs) {
2454                 struct tipc_msg *msg = buf_msg(crs);
2455
2456                 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2457                         struct tipc_msg *m = msg_get_wrapped(msg);
2458                         unchar* pos = (unchar*)m;
2459
2460                         msgcount = msg_msgcnt(msg);
2461                         while (msgcount--) {
2462                                 msg_set_seqno(m,msg_seqno(msg));
2463                                 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2464                                                  msg_link_selector(m));
2465                                 pos += align(msg_size(m));
2466                                 m = (struct tipc_msg *)pos;
2467                         }
2468                 } else {
2469                         tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2470                                          msg_link_selector(msg));
2471                 }
2472                 crs = crs->next;
2473         }
2474 }
2475
2476 void tipc_link_send_duplicate(struct link *l_ptr, struct link *tunnel)
2477 {
2478         struct sk_buff *iter;
2479         struct tipc_msg tunnel_hdr;
2480
2481         msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2482                  DUPLICATE_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
2483         msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2484         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2485         iter = l_ptr->first_out;
2486         while (iter) {
2487                 struct sk_buff *outbuf;
2488                 struct tipc_msg *msg = buf_msg(iter);
2489                 u32 length = msg_size(msg);
2490
2491                 if (msg_user(msg) == MSG_BUNDLER)
2492                         msg_set_type(msg, CLOSED_MSG);
2493                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));   /* Update */
2494                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2495                 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2496                 outbuf = buf_acquire(length + INT_H_SIZE);
2497                 if (outbuf == NULL) {
2498                         warn("Link changeover error, "
2499                              "unable to send duplicate msg\n");
2500                         return;
2501                 }
2502                 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2503                 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2504                                                length);
2505                 dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2506                     tunnel->b_ptr->net_plane);
2507                 msg_dbg(buf_msg(outbuf), ">SEND>");
2508                 tipc_link_send_buf(tunnel, outbuf);
2509                 if (!tipc_link_is_up(l_ptr))
2510                         return;
2511                 iter = iter->next;
2512         }
2513 }
2514
2515
2516
2517 /**
2518  * buf_extract - extracts embedded TIPC message from another message
2519  * @skb: encapsulating message buffer
2520  * @from_pos: offset to extract from
2521  *
2522  * Returns a new message buffer containing an embedded message.  The
2523  * encapsulating message itself is left unchanged.
2524  */
2525
2526 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2527 {
2528         struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2529         u32 size = msg_size(msg);
2530         struct sk_buff *eb;
2531
2532         eb = buf_acquire(size);
2533         if (eb)
2534                 skb_copy_to_linear_data(eb, msg, size);
2535         return eb;
2536 }
2537
2538 /*
2539  *  link_recv_changeover_msg(): Receive tunneled packet sent
2540  *  via other link. Node is locked. Return extracted buffer.
2541  */
2542
2543 static int link_recv_changeover_msg(struct link **l_ptr,
2544                                     struct sk_buff **buf)
2545 {
2546         struct sk_buff *tunnel_buf = *buf;
2547         struct link *dest_link;
2548         struct tipc_msg *msg;
2549         struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2550         u32 msg_typ = msg_type(tunnel_msg);
2551         u32 msg_count = msg_msgcnt(tunnel_msg);
2552
2553         dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2554         if (!dest_link) {
2555                 msg_dbg(tunnel_msg, "NOLINK/<REC<");
2556                 goto exit;
2557         }
2558         if (dest_link == *l_ptr) {
2559                 err("Unexpected changeover message on link <%s>\n",
2560                     (*l_ptr)->name);
2561                 goto exit;
2562         }
2563         dbg("%c<-%c:", dest_link->b_ptr->net_plane,
2564             (*l_ptr)->b_ptr->net_plane);
2565         *l_ptr = dest_link;
2566         msg = msg_get_wrapped(tunnel_msg);
2567
2568         if (msg_typ == DUPLICATE_MSG) {
2569                 if (less(msg_seqno(msg), mod(dest_link->next_in_no))) {
2570                         msg_dbg(tunnel_msg, "DROP/<REC<");
2571                         goto exit;
2572                 }
2573                 *buf = buf_extract(tunnel_buf,INT_H_SIZE);
2574                 if (*buf == NULL) {
2575                         warn("Link changeover error, duplicate msg dropped\n");
2576                         goto exit;
2577                 }
2578                 msg_dbg(tunnel_msg, "TNL<REC<");
2579                 buf_discard(tunnel_buf);
2580                 return 1;
2581         }
2582
2583         /* First original message ?: */
2584
2585         if (tipc_link_is_up(dest_link)) {
2586                 msg_dbg(tunnel_msg, "UP/FIRST/<REC<");
2587                 info("Resetting link <%s>, changeover initiated by peer\n",
2588                      dest_link->name);
2589                 tipc_link_reset(dest_link);
2590                 dest_link->exp_msg_count = msg_count;
2591                 dbg("Expecting %u tunnelled messages\n", msg_count);
2592                 if (!msg_count)
2593                         goto exit;
2594         } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2595                 msg_dbg(tunnel_msg, "BLK/FIRST/<REC<");
2596                 dest_link->exp_msg_count = msg_count;
2597                 dbg("Expecting %u tunnelled messages\n", msg_count);
2598                 if (!msg_count)
2599                         goto exit;
2600         }
2601
2602         /* Receive original message */
2603
2604         if (dest_link->exp_msg_count == 0) {
2605                 warn("Link switchover error, "
2606                      "got too many tunnelled messages\n");
2607                 msg_dbg(tunnel_msg, "OVERDUE/DROP/<REC<");
2608                 dbg_print_link(dest_link, "LINK:");
2609                 goto exit;
2610         }
2611         dest_link->exp_msg_count--;
2612         if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2613                 msg_dbg(tunnel_msg, "DROP/DUPL/<REC<");
2614                 goto exit;
2615         } else {
2616                 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2617                 if (*buf != NULL) {
2618                         msg_dbg(tunnel_msg, "TNL<REC<");
2619                         buf_discard(tunnel_buf);
2620                         return 1;
2621                 } else {
2622                         warn("Link changeover error, original msg dropped\n");
2623                 }
2624         }
2625 exit:
2626         *buf = NULL;
2627         buf_discard(tunnel_buf);
2628         return 0;
2629 }
2630
2631 /*
2632  *  Bundler functionality:
2633  */
2634 void tipc_link_recv_bundle(struct sk_buff *buf)
2635 {
2636         u32 msgcount = msg_msgcnt(buf_msg(buf));
2637         u32 pos = INT_H_SIZE;
2638         struct sk_buff *obuf;
2639
2640         msg_dbg(buf_msg(buf), "<BNDL<: ");
2641         while (msgcount--) {
2642                 obuf = buf_extract(buf, pos);
2643                 if (obuf == NULL) {
2644                         warn("Link unable to unbundle message(s)\n");
2645                         break;
2646                 }
2647                 pos += align(msg_size(buf_msg(obuf)));
2648                 msg_dbg(buf_msg(obuf), "     /");
2649                 tipc_net_route_msg(obuf);
2650         }
2651         buf_discard(buf);
2652 }
2653
2654 /*
2655  *  Fragmentation/defragmentation:
2656  */
2657
2658
2659 /*
2660  * tipc_link_send_long_buf: Entry for buffers needing fragmentation.
2661  * The buffer is complete, inclusive total message length.
2662  * Returns user data length.
2663  */
2664 int tipc_link_send_long_buf(struct link *l_ptr, struct sk_buff *buf)
2665 {
2666         struct tipc_msg *inmsg = buf_msg(buf);
2667         struct tipc_msg fragm_hdr;
2668         u32 insize = msg_size(inmsg);
2669         u32 dsz = msg_data_sz(inmsg);
2670         unchar *crs = buf->data;
2671         u32 rest = insize;
2672         u32 pack_sz = link_max_pkt(l_ptr);
2673         u32 fragm_sz = pack_sz - INT_H_SIZE;
2674         u32 fragm_no = 1;
2675         u32 destaddr = msg_destnode(inmsg);
2676
2677         if (msg_short(inmsg))
2678                 destaddr = l_ptr->addr;
2679
2680         if (msg_routed(inmsg))
2681                 msg_set_prevnode(inmsg, tipc_own_addr);
2682
2683         /* Prepare reusable fragment header: */
2684
2685         msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2686                  TIPC_OK, INT_H_SIZE, destaddr);
2687         msg_set_link_selector(&fragm_hdr, msg_link_selector(inmsg));
2688         msg_set_long_msgno(&fragm_hdr, mod(l_ptr->long_msg_seq_no++));
2689         msg_set_fragm_no(&fragm_hdr, fragm_no);
2690         l_ptr->stats.sent_fragmented++;
2691
2692         /* Chop up message: */
2693
2694         while (rest > 0) {
2695                 struct sk_buff *fragm;
2696
2697                 if (rest <= fragm_sz) {
2698                         fragm_sz = rest;
2699                         msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2700                 }
2701                 fragm = buf_acquire(fragm_sz + INT_H_SIZE);
2702                 if (fragm == NULL) {
2703                         warn("Link unable to fragment message\n");
2704                         dsz = -ENOMEM;
2705                         goto exit;
2706                 }
2707                 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2708                 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2709                 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2710                                                fragm_sz);
2711                 /*  Send queued messages first, if any: */
2712
2713                 l_ptr->stats.sent_fragments++;
2714                 tipc_link_send_buf(l_ptr, fragm);
2715                 if (!tipc_link_is_up(l_ptr))
2716                         return dsz;
2717                 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
2718                 rest -= fragm_sz;
2719                 crs += fragm_sz;
2720                 msg_set_type(&fragm_hdr, FRAGMENT);
2721         }
2722 exit:
2723         buf_discard(buf);
2724         return dsz;
2725 }
2726
2727 /*
2728  * A pending message being re-assembled must store certain values
2729  * to handle subsequent fragments correctly. The following functions
2730  * help storing these values in unused, available fields in the
2731  * pending message. This makes dynamic memory allocation unecessary.
2732  */
2733
2734 static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2735 {
2736         msg_set_seqno(buf_msg(buf), seqno);
2737 }
2738
2739 static u32 get_fragm_size(struct sk_buff *buf)
2740 {
2741         return msg_ack(buf_msg(buf));
2742 }
2743
2744 static void set_fragm_size(struct sk_buff *buf, u32 sz)
2745 {
2746         msg_set_ack(buf_msg(buf), sz);
2747 }
2748
2749 static u32 get_expected_frags(struct sk_buff *buf)
2750 {
2751         return msg_bcast_ack(buf_msg(buf));
2752 }
2753
2754 static void set_expected_frags(struct sk_buff *buf, u32 exp)
2755 {
2756         msg_set_bcast_ack(buf_msg(buf), exp);
2757 }
2758
2759 static u32 get_timer_cnt(struct sk_buff *buf)
2760 {
2761         return msg_reroute_cnt(buf_msg(buf));
2762 }
2763
2764 static void incr_timer_cnt(struct sk_buff *buf)
2765 {
2766         msg_incr_reroute_cnt(buf_msg(buf));
2767 }
2768
2769 /*
2770  * tipc_link_recv_fragment(): Called with node lock on. Returns
2771  * the reassembled buffer if message is complete.
2772  */
2773 int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2774                             struct tipc_msg **m)
2775 {
2776         struct sk_buff *prev = NULL;
2777         struct sk_buff *fbuf = *fb;
2778         struct tipc_msg *fragm = buf_msg(fbuf);
2779         struct sk_buff *pbuf = *pending;
2780         u32 long_msg_seq_no = msg_long_msgno(fragm);
2781
2782         *fb = NULL;
2783         msg_dbg(fragm,"FRG<REC<");
2784
2785         /* Is there an incomplete message waiting for this fragment? */
2786
2787         while (pbuf && ((msg_seqno(buf_msg(pbuf)) != long_msg_seq_no)
2788                         || (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2789                 prev = pbuf;
2790                 pbuf = pbuf->next;
2791         }
2792
2793         if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2794                 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2795                 u32 msg_sz = msg_size(imsg);
2796                 u32 fragm_sz = msg_data_sz(fragm);
2797                 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2798                 u32 max =  TIPC_MAX_USER_MSG_SIZE + LONG_H_SIZE;
2799                 if (msg_type(imsg) == TIPC_MCAST_MSG)
2800                         max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2801                 if (msg_size(imsg) > max) {
2802                         msg_dbg(fragm,"<REC<Oversized: ");
2803                         buf_discard(fbuf);
2804                         return 0;
2805                 }
2806                 pbuf = buf_acquire(msg_size(imsg));
2807                 if (pbuf != NULL) {
2808                         pbuf->next = *pending;
2809                         *pending = pbuf;
2810                         skb_copy_to_linear_data(pbuf, imsg,
2811                                                 msg_data_sz(fragm));
2812                         /*  Prepare buffer for subsequent fragments. */
2813
2814                         set_long_msg_seqno(pbuf, long_msg_seq_no);
2815                         set_fragm_size(pbuf,fragm_sz);
2816                         set_expected_frags(pbuf,exp_fragm_cnt - 1);
2817                 } else {
2818                         warn("Link unable to reassemble fragmented message\n");
2819                 }
2820                 buf_discard(fbuf);
2821                 return 0;
2822         } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2823                 u32 dsz = msg_data_sz(fragm);
2824                 u32 fsz = get_fragm_size(pbuf);
2825                 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2826                 u32 exp_frags = get_expected_frags(pbuf) - 1;
2827                 skb_copy_to_linear_data_offset(pbuf, crs,
2828                                                msg_data(fragm), dsz);
2829                 buf_discard(fbuf);
2830
2831                 /* Is message complete? */
2832
2833                 if (exp_frags == 0) {
2834                         if (prev)
2835                                 prev->next = pbuf->next;
2836                         else
2837                                 *pending = pbuf->next;
2838                         msg_reset_reroute_cnt(buf_msg(pbuf));
2839                         *fb = pbuf;
2840                         *m = buf_msg(pbuf);
2841                         return 1;
2842                 }
2843                 set_expected_frags(pbuf,exp_frags);
2844                 return 0;
2845         }
2846         dbg(" Discarding orphan fragment %x\n",fbuf);
2847         msg_dbg(fragm,"ORPHAN:");
2848         dbg("Pending long buffers:\n");
2849         dbg_print_buf_chain(*pending);
2850         buf_discard(fbuf);
2851         return 0;
2852 }
2853
2854 /**
2855  * link_check_defragm_bufs - flush stale incoming message fragments
2856  * @l_ptr: pointer to link
2857  */
2858
2859 static void link_check_defragm_bufs(struct link *l_ptr)
2860 {
2861         struct sk_buff *prev = NULL;
2862         struct sk_buff *next = NULL;
2863         struct sk_buff *buf = l_ptr->defragm_buf;
2864
2865         if (!buf)
2866                 return;
2867         if (!link_working_working(l_ptr))
2868                 return;
2869         while (buf) {
2870                 u32 cnt = get_timer_cnt(buf);
2871
2872                 next = buf->next;
2873                 if (cnt < 4) {
2874                         incr_timer_cnt(buf);
2875                         prev = buf;
2876                 } else {
2877                         dbg(" Discarding incomplete long buffer\n");
2878                         msg_dbg(buf_msg(buf), "LONG:");
2879                         dbg_print_link(l_ptr, "curr:");
2880                         dbg("Pending long buffers:\n");
2881                         dbg_print_buf_chain(l_ptr->defragm_buf);
2882                         if (prev)
2883                                 prev->next = buf->next;
2884                         else
2885                                 l_ptr->defragm_buf = buf->next;
2886                         buf_discard(buf);
2887                 }
2888                 buf = next;
2889         }
2890 }
2891
2892
2893
2894 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance)
2895 {
2896         l_ptr->tolerance = tolerance;
2897         l_ptr->continuity_interval =
2898                 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2899         l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2900 }
2901
2902
2903 void tipc_link_set_queue_limits(struct link *l_ptr, u32 window)
2904 {
2905         /* Data messages from this node, inclusive FIRST_FRAGM */
2906         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2907         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2908         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2909         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2910         /* Transiting data messages,inclusive FIRST_FRAGM */
2911         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2912         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2913         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2914         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2915         l_ptr->queue_limit[CONN_MANAGER] = 1200;
2916         l_ptr->queue_limit[ROUTE_DISTRIBUTOR] = 1200;
2917         l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2918         l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2919         /* FRAGMENT and LAST_FRAGMENT packets */
2920         l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2921 }
2922
2923 /**
2924  * link_find_link - locate link by name
2925  * @name - ptr to link name string
2926  * @node - ptr to area to be filled with ptr to associated node
2927  *
2928  * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2929  * this also prevents link deletion.
2930  *
2931  * Returns pointer to link (or 0 if invalid link name).
2932  */
2933
2934 static struct link *link_find_link(const char *name, struct node **node)
2935 {
2936         struct link_name link_name_parts;
2937         struct bearer *b_ptr;
2938         struct link *l_ptr;
2939
2940         if (!link_name_validate(name, &link_name_parts))
2941                 return NULL;
2942
2943         b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2944         if (!b_ptr)
2945                 return NULL;
2946
2947         *node = tipc_node_find(link_name_parts.addr_peer);
2948         if (!*node)
2949                 return NULL;
2950
2951         l_ptr = (*node)->links[b_ptr->identity];
2952         if (!l_ptr || strcmp(l_ptr->name, name))
2953                 return NULL;
2954
2955         return l_ptr;
2956 }
2957
2958 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2959                                      u16 cmd)
2960 {
2961         struct tipc_link_config *args;
2962         u32 new_value;
2963         struct link *l_ptr;
2964         struct node *node;
2965         int res;
2966
2967         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2968                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2969
2970         args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2971         new_value = ntohl(args->value);
2972
2973         if (!strcmp(args->name, tipc_bclink_name)) {
2974                 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2975                     (tipc_bclink_set_queue_limits(new_value) == 0))
2976                         return tipc_cfg_reply_none();
2977                 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2978                                                    " (cannot change setting on broadcast link)");
2979         }
2980
2981         read_lock_bh(&tipc_net_lock);
2982         l_ptr = link_find_link(args->name, &node);
2983         if (!l_ptr) {
2984                 read_unlock_bh(&tipc_net_lock);
2985                 return tipc_cfg_reply_error_string("link not found");
2986         }
2987
2988         tipc_node_lock(node);
2989         res = -EINVAL;
2990         switch (cmd) {
2991         case TIPC_CMD_SET_LINK_TOL:
2992                 if ((new_value >= TIPC_MIN_LINK_TOL) &&
2993                     (new_value <= TIPC_MAX_LINK_TOL)) {
2994                         link_set_supervision_props(l_ptr, new_value);
2995                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2996                                                  0, 0, new_value, 0, 0);
2997                         res = TIPC_OK;
2998                 }
2999                 break;
3000         case TIPC_CMD_SET_LINK_PRI:
3001                 if ((new_value >= TIPC_MIN_LINK_PRI) &&
3002                     (new_value <= TIPC_MAX_LINK_PRI)) {
3003                         l_ptr->priority = new_value;
3004                         tipc_link_send_proto_msg(l_ptr, STATE_MSG,
3005                                                  0, 0, 0, new_value, 0);
3006                         res = TIPC_OK;
3007                 }
3008                 break;
3009         case TIPC_CMD_SET_LINK_WINDOW:
3010                 if ((new_value >= TIPC_MIN_LINK_WIN) &&
3011                     (new_value <= TIPC_MAX_LINK_WIN)) {
3012                         tipc_link_set_queue_limits(l_ptr, new_value);
3013                         res = TIPC_OK;
3014                 }
3015                 break;
3016         }
3017         tipc_node_unlock(node);
3018
3019         read_unlock_bh(&tipc_net_lock);
3020         if (res)
3021                 return tipc_cfg_reply_error_string("cannot change link setting");
3022
3023         return tipc_cfg_reply_none();
3024 }
3025
3026 /**
3027  * link_reset_statistics - reset link statistics
3028  * @l_ptr: pointer to link
3029  */
3030
3031 static void link_reset_statistics(struct link *l_ptr)
3032 {
3033         memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
3034         l_ptr->stats.sent_info = l_ptr->next_out_no;
3035         l_ptr->stats.recv_info = l_ptr->next_in_no;
3036 }
3037
3038 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
3039 {
3040         char *link_name;
3041         struct link *l_ptr;
3042         struct node *node;
3043
3044         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3045                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3046
3047         link_name = (char *)TLV_DATA(req_tlv_area);
3048         if (!strcmp(link_name, tipc_bclink_name)) {
3049                 if (tipc_bclink_reset_stats())
3050                         return tipc_cfg_reply_error_string("link not found");
3051                 return tipc_cfg_reply_none();
3052         }
3053
3054         read_lock_bh(&tipc_net_lock);
3055         l_ptr = link_find_link(link_name, &node);
3056         if (!l_ptr) {
3057                 read_unlock_bh(&tipc_net_lock);
3058                 return tipc_cfg_reply_error_string("link not found");
3059         }
3060
3061         tipc_node_lock(node);
3062         link_reset_statistics(l_ptr);
3063         tipc_node_unlock(node);
3064         read_unlock_bh(&tipc_net_lock);
3065         return tipc_cfg_reply_none();
3066 }
3067
3068 /**
3069  * percent - convert count to a percentage of total (rounding up or down)
3070  */
3071
3072 static u32 percent(u32 count, u32 total)
3073 {
3074         return (count * 100 + (total / 2)) / total;
3075 }
3076
3077 /**
3078  * tipc_link_stats - print link statistics
3079  * @name: link name
3080  * @buf: print buffer area
3081  * @buf_size: size of print buffer area
3082  *
3083  * Returns length of print buffer data string (or 0 if error)
3084  */
3085
3086 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
3087 {
3088         struct print_buf pb;
3089         struct link *l_ptr;
3090         struct node *node;
3091         char *status;
3092         u32 profile_total = 0;
3093
3094         if (!strcmp(name, tipc_bclink_name))
3095                 return tipc_bclink_stats(buf, buf_size);
3096
3097         tipc_printbuf_init(&pb, buf, buf_size);
3098
3099         read_lock_bh(&tipc_net_lock);
3100         l_ptr = link_find_link(name, &node);
3101         if (!l_ptr) {
3102                 read_unlock_bh(&tipc_net_lock);
3103                 return 0;
3104         }
3105         tipc_node_lock(node);
3106
3107         if (tipc_link_is_active(l_ptr))
3108                 status = "ACTIVE";
3109         else if (tipc_link_is_up(l_ptr))
3110                 status = "STANDBY";
3111         else
3112                 status = "DEFUNCT";
3113         tipc_printf(&pb, "Link <%s>\n"
3114                          "  %s  MTU:%u  Priority:%u  Tolerance:%u ms"
3115                          "  Window:%u packets\n",
3116                     l_ptr->name, status, link_max_pkt(l_ptr),
3117                     l_ptr->priority, l_ptr->tolerance, l_ptr->queue_limit[0]);
3118         tipc_printf(&pb, "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
3119                     l_ptr->next_in_no - l_ptr->stats.recv_info,
3120                     l_ptr->stats.recv_fragments,
3121                     l_ptr->stats.recv_fragmented,
3122                     l_ptr->stats.recv_bundles,
3123                     l_ptr->stats.recv_bundled);
3124         tipc_printf(&pb, "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
3125                     l_ptr->next_out_no - l_ptr->stats.sent_info,
3126                     l_ptr->stats.sent_fragments,
3127                     l_ptr->stats.sent_fragmented,
3128                     l_ptr->stats.sent_bundles,
3129                     l_ptr->stats.sent_bundled);
3130         profile_total = l_ptr->stats.msg_length_counts;
3131         if (!profile_total)
3132                 profile_total = 1;
3133         tipc_printf(&pb, "  TX profile sample:%u packets  average:%u octets\n"
3134                          "  0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
3135                          "-16354:%u%% -32768:%u%% -66000:%u%%\n",
3136                     l_ptr->stats.msg_length_counts,
3137                     l_ptr->stats.msg_lengths_total / profile_total,
3138                     percent(l_ptr->stats.msg_length_profile[0], profile_total),
3139                     percent(l_ptr->stats.msg_length_profile[1], profile_total),
3140                     percent(l_ptr->stats.msg_length_profile[2], profile_total),
3141                     percent(l_ptr->stats.msg_length_profile[3], profile_total),
3142                     percent(l_ptr->stats.msg_length_profile[4], profile_total),
3143                     percent(l_ptr->stats.msg_length_profile[5], profile_total),
3144                     percent(l_ptr->stats.msg_length_profile[6], profile_total));
3145         tipc_printf(&pb, "  RX states:%u probes:%u naks:%u defs:%u dups:%u\n",
3146                     l_ptr->stats.recv_states,
3147                     l_ptr->stats.recv_probes,
3148                     l_ptr->stats.recv_nacks,
3149                     l_ptr->stats.deferred_recv,
3150                     l_ptr->stats.duplicates);
3151         tipc_printf(&pb, "  TX states:%u probes:%u naks:%u acks:%u dups:%u\n",
3152                     l_ptr->stats.sent_states,
3153                     l_ptr->stats.sent_probes,
3154                     l_ptr->stats.sent_nacks,
3155                     l_ptr->stats.sent_acks,
3156                     l_ptr->stats.retransmitted);
3157         tipc_printf(&pb, "  Congestion bearer:%u link:%u  Send queue max:%u avg:%u\n",
3158                     l_ptr->stats.bearer_congs,
3159                     l_ptr->stats.link_congs,
3160                     l_ptr->stats.max_queue_sz,
3161                     l_ptr->stats.queue_sz_counts
3162                     ? (l_ptr->stats.accu_queue_sz / l_ptr->stats.queue_sz_counts)
3163                     : 0);
3164
3165         tipc_node_unlock(node);
3166         read_unlock_bh(&tipc_net_lock);
3167         return tipc_printbuf_validate(&pb);
3168 }
3169
3170 #define MAX_LINK_STATS_INFO 2000
3171
3172 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
3173 {
3174         struct sk_buff *buf;
3175         struct tlv_desc *rep_tlv;
3176         int str_len;
3177
3178         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3179                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3180
3181         buf = tipc_cfg_reply_alloc(TLV_SPACE(MAX_LINK_STATS_INFO));
3182         if (!buf)
3183                 return NULL;
3184
3185         rep_tlv = (struct tlv_desc *)buf->data;
3186
3187         str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
3188                                   (char *)TLV_DATA(rep_tlv), MAX_LINK_STATS_INFO);
3189         if (!str_len) {
3190                 buf_discard(buf);
3191                 return tipc_cfg_reply_error_string("link not found");
3192         }
3193
3194         skb_put(buf, TLV_SPACE(str_len));
3195         TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
3196
3197         return buf;
3198 }
3199
3200 #if 0
3201 int link_control(const char *name, u32 op, u32 val)
3202 {
3203         int res = -EINVAL;
3204         struct link *l_ptr;
3205         u32 bearer_id;
3206         struct node * node;
3207         u32 a;
3208
3209         a = link_name2addr(name, &bearer_id);
3210         read_lock_bh(&tipc_net_lock);
3211         node = tipc_node_find(a);
3212         if (node) {
3213                 tipc_node_lock(node);
3214                 l_ptr = node->links[bearer_id];
3215                 if (l_ptr) {
3216                         if (op == TIPC_REMOVE_LINK) {
3217                                 struct bearer *b_ptr = l_ptr->b_ptr;
3218                                 spin_lock_bh(&b_ptr->publ.lock);
3219                                 tipc_link_delete(l_ptr);
3220                                 spin_unlock_bh(&b_ptr->publ.lock);
3221                         }
3222                         if (op == TIPC_CMD_BLOCK_LINK) {
3223                                 tipc_link_reset(l_ptr);
3224                                 l_ptr->blocked = 1;
3225                         }
3226                         if (op == TIPC_CMD_UNBLOCK_LINK) {
3227                                 l_ptr->blocked = 0;
3228                         }
3229                         res = TIPC_OK;
3230                 }
3231                 tipc_node_unlock(node);
3232         }
3233         read_unlock_bh(&tipc_net_lock);
3234         return res;
3235 }
3236 #endif
3237
3238 /**
3239  * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
3240  * @dest: network address of destination node
3241  * @selector: used to select from set of active links
3242  *
3243  * If no active link can be found, uses default maximum packet size.
3244  */
3245
3246 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
3247 {
3248         struct node *n_ptr;
3249         struct link *l_ptr;
3250         u32 res = MAX_PKT_DEFAULT;
3251
3252         if (dest == tipc_own_addr)
3253                 return MAX_MSG_SIZE;
3254
3255         read_lock_bh(&tipc_net_lock);
3256         n_ptr = tipc_node_select(dest, selector);
3257         if (n_ptr) {
3258                 tipc_node_lock(n_ptr);
3259                 l_ptr = n_ptr->active_links[selector & 1];
3260                 if (l_ptr)
3261                         res = link_max_pkt(l_ptr);
3262                 tipc_node_unlock(n_ptr);
3263         }
3264         read_unlock_bh(&tipc_net_lock);
3265         return res;
3266 }
3267
3268 #if 0
3269 static void link_dump_rec_queue(struct link *l_ptr)
3270 {
3271         struct sk_buff *crs;
3272
3273         if (!l_ptr->oldest_deferred_in) {
3274                 info("Reception queue empty\n");
3275                 return;
3276         }
3277         info("Contents of Reception queue:\n");
3278         crs = l_ptr->oldest_deferred_in;
3279         while (crs) {
3280                 if (crs->data == (void *)0x0000a3a3) {
3281                         info("buffer %x invalid\n", crs);
3282                         return;
3283                 }
3284                 msg_dbg(buf_msg(crs), "In rec queue: \n");
3285                 crs = crs->next;
3286         }
3287 }
3288 #endif
3289
3290 static void link_dump_send_queue(struct link *l_ptr)
3291 {
3292         if (l_ptr->next_out) {
3293                 info("\nContents of unsent queue:\n");
3294                 dbg_print_buf_chain(l_ptr->next_out);
3295         }
3296         info("\nContents of send queue:\n");
3297         if (l_ptr->first_out) {
3298                 dbg_print_buf_chain(l_ptr->first_out);
3299         }
3300         info("Empty send queue\n");
3301 }
3302
3303 static void link_print(struct link *l_ptr, struct print_buf *buf,
3304                        const char *str)
3305 {
3306         tipc_printf(buf, str);
3307         if (link_reset_reset(l_ptr) || link_reset_unknown(l_ptr))
3308                 return;
3309         tipc_printf(buf, "Link %x<%s>:",
3310                     l_ptr->addr, l_ptr->b_ptr->publ.name);
3311         tipc_printf(buf, ": NXO(%u):", mod(l_ptr->next_out_no));
3312         tipc_printf(buf, "NXI(%u):", mod(l_ptr->next_in_no));
3313         tipc_printf(buf, "SQUE");
3314         if (l_ptr->first_out) {
3315                 tipc_printf(buf, "[%u..", msg_seqno(buf_msg(l_ptr->first_out)));
3316                 if (l_ptr->next_out)
3317                         tipc_printf(buf, "%u..",
3318                                     msg_seqno(buf_msg(l_ptr->next_out)));
3319                 tipc_printf(buf, "%u]",
3320                             msg_seqno(buf_msg
3321                                       (l_ptr->last_out)), l_ptr->out_queue_size);
3322                 if ((mod(msg_seqno(buf_msg(l_ptr->last_out)) -
3323                          msg_seqno(buf_msg(l_ptr->first_out)))
3324                      != (l_ptr->out_queue_size - 1))
3325                     || (l_ptr->last_out->next != NULL)) {
3326                         tipc_printf(buf, "\nSend queue inconsistency\n");
3327                         tipc_printf(buf, "first_out= %x ", l_ptr->first_out);
3328                         tipc_printf(buf, "next_out= %x ", l_ptr->next_out);
3329                         tipc_printf(buf, "last_out= %x ", l_ptr->last_out);
3330                         link_dump_send_queue(l_ptr);
3331                 }
3332         } else
3333                 tipc_printf(buf, "[]");
3334         tipc_printf(buf, "SQSIZ(%u)", l_ptr->out_queue_size);
3335         if (l_ptr->oldest_deferred_in) {
3336                 u32 o = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
3337                 u32 n = msg_seqno(buf_msg(l_ptr->newest_deferred_in));
3338                 tipc_printf(buf, ":RQUE[%u..%u]", o, n);
3339                 if (l_ptr->deferred_inqueue_sz != mod((n + 1) - o)) {
3340                         tipc_printf(buf, ":RQSIZ(%u)",
3341                                     l_ptr->deferred_inqueue_sz);
3342                 }
3343         }
3344         if (link_working_unknown(l_ptr))
3345                 tipc_printf(buf, ":WU");
3346         if (link_reset_reset(l_ptr))
3347                 tipc_printf(buf, ":RR");
3348         if (link_reset_unknown(l_ptr))
3349                 tipc_printf(buf, ":RU");
3350         if (link_working_working(l_ptr))
3351                 tipc_printf(buf, ":WW");
3352         tipc_printf(buf, "\n");
3353 }
3354