net/tipc/link.c

   1 /*
   2  * net/tipc/link.c: TIPC link code
   3  *
   4  * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
   5  * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
   6  * All rights reserved.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions are met:
  10  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the names of the copyright holders nor the names of its
  17  *    contributors may be used to endorse or promote products derived from
  18  *    this software without specific prior written permission.
  19  *
  20  * Alternatively, this software may be distributed under the terms of the
  21  * GNU General Public License ("GPL") version 2 as published by the Free
  22  * Software Foundation.
  23  *
  24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  34  * POSSIBILITY OF SUCH DAMAGE.
  35  */
  36
  37 #include "core.h"
  38 #include "subscr.h"
  39 #include "link.h"
  40 #include "bcast.h"
  41 #include "socket.h"
  42 #include "name_distr.h"
  43 #include "discover.h"
  44 #include "netlink.h"
  45 #include "monitor.h"
  46 #include "trace.h"
  47 #include "crypto.h"
  48
  49 #include <linux/pkt_sched.h>
  50
  51 struct tipc_stats {
  52         u32 sent_pkts;
  53         u32 recv_pkts;
  54         u32 sent_states;
  55         u32 recv_states;
  56         u32 sent_probes;
  57         u32 recv_probes;
  58         u32 sent_nacks;
  59         u32 recv_nacks;
  60         u32 sent_acks;
  61         u32 sent_bundled;
  62         u32 sent_bundles;
  63         u32 recv_bundled;
  64         u32 recv_bundles;
  65         u32 retransmitted;
  66         u32 sent_fragmented;
  67         u32 sent_fragments;
  68         u32 recv_fragmented;
  69         u32 recv_fragments;
  70         u32 link_congs;         /* # port sends blocked by congestion */
  71         u32 deferred_recv;
  72         u32 duplicates;
  73         u32 max_queue_sz;       /* send queue size high water mark */
  74         u32 accu_queue_sz;      /* used for send queue size profiling */
  75         u32 queue_sz_counts;    /* used for send queue size profiling */
  76         u32 msg_length_counts;  /* used for message length profiling */
  77         u32 msg_lengths_total;  /* used for message length profiling */
  78         u32 msg_length_profile[7]; /* used for msg. length profiling */
  79 };
  80
  81 /**
  82  * struct tipc_link - TIPC link data structure
  83  * @addr: network address of link's peer node
  84  * @name: link name character string
  85  * @media_addr: media address to use when sending messages over link
  86  * @timer: link timer
  87  * @net: pointer to namespace struct
  88  * @refcnt: reference counter for permanent references (owner node & timer)
  89  * @peer_session: link session # being used by peer end of link
  90  * @peer_bearer_id: bearer id used by link's peer endpoint
  91  * @bearer_id: local bearer id used by link
  92  * @tolerance: minimum link continuity loss needed to reset link [in ms]
  93  * @abort_limit: # of unacknowledged continuity probes needed to reset link
  94  * @state: current state of link FSM
  95  * @peer_caps: bitmap describing capabilities of peer node
  96  * @silent_intv_cnt: # of timer intervals without any reception from peer
  97  * @proto_msg: template for control messages generated by link
  98  * @pmsg: convenience pointer to "proto_msg" field
  99  * @priority: current link priority
 100  * @net_plane: current link network plane ('A' through 'H')
 101  * @mon_state: cookie with information needed by link monitor
 102  * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
 103  * @exp_msg_count: # of tunnelled messages expected during link changeover
 104  * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
 105  * @mtu: current maximum packet size for this link
 106  * @advertised_mtu: advertised own mtu when link is being established
 107  * @transmitq: queue for sent, non-acked messages
 108  * @backlogq: queue for messages waiting to be sent
 109  * @snt_nxt: next sequence number to use for outbound messages
 110  * @ackers: # of peers that needs to ack each packet before it can be released
 111  * @acked: # last packet acked by a certain peer. Used for broadcast.
 112  * @rcv_nxt: next sequence number to expect for inbound messages
 113  * @deferred_queue: deferred queue saved OOS b'cast message received from node
 114  * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
 115  * @inputq: buffer queue for messages to be delivered upwards
 116  * @namedq: buffer queue for name table messages to be delivered upwards
 117  * @next_out: ptr to first unsent outbound message in queue
 118  * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
 119  * @long_msg_seq_no: next identifier to use for outbound fragmented messages
 120  * @reasm_buf: head of partially reassembled inbound message fragments
 121  * @bc_rcvr: marks that this is a broadcast receiver link
 122  * @stats: collects statistics regarding link activity
 123  */
 124 struct tipc_link {
 125         u32 addr;
 126         char name[TIPC_MAX_LINK_NAME];
 127         struct net *net;
 128
 129         /* Management and link supervision data */
 130         u16 peer_session;
 131         u16 session;
 132         u16 snd_nxt_state;
 133         u16 rcv_nxt_state;
 134         u32 peer_bearer_id;
 135         u32 bearer_id;
 136         u32 tolerance;
 137         u32 abort_limit;
 138         u32 state;
 139         u16 peer_caps;
 140         bool in_session;
 141         bool active;
 142         u32 silent_intv_cnt;
 143         char if_name[TIPC_MAX_IF_NAME];
 144         u32 priority;
 145         char net_plane;
 146         struct tipc_mon_state mon_state;
 147         u16 rst_cnt;
 148
 149         /* Failover/synch */
 150         u16 drop_point;
 151         struct sk_buff *failover_reasm_skb;
 152         struct sk_buff_head failover_deferdq;
 153
 154         /* Max packet negotiation */
 155         u16 mtu;
 156         u16 advertised_mtu;
 157
 158         /* Sending */
 159         struct sk_buff_head transmq;
 160         struct sk_buff_head backlogq;
 161         struct {
 162                 u16 len;
 163                 u16 limit;
 164                 struct sk_buff *target_bskb;
 165         } backlog[5];
 166         u16 snd_nxt;
 167         u16 window;
 168
 169         /* Reception */
 170         u16 rcv_nxt;
 171         u32 rcv_unacked;
 172         struct sk_buff_head deferdq;
 173         struct sk_buff_head *inputq;
 174         struct sk_buff_head *namedq;
 175
 176         /* Congestion handling */
 177         struct sk_buff_head wakeupq;
 178
 179         /* Fragmentation/reassembly */
 180         struct sk_buff *reasm_buf;
 181         struct sk_buff *reasm_tnlmsg;
 182
 183         /* Broadcast */
 184         u16 ackers;
 185         u16 acked;
 186         struct tipc_link *bc_rcvlink;
 187         struct tipc_link *bc_sndlink;
 188         u8 nack_state;
 189         bool bc_peer_is_up;
 190
 191         /* Statistics */
 192         struct tipc_stats stats;
 193 };
 194
 195 /*
 196  * Error message prefixes
 197  */
 198 static const char *link_co_err = "Link tunneling error, ";
 199 static const char *link_rst_msg = "Resetting link ";
 200
 201 /* Send states for broadcast NACKs
 202  */
 203 enum {
 204         BC_NACK_SND_CONDITIONAL,
 205         BC_NACK_SND_UNCONDITIONAL,
 206         BC_NACK_SND_SUPPRESS,
 207 };
 208
 209 #define TIPC_BC_RETR_LIM  (jiffies + msecs_to_jiffies(10))
 210 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
 211
 212 /*
 213  * Interval between NACKs when packets arrive out of order
 214  */
 215 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
 216
 217 /* Link FSM states:
 218  */
 219 enum {
 220         LINK_ESTABLISHED     = 0xe,
 221         LINK_ESTABLISHING    = 0xe  << 4,
 222         LINK_RESET           = 0x1  << 8,
 223         LINK_RESETTING       = 0x2  << 12,
 224         LINK_PEER_RESET      = 0xd  << 16,
 225         LINK_FAILINGOVER     = 0xf  << 20,
 226         LINK_SYNCHING        = 0xc  << 24
 227 };
 228
 229 /* Link FSM state checking routines
 230  */
 231 static int link_is_up(struct tipc_link *l)
 232 {
 233         return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
 234 }
 235
 236 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
 237                                struct sk_buff_head *xmitq);
 238 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
 239                                       bool probe_reply, u16 rcvgap,
 240                                       int tolerance, int priority,
 241                                       struct sk_buff_head *xmitq);
 242 static void link_print(struct tipc_link *l, const char *str);
 243 static int tipc_link_build_nack_msg(struct tipc_link *l,
 244                                     struct sk_buff_head *xmitq);
 245 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
 246                                         struct sk_buff_head *xmitq);
 247 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
 248 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data);
 249 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
 250                                      struct tipc_gap_ack_blks *ga,
 251                                      struct sk_buff_head *xmitq);
 252
 253 /*
 254  *  Simple non-static link routines (i.e. referenced outside this file)
 255  */
 256 bool tipc_link_is_up(struct tipc_link *l)
 257 {
 258         return link_is_up(l);
 259 }
 260
 261 bool tipc_link_peer_is_down(struct tipc_link *l)
 262 {
 263         return l->state == LINK_PEER_RESET;
 264 }
 265
 266 bool tipc_link_is_reset(struct tipc_link *l)
 267 {
 268         return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
 269 }
 270
 271 bool tipc_link_is_establishing(struct tipc_link *l)
 272 {
 273         return l->state == LINK_ESTABLISHING;
 274 }
 275
 276 bool tipc_link_is_synching(struct tipc_link *l)
 277 {
 278         return l->state == LINK_SYNCHING;
 279 }
 280
 281 bool tipc_link_is_failingover(struct tipc_link *l)
 282 {
 283         return l->state == LINK_FAILINGOVER;
 284 }
 285
 286 bool tipc_link_is_blocked(struct tipc_link *l)
 287 {
 288         return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
 289 }
 290
 291 static bool link_is_bc_sndlink(struct tipc_link *l)
 292 {
 293         return !l->bc_sndlink;
 294 }
 295
 296 static bool link_is_bc_rcvlink(struct tipc_link *l)
 297 {
 298         return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
 299 }
 300
 301 void tipc_link_set_active(struct tipc_link *l, bool active)
 302 {
 303         l->active = active;
 304 }
 305
 306 u32 tipc_link_id(struct tipc_link *l)
 307 {
 308         return l->peer_bearer_id << 16 | l->bearer_id;
 309 }
 310
 311 int tipc_link_window(struct tipc_link *l)
 312 {
 313         return l->window;
 314 }
 315
 316 int tipc_link_prio(struct tipc_link *l)
 317 {
 318         return l->priority;
 319 }
 320
 321 unsigned long tipc_link_tolerance(struct tipc_link *l)
 322 {
 323         return l->tolerance;
 324 }
 325
 326 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
 327 {
 328         return l->inputq;
 329 }
 330
 331 char tipc_link_plane(struct tipc_link *l)
 332 {
 333         return l->net_plane;
 334 }
 335
 336 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
 337 {
 338         l->peer_caps = capabilities;
 339 }
 340
 341 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
 342                            struct tipc_link *uc_l,
 343                            struct sk_buff_head *xmitq)
 344 {
 345         struct tipc_link *rcv_l = uc_l->bc_rcvlink;
 346
 347         snd_l->ackers++;
 348         rcv_l->acked = snd_l->snd_nxt - 1;
 349         snd_l->state = LINK_ESTABLISHED;
 350         tipc_link_build_bc_init_msg(uc_l, xmitq);
 351 }
 352
 353 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
 354                               struct tipc_link *rcv_l,
 355                               struct sk_buff_head *xmitq)
 356 {
 357         u16 ack = snd_l->snd_nxt - 1;
 358
 359         snd_l->ackers--;
 360         rcv_l->bc_peer_is_up = true;
 361         rcv_l->state = LINK_ESTABLISHED;
 362         tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
 363         trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
 364         tipc_link_reset(rcv_l);
 365         rcv_l->state = LINK_RESET;
 366         if (!snd_l->ackers) {
 367                 trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
 368                 tipc_link_reset(snd_l);
 369                 snd_l->state = LINK_RESET;
 370                 __skb_queue_purge(xmitq);
 371         }
 372 }
 373
 374 int tipc_link_bc_peers(struct tipc_link *l)
 375 {
 376         return l->ackers;
 377 }
 378
 379 static u16 link_bc_rcv_gap(struct tipc_link *l)
 380 {
 381         struct sk_buff *skb = skb_peek(&l->deferdq);
 382         u16 gap = 0;
 383
 384         if (more(l->snd_nxt, l->rcv_nxt))
 385                 gap = l->snd_nxt - l->rcv_nxt;
 386         if (skb)
 387                 gap = buf_seqno(skb) - l->rcv_nxt;
 388         return gap;
 389 }
 390
 391 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
 392 {
 393         l->mtu = mtu;
 394 }
 395
 396 int tipc_link_mtu(struct tipc_link *l)
 397 {
 398         return l->mtu;
 399 }
 400
 401 int tipc_link_mss(struct tipc_link *l)
 402 {
 403 #ifdef CONFIG_TIPC_CRYPTO
 404         return l->mtu - INT_H_SIZE - EMSG_OVERHEAD;
 405 #else
 406         return l->mtu - INT_H_SIZE;
 407 #endif
 408 }
 409
 410 u16 tipc_link_rcv_nxt(struct tipc_link *l)
 411 {
 412         return l->rcv_nxt;
 413 }
 414
 415 u16 tipc_link_acked(struct tipc_link *l)
 416 {
 417         return l->acked;
 418 }
 419
 420 char *tipc_link_name(struct tipc_link *l)
 421 {
 422         return l->name;
 423 }
 424
 425 u32 tipc_link_state(struct tipc_link *l)
 426 {
 427         return l->state;
 428 }
 429
 430 /**
 431  * tipc_link_create - create a new link
 432  * @n: pointer to associated node
 433  * @if_name: associated interface name
 434  * @bearer_id: id (index) of associated bearer
 435  * @tolerance: link tolerance to be used by link
 436  * @net_plane: network plane (A,B,c..) this link belongs to
 437  * @mtu: mtu to be advertised by link
 438  * @priority: priority to be used by link
 439  * @window: send window to be used by link
 440  * @session: session to be used by link
 441  * @ownnode: identity of own node
 442  * @peer: node id of peer node
 443  * @peer_caps: bitmap describing peer node capabilities
 444  * @bc_sndlink: the namespace global link used for broadcast sending
 445  * @bc_rcvlink: the peer specific link used for broadcast reception
 446  * @inputq: queue to put messages ready for delivery
 447  * @namedq: queue to put binding table update messages ready for delivery
 448  * @link: return value, pointer to put the created link
 449  *
 450  * Returns true if link was created, otherwise false
 451  */
 452 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
 453                       int tolerance, char net_plane, u32 mtu, int priority,
 454                       int window, u32 session, u32 self,
 455                       u32 peer, u8 *peer_id, u16 peer_caps,
 456                       struct tipc_link *bc_sndlink,
 457                       struct tipc_link *bc_rcvlink,
 458                       struct sk_buff_head *inputq,
 459                       struct sk_buff_head *namedq,
 460                       struct tipc_link **link)
 461 {
 462         char peer_str[NODE_ID_STR_LEN] = {0,};
 463         char self_str[NODE_ID_STR_LEN] = {0,};
 464         struct tipc_link *l;
 465
 466         l = kzalloc(sizeof(*l), GFP_ATOMIC);
 467         if (!l)
 468                 return false;
 469         *link = l;
 470         l->session = session;
 471
 472         /* Set link name for unicast links only */
 473         if (peer_id) {
 474                 tipc_nodeid2string(self_str, tipc_own_id(net));
 475                 if (strlen(self_str) > 16)
 476                         sprintf(self_str, "%x", self);
 477                 tipc_nodeid2string(peer_str, peer_id);
 478                 if (strlen(peer_str) > 16)
 479                         sprintf(peer_str, "%x", peer);
 480         }
 481         /* Peer i/f name will be completed by reset/activate message */
 482         snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
 483                  self_str, if_name, peer_str);
 484
 485         strcpy(l->if_name, if_name);
 486         l->addr = peer;
 487         l->peer_caps = peer_caps;
 488         l->net = net;
 489         l->in_session = false;
 490         l->bearer_id = bearer_id;
 491         l->tolerance = tolerance;
 492         if (bc_rcvlink)
 493                 bc_rcvlink->tolerance = tolerance;
 494         l->net_plane = net_plane;
 495         l->advertised_mtu = mtu;
 496         l->mtu = mtu;
 497         l->priority = priority;
 498         tipc_link_set_queue_limits(l, window);
 499         l->ackers = 1;
 500         l->bc_sndlink = bc_sndlink;
 501         l->bc_rcvlink = bc_rcvlink;
 502         l->inputq = inputq;
 503         l->namedq = namedq;
 504         l->state = LINK_RESETTING;
 505         __skb_queue_head_init(&l->transmq);
 506         __skb_queue_head_init(&l->backlogq);
 507         __skb_queue_head_init(&l->deferdq);
 508         __skb_queue_head_init(&l->failover_deferdq);
 509         skb_queue_head_init(&l->wakeupq);
 510         skb_queue_head_init(l->inputq);
 511         return true;
 512 }
 513
 514 /**
 515  * tipc_link_bc_create - create new link to be used for broadcast
 516  * @n: pointer to associated node
 517  * @mtu: mtu to be used initially if no peers
 518  * @window: send window to be used
 519  * @inputq: queue to put messages ready for delivery
 520  * @namedq: queue to put binding table update messages ready for delivery
 521  * @link: return value, pointer to put the created link
 522  *
 523  * Returns true if link was created, otherwise false
 524  */
 525 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
 526                          int mtu, int window, u16 peer_caps,
 527                          struct sk_buff_head *inputq,
 528                          struct sk_buff_head *namedq,
 529                          struct tipc_link *bc_sndlink,
 530                          struct tipc_link **link)
 531 {
 532         struct tipc_link *l;
 533
 534         if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
 535                               0, ownnode, peer, NULL, peer_caps, bc_sndlink,
 536                               NULL, inputq, namedq, link))
 537                 return false;
 538
 539         l = *link;
 540         strcpy(l->name, tipc_bclink_name);
 541         trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
 542         tipc_link_reset(l);
 543         l->state = LINK_RESET;
 544         l->ackers = 0;
 545         l->bc_rcvlink = l;
 546
 547         /* Broadcast send link is always up */
 548         if (link_is_bc_sndlink(l))
 549                 l->state = LINK_ESTABLISHED;
 550
 551         /* Disable replicast if even a single peer doesn't support it */
 552         if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
 553                 tipc_bcast_disable_rcast(net);
 554
 555         return true;
 556 }
 557
 558 /**
 559  * tipc_link_fsm_evt - link finite state machine
 560  * @l: pointer to link
 561  * @evt: state machine event to be processed
 562  */
 563 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
 564 {
 565         int rc = 0;
 566         int old_state = l->state;
 567
 568         switch (l->state) {
 569         case LINK_RESETTING:
 570                 switch (evt) {
 571                 case LINK_PEER_RESET_EVT:
 572                         l->state = LINK_PEER_RESET;
 573                         break;
 574                 case LINK_RESET_EVT:
 575                         l->state = LINK_RESET;
 576                         break;
 577                 case LINK_FAILURE_EVT:
 578                 case LINK_FAILOVER_BEGIN_EVT:
 579                 case LINK_ESTABLISH_EVT:
 580                 case LINK_FAILOVER_END_EVT:
 581                 case LINK_SYNCH_BEGIN_EVT:
 582                 case LINK_SYNCH_END_EVT:
 583                 default:
 584                         goto illegal_evt;
 585                 }
 586                 break;
 587         case LINK_RESET:
 588                 switch (evt) {
 589                 case LINK_PEER_RESET_EVT:
 590                         l->state = LINK_ESTABLISHING;
 591                         break;
 592                 case LINK_FAILOVER_BEGIN_EVT:
 593                         l->state = LINK_FAILINGOVER;
 594                 case LINK_FAILURE_EVT:
 595                 case LINK_RESET_EVT:
 596                 case LINK_ESTABLISH_EVT:
 597                 case LINK_FAILOVER_END_EVT:
 598                         break;
 599                 case LINK_SYNCH_BEGIN_EVT:
 600                 case LINK_SYNCH_END_EVT:
 601                 default:
 602                         goto illegal_evt;
 603                 }
 604                 break;
 605         case LINK_PEER_RESET:
 606                 switch (evt) {
 607                 case LINK_RESET_EVT:
 608                         l->state = LINK_ESTABLISHING;
 609                         break;
 610                 case LINK_PEER_RESET_EVT:
 611                 case LINK_ESTABLISH_EVT:
 612                 case LINK_FAILURE_EVT:
 613                         break;
 614                 case LINK_SYNCH_BEGIN_EVT:
 615                 case LINK_SYNCH_END_EVT:
 616                 case LINK_FAILOVER_BEGIN_EVT:
 617                 case LINK_FAILOVER_END_EVT:
 618                 default:
 619                         goto illegal_evt;
 620                 }
 621                 break;
 622         case LINK_FAILINGOVER:
 623                 switch (evt) {
 624                 case LINK_FAILOVER_END_EVT:
 625                         l->state = LINK_RESET;
 626                         break;
 627                 case LINK_PEER_RESET_EVT:
 628                 case LINK_RESET_EVT:
 629                 case LINK_ESTABLISH_EVT:
 630                 case LINK_FAILURE_EVT:
 631                         break;
 632                 case LINK_FAILOVER_BEGIN_EVT:
 633                 case LINK_SYNCH_BEGIN_EVT:
 634                 case LINK_SYNCH_END_EVT:
 635                 default:
 636                         goto illegal_evt;
 637                 }
 638                 break;
 639         case LINK_ESTABLISHING:
 640                 switch (evt) {
 641                 case LINK_ESTABLISH_EVT:
 642                         l->state = LINK_ESTABLISHED;
 643                         break;
 644                 case LINK_FAILOVER_BEGIN_EVT:
 645                         l->state = LINK_FAILINGOVER;
 646                         break;
 647                 case LINK_RESET_EVT:
 648                         l->state = LINK_RESET;
 649                         break;
 650                 case LINK_FAILURE_EVT:
 651                 case LINK_PEER_RESET_EVT:
 652                 case LINK_SYNCH_BEGIN_EVT:
 653                 case LINK_FAILOVER_END_EVT:
 654                         break;
 655                 case LINK_SYNCH_END_EVT:
 656                 default:
 657                         goto illegal_evt;
 658                 }
 659                 break;
 660         case LINK_ESTABLISHED:
 661                 switch (evt) {
 662                 case LINK_PEER_RESET_EVT:
 663                         l->state = LINK_PEER_RESET;
 664                         rc |= TIPC_LINK_DOWN_EVT;
 665                         break;
 666                 case LINK_FAILURE_EVT:
 667                         l->state = LINK_RESETTING;
 668                         rc |= TIPC_LINK_DOWN_EVT;
 669                         break;
 670                 case LINK_RESET_EVT:
 671                         l->state = LINK_RESET;
 672                         break;
 673                 case LINK_ESTABLISH_EVT:
 674                 case LINK_SYNCH_END_EVT:
 675                         break;
 676                 case LINK_SYNCH_BEGIN_EVT:
 677                         l->state = LINK_SYNCHING;
 678                         break;
 679                 case LINK_FAILOVER_BEGIN_EVT:
 680                 case LINK_FAILOVER_END_EVT:
 681                 default:
 682                         goto illegal_evt;
 683                 }
 684                 break;
 685         case LINK_SYNCHING:
 686                 switch (evt) {
 687                 case LINK_PEER_RESET_EVT:
 688                         l->state = LINK_PEER_RESET;
 689                         rc |= TIPC_LINK_DOWN_EVT;
 690                         break;
 691                 case LINK_FAILURE_EVT:
 692                         l->state = LINK_RESETTING;
 693                         rc |= TIPC_LINK_DOWN_EVT;
 694                         break;
 695                 case LINK_RESET_EVT:
 696                         l->state = LINK_RESET;
 697                         break;
 698                 case LINK_ESTABLISH_EVT:
 699                 case LINK_SYNCH_BEGIN_EVT:
 700                         break;
 701                 case LINK_SYNCH_END_EVT:
 702                         l->state = LINK_ESTABLISHED;
 703                         break;
 704                 case LINK_FAILOVER_BEGIN_EVT:
 705                 case LINK_FAILOVER_END_EVT:
 706                 default:
 707                         goto illegal_evt;
 708                 }
 709                 break;
 710         default:
 711                 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
 712         }
 713         trace_tipc_link_fsm(l->name, old_state, l->state, evt);
 714         return rc;
 715 illegal_evt:
 716         pr_err("Illegal FSM event %x in state %x on link %s\n",
 717                evt, l->state, l->name);
 718         trace_tipc_link_fsm(l->name, old_state, l->state, evt);
 719         return rc;
 720 }
 721
 722 /* link_profile_stats - update statistical profiling of traffic
 723  */
 724 static void link_profile_stats(struct tipc_link *l)
 725 {
 726         struct sk_buff *skb;
 727         struct tipc_msg *msg;
 728         int length;
 729
 730         /* Update counters used in statistical profiling of send traffic */
 731         l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
 732         l->stats.queue_sz_counts++;
 733
 734         skb = skb_peek(&l->transmq);
 735         if (!skb)
 736                 return;
 737         msg = buf_msg(skb);
 738         length = msg_size(msg);
 739
 740         if (msg_user(msg) == MSG_FRAGMENTER) {
 741                 if (msg_type(msg) != FIRST_FRAGMENT)
 742                         return;
 743                 length = msg_size(msg_inner_hdr(msg));
 744         }
 745         l->stats.msg_lengths_total += length;
 746         l->stats.msg_length_counts++;
 747         if (length <= 64)
 748                 l->stats.msg_length_profile[0]++;
 749         else if (length <= 256)
 750                 l->stats.msg_length_profile[1]++;
 751         else if (length <= 1024)
 752                 l->stats.msg_length_profile[2]++;
 753         else if (length <= 4096)
 754                 l->stats.msg_length_profile[3]++;
 755         else if (length <= 16384)
 756                 l->stats.msg_length_profile[4]++;
 757         else if (length <= 32768)
 758                 l->stats.msg_length_profile[5]++;
 759         else
 760                 l->stats.msg_length_profile[6]++;
 761 }
 762
 763 /**
 764  * tipc_link_too_silent - check if link is "too silent"
 765  * @l: tipc link to be checked
 766  *
 767  * Returns true if the link 'silent_intv_cnt' is about to reach the
 768  * 'abort_limit' value, otherwise false
 769  */
 770 bool tipc_link_too_silent(struct tipc_link *l)
 771 {
 772         return (l->silent_intv_cnt + 2 > l->abort_limit);
 773 }
 774
 775 /* tipc_link_timeout - perform periodic task as instructed from node timeout
 776  */
 777 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
 778 {
 779         int mtyp = 0;
 780         int rc = 0;
 781         bool state = false;
 782         bool probe = false;
 783         bool setup = false;
 784         u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
 785         u16 bc_acked = l->bc_rcvlink->acked;
 786         struct tipc_mon_state *mstate = &l->mon_state;
 787
 788         trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
 789         trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
 790         switch (l->state) {
 791         case LINK_ESTABLISHED:
 792         case LINK_SYNCHING:
 793                 mtyp = STATE_MSG;
 794                 link_profile_stats(l);
 795                 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
 796                 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
 797                         return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
 798                 state = bc_acked != bc_snt;
 799                 state |= l->bc_rcvlink->rcv_unacked;
 800                 state |= l->rcv_unacked;
 801                 state |= !skb_queue_empty(&l->transmq);
 802                 state |= !skb_queue_empty(&l->deferdq);
 803                 probe = mstate->probing;
 804                 probe |= l->silent_intv_cnt;
 805                 if (probe || mstate->monitoring)
 806                         l->silent_intv_cnt++;
 807                 break;
 808         case LINK_RESET:
 809                 setup = l->rst_cnt++ <= 4;
 810                 setup |= !(l->rst_cnt % 16);
 811                 mtyp = RESET_MSG;
 812                 break;
 813         case LINK_ESTABLISHING:
 814                 setup = true;
 815                 mtyp = ACTIVATE_MSG;
 816                 break;
 817         case LINK_PEER_RESET:
 818         case LINK_RESETTING:
 819         case LINK_FAILINGOVER:
 820                 break;
 821         default:
 822                 break;
 823         }
 824
 825         if (state || probe || setup)
 826                 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
 827
 828         return rc;
 829 }
 830
 831 /**
 832  * link_schedule_user - schedule a message sender for wakeup after congestion
 833  * @l: congested link
 834  * @hdr: header of message that is being sent
 835  * Create pseudo msg to send back to user when congestion abates
 836  */
 837 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
 838 {
 839         u32 dnode = tipc_own_addr(l->net);
 840         u32 dport = msg_origport(hdr);
 841         struct sk_buff *skb;
 842
 843         /* Create and schedule wakeup pseudo message */
 844         skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
 845                               dnode, l->addr, dport, 0, 0);
 846         if (!skb)
 847                 return -ENOBUFS;
 848         msg_set_dest_droppable(buf_msg(skb), true);
 849         TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
 850         skb_queue_tail(&l->wakeupq, skb);
 851         l->stats.link_congs++;
 852         trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
 853         return -ELINKCONG;
 854 }
 855
 856 /**
 857  * link_prepare_wakeup - prepare users for wakeup after congestion
 858  * @l: congested link
 859  * Wake up a number of waiting users, as permitted by available space
 860  * in the send queue
 861  */
 862 static void link_prepare_wakeup(struct tipc_link *l)
 863 {
 864         struct sk_buff_head *wakeupq = &l->wakeupq;
 865         struct sk_buff_head *inputq = l->inputq;
 866         struct sk_buff *skb, *tmp;
 867         struct sk_buff_head tmpq;
 868         int avail[5] = {0,};
 869         int imp = 0;
 870
 871         __skb_queue_head_init(&tmpq);
 872
 873         for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
 874                 avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
 875
 876         skb_queue_walk_safe(wakeupq, skb, tmp) {
 877                 imp = TIPC_SKB_CB(skb)->chain_imp;
 878                 if (avail[imp] <= 0)
 879                         continue;
 880                 avail[imp]--;
 881                 __skb_unlink(skb, wakeupq);
 882                 __skb_queue_tail(&tmpq, skb);
 883         }
 884
 885         spin_lock_bh(&inputq->lock);
 886         skb_queue_splice_tail(&tmpq, inputq);
 887         spin_unlock_bh(&inputq->lock);
 888
 889 }
 890
 891 void tipc_link_reset(struct tipc_link *l)
 892 {
 893         struct sk_buff_head list;
 894         u32 imp;
 895
 896         __skb_queue_head_init(&list);
 897
 898         l->in_session = false;
 899         /* Force re-synch of peer session number before establishing */
 900         l->peer_session--;
 901         l->session++;
 902         l->mtu = l->advertised_mtu;
 903
 904         spin_lock_bh(&l->wakeupq.lock);
 905         skb_queue_splice_init(&l->wakeupq, &list);
 906         spin_unlock_bh(&l->wakeupq.lock);
 907
 908         spin_lock_bh(&l->inputq->lock);
 909         skb_queue_splice_init(&list, l->inputq);
 910         spin_unlock_bh(&l->inputq->lock);
 911
 912         __skb_queue_purge(&l->transmq);
 913         __skb_queue_purge(&l->deferdq);
 914         __skb_queue_purge(&l->backlogq);
 915         __skb_queue_purge(&l->failover_deferdq);
 916         for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
 917                 l->backlog[imp].len = 0;
 918                 l->backlog[imp].target_bskb = NULL;
 919         }
 920         kfree_skb(l->reasm_buf);
 921         kfree_skb(l->reasm_tnlmsg);
 922         kfree_skb(l->failover_reasm_skb);
 923         l->reasm_buf = NULL;
 924         l->reasm_tnlmsg = NULL;
 925         l->failover_reasm_skb = NULL;
 926         l->rcv_unacked = 0;
 927         l->snd_nxt = 1;
 928         l->rcv_nxt = 1;
 929         l->snd_nxt_state = 1;
 930         l->rcv_nxt_state = 1;
 931         l->acked = 0;
 932         l->silent_intv_cnt = 0;
 933         l->rst_cnt = 0;
 934         l->bc_peer_is_up = false;
 935         memset(&l->mon_state, 0, sizeof(l->mon_state));
 936         tipc_link_reset_stats(l);
 937 }
 938
 939 /**
 940  * tipc_link_xmit(): enqueue buffer list according to queue situation
 941  * @link: link to use
 942  * @list: chain of buffers containing message
 943  * @xmitq: returned list of packets to be sent by caller
 944  *
 945  * Consumes the buffer chain.
 946  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
 947  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
 948  */
 949 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
 950                    struct sk_buff_head *xmitq)
 951 {
 952         struct tipc_msg *hdr = buf_msg(skb_peek(list));
 953         struct sk_buff_head *backlogq = &l->backlogq;
 954         struct sk_buff_head *transmq = &l->transmq;
 955         struct sk_buff *skb, *_skb;
 956         u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
 957         u16 ack = l->rcv_nxt - 1;
 958         u16 seqno = l->snd_nxt;
 959         int pkt_cnt = skb_queue_len(list);
 960         int imp = msg_importance(hdr);
 961         unsigned int mss = tipc_link_mss(l);
 962         unsigned int maxwin = l->window;
 963         unsigned int mtu = l->mtu;
 964         bool new_bundle;
 965         int rc = 0;
 966
 967         if (unlikely(msg_size(hdr) > mtu)) {
 968                 pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
 969                         skb_queue_len(list), msg_user(hdr),
 970                         msg_type(hdr), msg_size(hdr), mtu);
 971                 __skb_queue_purge(list);
 972                 return -EMSGSIZE;
 973         }
 974
 975         /* Allow oversubscription of one data msg per source at congestion */
 976         if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
 977                 if (imp == TIPC_SYSTEM_IMPORTANCE) {
 978                         pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
 979                         return -ENOBUFS;
 980                 }
 981                 rc = link_schedule_user(l, hdr);
 982         }
 983
 984         if (pkt_cnt > 1) {
 985                 l->stats.sent_fragmented++;
 986                 l->stats.sent_fragments += pkt_cnt;
 987         }
 988
 989         /* Prepare each packet for sending, and add to relevant queue: */
 990         while ((skb = __skb_dequeue(list))) {
 991                 if (likely(skb_queue_len(transmq) < maxwin)) {
 992                         hdr = buf_msg(skb);
 993                         msg_set_seqno(hdr, seqno);
 994                         msg_set_ack(hdr, ack);
 995                         msg_set_bcast_ack(hdr, bc_ack);
 996                         _skb = skb_clone(skb, GFP_ATOMIC);
 997                         if (!_skb) {
 998                                 kfree_skb(skb);
 999                                 __skb_queue_purge(list);
1000                                 return -ENOBUFS;
1001                         }
1002                         __skb_queue_tail(transmq, skb);
1003                         /* next retransmit attempt */
1004                         if (link_is_bc_sndlink(l))
1005                                 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1006                         __skb_queue_tail(xmitq, _skb);
1007                         TIPC_SKB_CB(skb)->ackers = l->ackers;
1008                         l->rcv_unacked = 0;
1009                         l->stats.sent_pkts++;
1010                         seqno++;
1011                         continue;
1012                 }
1013                 if (tipc_msg_try_bundle(l->backlog[imp].target_bskb, &skb,
1014                                         mss, l->addr, &new_bundle)) {
1015                         if (skb) {
1016                                 /* Keep a ref. to the skb for next try */
1017                                 l->backlog[imp].target_bskb = skb;
1018                                 l->backlog[imp].len++;
1019                                 __skb_queue_tail(backlogq, skb);
1020                         } else {
1021                                 if (new_bundle) {
1022                                         l->stats.sent_bundles++;
1023                                         l->stats.sent_bundled++;
1024                                 }
1025                                 l->stats.sent_bundled++;
1026                         }
1027                         continue;
1028                 }
1029                 l->backlog[imp].target_bskb = NULL;
1030                 l->backlog[imp].len += (1 + skb_queue_len(list));
1031                 __skb_queue_tail(backlogq, skb);
1032                 skb_queue_splice_tail_init(list, backlogq);
1033         }
1034         l->snd_nxt = seqno;
1035         return rc;
1036 }
1037
1038 static void tipc_link_advance_backlog(struct tipc_link *l,
1039                                       struct sk_buff_head *xmitq)
1040 {
1041         struct sk_buff *skb, *_skb;
1042         struct tipc_msg *hdr;
1043         u16 seqno = l->snd_nxt;
1044         u16 ack = l->rcv_nxt - 1;
1045         u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1046         u32 imp;
1047
1048         while (skb_queue_len(&l->transmq) < l->window) {
1049                 skb = skb_peek(&l->backlogq);
1050                 if (!skb)
1051                         break;
1052                 _skb = skb_clone(skb, GFP_ATOMIC);
1053                 if (!_skb)
1054                         break;
1055                 __skb_dequeue(&l->backlogq);
1056                 hdr = buf_msg(skb);
1057                 imp = msg_importance(hdr);
1058                 l->backlog[imp].len--;
1059                 if (unlikely(skb == l->backlog[imp].target_bskb))
1060                         l->backlog[imp].target_bskb = NULL;
1061                 __skb_queue_tail(&l->transmq, skb);
1062                 /* next retransmit attempt */
1063                 if (link_is_bc_sndlink(l))
1064                         TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1065
1066                 __skb_queue_tail(xmitq, _skb);
1067                 TIPC_SKB_CB(skb)->ackers = l->ackers;
1068                 msg_set_seqno(hdr, seqno);
1069                 msg_set_ack(hdr, ack);
1070                 msg_set_bcast_ack(hdr, bc_ack);
1071                 l->rcv_unacked = 0;
1072                 l->stats.sent_pkts++;
1073                 seqno++;
1074         }
1075         l->snd_nxt = seqno;
1076 }
1077
1078 /**
1079  * link_retransmit_failure() - Detect repeated retransmit failures
1080  * @l: tipc link sender
1081  * @r: tipc link receiver (= l in case of unicast)
1082  * @rc: returned code
1083  *
1084  * Return: true if the repeated retransmit failures happens, otherwise
1085  * false
1086  */
1087 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1088                                     int *rc)
1089 {
1090         struct sk_buff *skb = skb_peek(&l->transmq);
1091         struct tipc_msg *hdr;
1092
1093         if (!skb)
1094                 return false;
1095
1096         if (!TIPC_SKB_CB(skb)->retr_cnt)
1097                 return false;
1098
1099         if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1100                         msecs_to_jiffies(r->tolerance * 10)))
1101                 return false;
1102
1103         hdr = buf_msg(skb);
1104         if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1105                 return false;
1106
1107         pr_warn("Retransmission failure on link <%s>\n", l->name);
1108         link_print(l, "State of link ");
1109         pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1110                 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1111         pr_info("sqno %u, prev: %x, dest: %x\n",
1112                 msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1113         pr_info("retr_stamp %d, retr_cnt %d\n",
1114                 jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1115                 TIPC_SKB_CB(skb)->retr_cnt);
1116
1117         trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1118         trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1119         trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1120
1121         if (link_is_bc_sndlink(l)) {
1122                 r->state = LINK_RESET;
1123                 *rc = TIPC_LINK_DOWN_EVT;
1124         } else {
1125                 *rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1126         }
1127
1128         return true;
1129 }
1130
1131 /* tipc_link_bc_retrans() - retransmit zero or more packets
1132  * @l: the link to transmit on
1133  * @r: the receiving link ordering the retransmit. Same as l if unicast
1134  * @from: retransmit from (inclusive) this sequence number
1135  * @to: retransmit to (inclusive) this sequence number
1136  * xmitq: queue for accumulating the retransmitted packets
1137  */
1138 static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r,
1139                                 u16 from, u16 to, struct sk_buff_head *xmitq)
1140 {
1141         struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1142         u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1143         u16 ack = l->rcv_nxt - 1;
1144         struct tipc_msg *hdr;
1145         int rc = 0;
1146
1147         if (!skb)
1148                 return 0;
1149         if (less(to, from))
1150                 return 0;
1151
1152         trace_tipc_link_retrans(r, from, to, &l->transmq);
1153
1154         if (link_retransmit_failure(l, r, &rc))
1155                 return rc;
1156
1157         skb_queue_walk(&l->transmq, skb) {
1158                 hdr = buf_msg(skb);
1159                 if (less(msg_seqno(hdr), from))
1160                         continue;
1161                 if (more(msg_seqno(hdr), to))
1162                         break;
1163
1164                 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1165                         continue;
1166                 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1167                 _skb = pskb_copy(skb, GFP_ATOMIC);
1168                 if (!_skb)
1169                         return 0;
1170                 hdr = buf_msg(_skb);
1171                 msg_set_ack(hdr, ack);
1172                 msg_set_bcast_ack(hdr, bc_ack);
1173                 _skb->priority = TC_PRIO_CONTROL;
1174                 __skb_queue_tail(xmitq, _skb);
1175                 l->stats.retransmitted++;
1176
1177                 /* Increase actual retrans counter & mark first time */
1178                 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1179                         TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1180         }
1181         return 0;
1182 }
1183
1184 /* tipc_data_input - deliver data and name distr msgs to upper layer
1185  *
1186  * Consumes buffer if message is of right type
1187  * Node lock must be held
1188  */
1189 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1190                             struct sk_buff_head *inputq)
1191 {
1192         struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1193         struct tipc_msg *hdr = buf_msg(skb);
1194
1195         switch (msg_user(hdr)) {
1196         case TIPC_LOW_IMPORTANCE:
1197         case TIPC_MEDIUM_IMPORTANCE:
1198         case TIPC_HIGH_IMPORTANCE:
1199         case TIPC_CRITICAL_IMPORTANCE:
1200                 if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1201                         skb_queue_tail(mc_inputq, skb);
1202                         return true;
1203                 }
1204                 /* fall through */
1205         case CONN_MANAGER:
1206                 skb_queue_tail(inputq, skb);
1207                 return true;
1208         case GROUP_PROTOCOL:
1209                 skb_queue_tail(mc_inputq, skb);
1210                 return true;
1211         case NAME_DISTRIBUTOR:
1212                 l->bc_rcvlink->state = LINK_ESTABLISHED;
1213                 skb_queue_tail(l->namedq, skb);
1214                 return true;
1215         case MSG_BUNDLER:
1216         case TUNNEL_PROTOCOL:
1217         case MSG_FRAGMENTER:
1218         case BCAST_PROTOCOL:
1219                 return false;
1220         default:
1221                 pr_warn("Dropping received illegal msg type\n");
1222                 kfree_skb(skb);
1223                 return true;
1224         };
1225 }
1226
1227 /* tipc_link_input - process packet that has passed link protocol check
1228  *
1229  * Consumes buffer
1230  */
1231 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1232                            struct sk_buff_head *inputq,
1233                            struct sk_buff **reasm_skb)
1234 {
1235         struct tipc_msg *hdr = buf_msg(skb);
1236         struct sk_buff *iskb;
1237         struct sk_buff_head tmpq;
1238         int usr = msg_user(hdr);
1239         int pos = 0;
1240
1241         if (usr == MSG_BUNDLER) {
1242                 skb_queue_head_init(&tmpq);
1243                 l->stats.recv_bundles++;
1244                 l->stats.recv_bundled += msg_msgcnt(hdr);
1245                 while (tipc_msg_extract(skb, &iskb, &pos))
1246                         tipc_data_input(l, iskb, &tmpq);
1247                 tipc_skb_queue_splice_tail(&tmpq, inputq);
1248                 return 0;
1249         } else if (usr == MSG_FRAGMENTER) {
1250                 l->stats.recv_fragments++;
1251                 if (tipc_buf_append(reasm_skb, &skb)) {
1252                         l->stats.recv_fragmented++;
1253                         tipc_data_input(l, skb, inputq);
1254                 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1255                         pr_warn_ratelimited("Unable to build fragment list\n");
1256                         return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1257                 }
1258                 return 0;
1259         } else if (usr == BCAST_PROTOCOL) {
1260                 tipc_bcast_lock(l->net);
1261                 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1262                 tipc_bcast_unlock(l->net);
1263         }
1264
1265         kfree_skb(skb);
1266         return 0;
1267 }
1268
1269 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1270  *                       inner message along with the ones in the old link's
1271  *                       deferdq
1272  * @l: tunnel link
1273  * @skb: TUNNEL_PROTOCOL message
1274  * @inputq: queue to put messages ready for delivery
1275  */
1276 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1277                              struct sk_buff_head *inputq)
1278 {
1279         struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1280         struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1281         struct sk_buff_head *fdefq = &l->failover_deferdq;
1282         struct tipc_msg *hdr = buf_msg(skb);
1283         struct sk_buff *iskb;
1284         int ipos = 0;
1285         int rc = 0;
1286         u16 seqno;
1287
1288         if (msg_type(hdr) == SYNCH_MSG) {
1289                 kfree_skb(skb);
1290                 return 0;
1291         }
1292
1293         /* Not a fragment? */
1294         if (likely(!msg_nof_fragms(hdr))) {
1295                 if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1296                         pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1297                                             skb_queue_len(fdefq));
1298                         return 0;
1299                 }
1300                 kfree_skb(skb);
1301         } else {
1302                 /* Set fragment type for buf_append */
1303                 if (msg_fragm_no(hdr) == 1)
1304                         msg_set_type(hdr, FIRST_FRAGMENT);
1305                 else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1306                         msg_set_type(hdr, FRAGMENT);
1307                 else
1308                         msg_set_type(hdr, LAST_FRAGMENT);
1309
1310                 if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1311                         /* Successful but non-complete reassembly? */
1312                         if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1313                                 return 0;
1314                         pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1315                         return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1316                 }
1317                 iskb = skb;
1318         }
1319
1320         do {
1321                 seqno = buf_seqno(iskb);
1322                 if (unlikely(less(seqno, l->drop_point))) {
1323                         kfree_skb(iskb);
1324                         continue;
1325                 }
1326                 if (unlikely(seqno != l->drop_point)) {
1327                         __tipc_skb_queue_sorted(fdefq, seqno, iskb);
1328                         continue;
1329                 }
1330
1331                 l->drop_point++;
1332                 if (!tipc_data_input(l, iskb, inputq))
1333                         rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1334                 if (unlikely(rc))
1335                         break;
1336         } while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1337
1338         return rc;
1339 }
1340
1341 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1342 {
1343         bool released = false;
1344         struct sk_buff *skb, *tmp;
1345
1346         skb_queue_walk_safe(&l->transmq, skb, tmp) {
1347                 if (more(buf_seqno(skb), acked))
1348                         break;
1349                 __skb_unlink(skb, &l->transmq);
1350                 kfree_skb(skb);
1351                 released = true;
1352         }
1353         return released;
1354 }
1355
1356 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1357  * @l: tipc link that data have come with gaps in sequence if any
1358  * @data: data buffer to store the Gap ACK blocks after built
1359  *
1360  * returns the actual allocated memory size
1361  */
1362 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data)
1363 {
1364         struct sk_buff *skb = skb_peek(&l->deferdq);
1365         struct tipc_gap_ack_blks *ga = data;
1366         u16 len, expect, seqno = 0;
1367         u8 n = 0;
1368
1369         if (!skb)
1370                 goto exit;
1371
1372         expect = buf_seqno(skb);
1373         skb_queue_walk(&l->deferdq, skb) {
1374                 seqno = buf_seqno(skb);
1375                 if (unlikely(more(seqno, expect))) {
1376                         ga->gacks[n].ack = htons(expect - 1);
1377                         ga->gacks[n].gap = htons(seqno - expect);
1378                         if (++n >= MAX_GAP_ACK_BLKS) {
1379                                 pr_info_ratelimited("Too few Gap ACK blocks!\n");
1380                                 goto exit;
1381                         }
1382                 } else if (unlikely(less(seqno, expect))) {
1383                         pr_warn("Unexpected skb in deferdq!\n");
1384                         continue;
1385                 }
1386                 expect = seqno + 1;
1387         }
1388
1389         /* last block */
1390         ga->gacks[n].ack = htons(seqno);
1391         ga->gacks[n].gap = 0;
1392         n++;
1393
1394 exit:
1395         len = tipc_gap_ack_blks_sz(n);
1396         ga->len = htons(len);
1397         ga->gack_cnt = n;
1398         return len;
1399 }
1400
1401 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1402  *                             acked packets, also doing retransmissions if
1403  *                             gaps found
1404  * @l: tipc link with transmq queue to be advanced
1405  * @acked: seqno of last packet acked by peer without any gaps before
1406  * @gap: # of gap packets
1407  * @ga: buffer pointer to Gap ACK blocks from peer
1408  * @xmitq: queue for accumulating the retransmitted packets if any
1409  *
1410  * In case of a repeated retransmit failures, the call will return shortly
1411  * with a returned code (e.g. TIPC_LINK_DOWN_EVT)
1412  */
1413 static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
1414                                      struct tipc_gap_ack_blks *ga,
1415                                      struct sk_buff_head *xmitq)
1416 {
1417         struct sk_buff *skb, *_skb, *tmp;
1418         struct tipc_msg *hdr;
1419         u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1420         u16 ack = l->rcv_nxt - 1;
1421         bool passed = false;
1422         u16 seqno, n = 0;
1423         int rc = 0;
1424
1425         skb_queue_walk_safe(&l->transmq, skb, tmp) {
1426                 seqno = buf_seqno(skb);
1427
1428 next_gap_ack:
1429                 if (less_eq(seqno, acked)) {
1430                         /* release skb */
1431                         __skb_unlink(skb, &l->transmq);
1432                         kfree_skb(skb);
1433                 } else if (less_eq(seqno, acked + gap)) {
1434                         /* First, check if repeated retrans failures occurs? */
1435                         if (!passed && link_retransmit_failure(l, l, &rc))
1436                                 return rc;
1437                         passed = true;
1438
1439                         /* retransmit skb if unrestricted*/
1440                         if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1441                                 continue;
1442                         TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
1443                         _skb = pskb_copy(skb, GFP_ATOMIC);
1444                         if (!_skb)
1445                                 continue;
1446                         hdr = buf_msg(_skb);
1447                         msg_set_ack(hdr, ack);
1448                         msg_set_bcast_ack(hdr, bc_ack);
1449                         _skb->priority = TC_PRIO_CONTROL;
1450                         __skb_queue_tail(xmitq, _skb);
1451                         l->stats.retransmitted++;
1452
1453                         /* Increase actual retrans counter & mark first time */
1454                         if (!TIPC_SKB_CB(skb)->retr_cnt++)
1455                                 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1456                 } else {
1457                         /* retry with Gap ACK blocks if any */
1458                         if (!ga || n >= ga->gack_cnt)
1459                                 break;
1460                         acked = ntohs(ga->gacks[n].ack);
1461                         gap = ntohs(ga->gacks[n].gap);
1462                         n++;
1463                         goto next_gap_ack;
1464                 }
1465         }
1466
1467         return 0;
1468 }
1469
1470 /* tipc_link_build_state_msg: prepare link state message for transmission
1471  *
1472  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1473  * risk of ack storms towards the sender
1474  */
1475 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1476 {
1477         if (!l)
1478                 return 0;
1479
1480         /* Broadcast ACK must be sent via a unicast link => defer to caller */
1481         if (link_is_bc_rcvlink(l)) {
1482                 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1483                         return 0;
1484                 l->rcv_unacked = 0;
1485
1486                 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1487                 l->snd_nxt = l->rcv_nxt;
1488                 return TIPC_LINK_SND_STATE;
1489         }
1490
1491         /* Unicast ACK */
1492         l->rcv_unacked = 0;
1493         l->stats.sent_acks++;
1494         tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1495         return 0;
1496 }
1497
1498 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1499  */
1500 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1501 {
1502         int mtyp = RESET_MSG;
1503         struct sk_buff *skb;
1504
1505         if (l->state == LINK_ESTABLISHING)
1506                 mtyp = ACTIVATE_MSG;
1507
1508         tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1509
1510         /* Inform peer that this endpoint is going down if applicable */
1511         skb = skb_peek_tail(xmitq);
1512         if (skb && (l->state == LINK_RESET))
1513                 msg_set_peer_stopping(buf_msg(skb), 1);
1514 }
1515
1516 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1517  * Note that sending of broadcast NACK is coordinated among nodes, to
1518  * reduce the risk of NACK storms towards the sender
1519  */
1520 static int tipc_link_build_nack_msg(struct tipc_link *l,
1521                                     struct sk_buff_head *xmitq)
1522 {
1523         u32 def_cnt = ++l->stats.deferred_recv;
1524         u32 defq_len = skb_queue_len(&l->deferdq);
1525         int match1, match2;
1526
1527         if (link_is_bc_rcvlink(l)) {
1528                 match1 = def_cnt & 0xf;
1529                 match2 = tipc_own_addr(l->net) & 0xf;
1530                 if (match1 == match2)
1531                         return TIPC_LINK_SND_STATE;
1532                 return 0;
1533         }
1534
1535         if (defq_len >= 3 && !((defq_len - 3) % 16))
1536                 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1537         return 0;
1538 }
1539
1540 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1541  * @l: the link that should handle the message
1542  * @skb: TIPC packet
1543  * @xmitq: queue to place packets to be sent after this call
1544  */
1545 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1546                   struct sk_buff_head *xmitq)
1547 {
1548         struct sk_buff_head *defq = &l->deferdq;
1549         struct tipc_msg *hdr = buf_msg(skb);
1550         u16 seqno, rcv_nxt, win_lim;
1551         int rc = 0;
1552
1553         /* Verify and update link state */
1554         if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1555                 return tipc_link_proto_rcv(l, skb, xmitq);
1556
1557         /* Don't send probe at next timeout expiration */
1558         l->silent_intv_cnt = 0;
1559
1560         do {
1561                 hdr = buf_msg(skb);
1562                 seqno = msg_seqno(hdr);
1563                 rcv_nxt = l->rcv_nxt;
1564                 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1565
1566                 if (unlikely(!link_is_up(l))) {
1567                         if (l->state == LINK_ESTABLISHING)
1568                                 rc = TIPC_LINK_UP_EVT;
1569                         goto drop;
1570                 }
1571
1572                 /* Drop if outside receive window */
1573                 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1574                         l->stats.duplicates++;
1575                         goto drop;
1576                 }
1577
1578                 /* Forward queues and wake up waiting users */
1579                 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1580                         tipc_link_advance_backlog(l, xmitq);
1581                         if (unlikely(!skb_queue_empty(&l->wakeupq)))
1582                                 link_prepare_wakeup(l);
1583                 }
1584
1585                 /* Defer delivery if sequence gap */
1586                 if (unlikely(seqno != rcv_nxt)) {
1587                         __tipc_skb_queue_sorted(defq, seqno, skb);
1588                         rc |= tipc_link_build_nack_msg(l, xmitq);
1589                         break;
1590                 }
1591
1592                 /* Deliver packet */
1593                 l->rcv_nxt++;
1594                 l->stats.recv_pkts++;
1595
1596                 if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1597                         rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1598                 else if (!tipc_data_input(l, skb, l->inputq))
1599                         rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1600                 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1601                         rc |= tipc_link_build_state_msg(l, xmitq);
1602                 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1603                         break;
1604         } while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1605
1606         return rc;
1607 drop:
1608         kfree_skb(skb);
1609         return rc;
1610 }
1611
1612 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1613                                       bool probe_reply, u16 rcvgap,
1614                                       int tolerance, int priority,
1615                                       struct sk_buff_head *xmitq)
1616 {
1617         struct tipc_link *bcl = l->bc_rcvlink;
1618         struct sk_buff *skb;
1619         struct tipc_msg *hdr;
1620         struct sk_buff_head *dfq = &l->deferdq;
1621         bool node_up = link_is_up(bcl);
1622         struct tipc_mon_state *mstate = &l->mon_state;
1623         int dlen = 0;
1624         void *data;
1625         u16 glen = 0;
1626
1627         /* Don't send protocol message during reset or link failover */
1628         if (tipc_link_is_blocked(l))
1629                 return;
1630
1631         if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1632                 return;
1633
1634         if (!skb_queue_empty(dfq))
1635                 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1636
1637         skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1638                               tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1639                               l->addr, tipc_own_addr(l->net), 0, 0, 0);
1640         if (!skb)
1641                 return;
1642
1643         hdr = buf_msg(skb);
1644         data = msg_data(hdr);
1645         msg_set_session(hdr, l->session);
1646         msg_set_bearer_id(hdr, l->bearer_id);
1647         msg_set_net_plane(hdr, l->net_plane);
1648         msg_set_next_sent(hdr, l->snd_nxt);
1649         msg_set_ack(hdr, l->rcv_nxt - 1);
1650         msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1651         msg_set_bc_ack_invalid(hdr, !node_up);
1652         msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1653         msg_set_link_tolerance(hdr, tolerance);
1654         msg_set_linkprio(hdr, priority);
1655         msg_set_redundant_link(hdr, node_up);
1656         msg_set_seq_gap(hdr, 0);
1657         msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1658
1659         if (mtyp == STATE_MSG) {
1660                 if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1661                         msg_set_seqno(hdr, l->snd_nxt_state++);
1662                 msg_set_seq_gap(hdr, rcvgap);
1663                 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1664                 msg_set_probe(hdr, probe);
1665                 msg_set_is_keepalive(hdr, probe || probe_reply);
1666                 if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1667                         glen = tipc_build_gap_ack_blks(l, data);
1668                 tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1669                 msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1670                 skb_trim(skb, INT_H_SIZE + glen + dlen);
1671                 l->stats.sent_states++;
1672                 l->rcv_unacked = 0;
1673         } else {
1674                 /* RESET_MSG or ACTIVATE_MSG */
1675                 if (mtyp == ACTIVATE_MSG) {
1676                         msg_set_dest_session_valid(hdr, 1);
1677                         msg_set_dest_session(hdr, l->peer_session);
1678                 }
1679                 msg_set_max_pkt(hdr, l->advertised_mtu);
1680                 strcpy(data, l->if_name);
1681                 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1682                 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1683         }
1684         if (probe)
1685                 l->stats.sent_probes++;
1686         if (rcvgap)
1687                 l->stats.sent_nacks++;
1688         skb->priority = TC_PRIO_CONTROL;
1689         __skb_queue_tail(xmitq, skb);
1690         trace_tipc_proto_build(skb, false, l->name);
1691 }
1692
1693 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1694                                     struct sk_buff_head *xmitq)
1695 {
1696         u32 onode = tipc_own_addr(l->net);
1697         struct tipc_msg *hdr, *ihdr;
1698         struct sk_buff_head tnlq;
1699         struct sk_buff *skb;
1700         u32 dnode = l->addr;
1701
1702         __skb_queue_head_init(&tnlq);
1703         skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1704                               INT_H_SIZE, BASIC_H_SIZE,
1705                               dnode, onode, 0, 0, 0);
1706         if (!skb) {
1707                 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1708                 return;
1709         }
1710
1711         hdr = buf_msg(skb);
1712         msg_set_msgcnt(hdr, 1);
1713         msg_set_bearer_id(hdr, l->peer_bearer_id);
1714
1715         ihdr = (struct tipc_msg *)msg_data(hdr);
1716         tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1717                       BASIC_H_SIZE, dnode);
1718         msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1719         __skb_queue_tail(&tnlq, skb);
1720         tipc_link_xmit(l, &tnlq, xmitq);
1721 }
1722
1723 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1724  * with contents of the link's transmit and backlog queues.
1725  */
1726 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1727                            int mtyp, struct sk_buff_head *xmitq)
1728 {
1729         struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1730         struct sk_buff *skb, *tnlskb;
1731         struct tipc_msg *hdr, tnlhdr;
1732         struct sk_buff_head *queue = &l->transmq;
1733         struct sk_buff_head tmpxq, tnlq, frags;
1734         u16 pktlen, pktcnt, seqno = l->snd_nxt;
1735         bool pktcnt_need_update = false;
1736         u16 syncpt;
1737         int rc;
1738
1739         if (!tnl)
1740                 return;
1741
1742         __skb_queue_head_init(&tnlq);
1743         /* Link Synching:
1744          * From now on, send only one single ("dummy") SYNCH message
1745          * to peer. The SYNCH message does not contain any data, just
1746          * a header conveying the synch point to the peer.
1747          */
1748         if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1749                 tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1750                                          INT_H_SIZE, 0, l->addr,
1751                                          tipc_own_addr(l->net),
1752                                          0, 0, 0);
1753                 if (!tnlskb) {
1754                         pr_warn("%sunable to create dummy SYNCH_MSG\n",
1755                                 link_co_err);
1756                         return;
1757                 }
1758
1759                 hdr = buf_msg(tnlskb);
1760                 syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
1761                 msg_set_syncpt(hdr, syncpt);
1762                 msg_set_bearer_id(hdr, l->peer_bearer_id);
1763                 __skb_queue_tail(&tnlq, tnlskb);
1764                 tipc_link_xmit(tnl, &tnlq, xmitq);
1765                 return;
1766         }
1767
1768         __skb_queue_head_init(&tmpxq);
1769         __skb_queue_head_init(&frags);
1770         /* At least one packet required for safe algorithm => add dummy */
1771         skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1772                               BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1773                               0, 0, TIPC_ERR_NO_PORT);
1774         if (!skb) {
1775                 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1776                 return;
1777         }
1778         __skb_queue_tail(&tnlq, skb);
1779         tipc_link_xmit(l, &tnlq, &tmpxq);
1780         __skb_queue_purge(&tmpxq);
1781
1782         /* Initialize reusable tunnel packet header */
1783         tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1784                       mtyp, INT_H_SIZE, l->addr);
1785         if (mtyp == SYNCH_MSG)
1786                 pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
1787         else
1788                 pktcnt = skb_queue_len(&l->transmq);
1789         pktcnt += skb_queue_len(&l->backlogq);
1790         msg_set_msgcnt(&tnlhdr, pktcnt);
1791         msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1792 tnl:
1793         /* Wrap each packet into a tunnel packet */
1794         skb_queue_walk(queue, skb) {
1795                 hdr = buf_msg(skb);
1796                 if (queue == &l->backlogq)
1797                         msg_set_seqno(hdr, seqno++);
1798                 pktlen = msg_size(hdr);
1799
1800                 /* Tunnel link MTU is not large enough? This could be
1801                  * due to:
1802                  * 1) Link MTU has just changed or set differently;
1803                  * 2) Or FAILOVER on the top of a SYNCH message
1804                  *
1805                  * The 2nd case should not happen if peer supports
1806                  * TIPC_TUNNEL_ENHANCED
1807                  */
1808                 if (pktlen > tnl->mtu - INT_H_SIZE) {
1809                         if (mtyp == FAILOVER_MSG &&
1810                             (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1811                                 rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
1812                                                        &frags);
1813                                 if (rc) {
1814                                         pr_warn("%sunable to frag msg: rc %d\n",
1815                                                 link_co_err, rc);
1816                                         return;
1817                                 }
1818                                 pktcnt += skb_queue_len(&frags) - 1;
1819                                 pktcnt_need_update = true;
1820                                 skb_queue_splice_tail_init(&frags, &tnlq);
1821                                 continue;
1822                         }
1823                         /* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
1824                          * => Just warn it and return!
1825                          */
1826                         pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
1827                                             link_co_err, msg_user(hdr),
1828                                             msg_type(hdr), msg_size(hdr));
1829                         return;
1830                 }
1831
1832                 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1833                 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1834                 if (!tnlskb) {
1835                         pr_warn("%sunable to send packet\n", link_co_err);
1836                         return;
1837                 }
1838                 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1839                 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1840                 __skb_queue_tail(&tnlq, tnlskb);
1841         }
1842         if (queue != &l->backlogq) {
1843                 queue = &l->backlogq;
1844                 goto tnl;
1845         }
1846
1847         if (pktcnt_need_update)
1848                 skb_queue_walk(&tnlq, skb) {
1849                         hdr = buf_msg(skb);
1850                         msg_set_msgcnt(hdr, pktcnt);
1851                 }
1852
1853         tipc_link_xmit(tnl, &tnlq, xmitq);
1854
1855         if (mtyp == FAILOVER_MSG) {
1856                 tnl->drop_point = l->rcv_nxt;
1857                 tnl->failover_reasm_skb = l->reasm_buf;
1858                 l->reasm_buf = NULL;
1859
1860                 /* Failover the link's deferdq */
1861                 if (unlikely(!skb_queue_empty(fdefq))) {
1862                         pr_warn("Link failover deferdq not empty: %d!\n",
1863                                 skb_queue_len(fdefq));
1864                         __skb_queue_purge(fdefq);
1865                 }
1866                 skb_queue_splice_init(&l->deferdq, fdefq);
1867         }
1868 }
1869
1870 /**
1871  * tipc_link_failover_prepare() - prepare tnl for link failover
1872  *
1873  * This is a special version of the precursor - tipc_link_tnl_prepare(),
1874  * see the tipc_node_link_failover() for details
1875  *
1876  * @l: failover link
1877  * @tnl: tunnel link
1878  * @xmitq: queue for messages to be xmited
1879  */
1880 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
1881                                 struct sk_buff_head *xmitq)
1882 {
1883         struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1884
1885         tipc_link_create_dummy_tnl_msg(tnl, xmitq);
1886
1887         /* This failover link endpoint was never established before,
1888          * so it has not received anything from peer.
1889          * Otherwise, it must be a normal failover situation or the
1890          * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
1891          * would have to start over from scratch instead.
1892          */
1893         tnl->drop_point = 1;
1894         tnl->failover_reasm_skb = NULL;
1895
1896         /* Initiate the link's failover deferdq */
1897         if (unlikely(!skb_queue_empty(fdefq))) {
1898                 pr_warn("Link failover deferdq not empty: %d!\n",
1899                         skb_queue_len(fdefq));
1900                 __skb_queue_purge(fdefq);
1901         }
1902 }
1903
1904 /* tipc_link_validate_msg(): validate message against current link state
1905  * Returns true if message should be accepted, otherwise false
1906  */
1907 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
1908 {
1909         u16 curr_session = l->peer_session;
1910         u16 session = msg_session(hdr);
1911         int mtyp = msg_type(hdr);
1912
1913         if (msg_user(hdr) != LINK_PROTOCOL)
1914                 return true;
1915
1916         switch (mtyp) {
1917         case RESET_MSG:
1918                 if (!l->in_session)
1919                         return true;
1920                 /* Accept only RESET with new session number */
1921                 return more(session, curr_session);
1922         case ACTIVATE_MSG:
1923                 if (!l->in_session)
1924                         return true;
1925                 /* Accept only ACTIVATE with new or current session number */
1926                 return !less(session, curr_session);
1927         case STATE_MSG:
1928                 /* Accept only STATE with current session number */
1929                 if (!l->in_session)
1930                         return false;
1931                 if (session != curr_session)
1932                         return false;
1933                 /* Extra sanity check */
1934                 if (!link_is_up(l) && msg_ack(hdr))
1935                         return false;
1936                 if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
1937                         return true;
1938                 /* Accept only STATE with new sequence number */
1939                 return !less(msg_seqno(hdr), l->rcv_nxt_state);
1940         default:
1941                 return false;
1942         }
1943 }
1944
1945 /* tipc_link_proto_rcv(): receive link level protocol message :
1946  * Note that network plane id propagates through the network, and may
1947  * change at any time. The node with lowest numerical id determines
1948  * network plane
1949  */
1950 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1951                                struct sk_buff_head *xmitq)
1952 {
1953         struct tipc_msg *hdr = buf_msg(skb);
1954         struct tipc_gap_ack_blks *ga = NULL;
1955         u16 rcvgap = 0;
1956         u16 ack = msg_ack(hdr);
1957         u16 gap = msg_seq_gap(hdr);
1958         u16 peers_snd_nxt =  msg_next_sent(hdr);
1959         u16 peers_tol = msg_link_tolerance(hdr);
1960         u16 peers_prio = msg_linkprio(hdr);
1961         u16 rcv_nxt = l->rcv_nxt;
1962         u16 dlen = msg_data_sz(hdr);
1963         int mtyp = msg_type(hdr);
1964         bool reply = msg_probe(hdr);
1965         u16 glen = 0;
1966         void *data;
1967         char *if_name;
1968         int rc = 0;
1969
1970         trace_tipc_proto_rcv(skb, false, l->name);
1971         if (tipc_link_is_blocked(l) || !xmitq)
1972                 goto exit;
1973
1974         if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1975                 l->net_plane = msg_net_plane(hdr);
1976
1977         skb_linearize(skb);
1978         hdr = buf_msg(skb);
1979         data = msg_data(hdr);
1980
1981         if (!tipc_link_validate_msg(l, hdr)) {
1982                 trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
1983                 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
1984                 goto exit;
1985         }
1986
1987         switch (mtyp) {
1988         case RESET_MSG:
1989         case ACTIVATE_MSG:
1990                 /* Complete own link name with peer's interface name */
1991                 if_name =  strrchr(l->name, ':') + 1;
1992                 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1993                         break;
1994                 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1995                         break;
1996                 strncpy(if_name, data, TIPC_MAX_IF_NAME);
1997
1998                 /* Update own tolerance if peer indicates a non-zero value */
1999                 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2000                         l->tolerance = peers_tol;
2001                         l->bc_rcvlink->tolerance = peers_tol;
2002                 }
2003                 /* Update own priority if peer's priority is higher */
2004                 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
2005                         l->priority = peers_prio;
2006
2007                 /* If peer is going down we want full re-establish cycle */
2008                 if (msg_peer_stopping(hdr)) {
2009                         rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2010                         break;
2011                 }
2012
2013                 /* If this endpoint was re-created while peer was ESTABLISHING
2014                  * it doesn't know current session number. Force re-synch.
2015                  */
2016                 if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2017                     l->session != msg_dest_session(hdr)) {
2018                         if (less(l->session, msg_dest_session(hdr)))
2019                                 l->session = msg_dest_session(hdr) + 1;
2020                         break;
2021                 }
2022
2023                 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2024                 if (mtyp == RESET_MSG || !link_is_up(l))
2025                         rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2026
2027                 /* ACTIVATE_MSG takes up link if it was already locally reset */
2028                 if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2029                         rc = TIPC_LINK_UP_EVT;
2030
2031                 l->peer_session = msg_session(hdr);
2032                 l->in_session = true;
2033                 l->peer_bearer_id = msg_bearer_id(hdr);
2034                 if (l->mtu > msg_max_pkt(hdr))
2035                         l->mtu = msg_max_pkt(hdr);
2036                 break;
2037
2038         case STATE_MSG:
2039                 l->rcv_nxt_state = msg_seqno(hdr) + 1;
2040
2041                 /* Update own tolerance if peer indicates a non-zero value */
2042                 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2043                         l->tolerance = peers_tol;
2044                         l->bc_rcvlink->tolerance = peers_tol;
2045                 }
2046                 /* Update own prio if peer indicates a different value */
2047                 if ((peers_prio != l->priority) &&
2048                     in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2049                         l->priority = peers_prio;
2050                         rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2051                 }
2052
2053                 l->silent_intv_cnt = 0;
2054                 l->stats.recv_states++;
2055                 if (msg_probe(hdr))
2056                         l->stats.recv_probes++;
2057
2058                 if (!link_is_up(l)) {
2059                         if (l->state == LINK_ESTABLISHING)
2060                                 rc = TIPC_LINK_UP_EVT;
2061                         break;
2062                 }
2063
2064                 /* Receive Gap ACK blocks from peer if any */
2065                 if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
2066                         ga = (struct tipc_gap_ack_blks *)data;
2067                         glen = ntohs(ga->len);
2068                         /* sanity check: if failed, ignore Gap ACK blocks */
2069                         if (glen != tipc_gap_ack_blks_sz(ga->gack_cnt))
2070                                 ga = NULL;
2071                 }
2072
2073                 tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2074                              &l->mon_state, l->bearer_id);
2075
2076                 /* Send NACK if peer has sent pkts we haven't received yet */
2077                 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
2078                         rcvgap = peers_snd_nxt - l->rcv_nxt;
2079                 if (rcvgap || reply)
2080                         tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2081                                                   rcvgap, 0, 0, xmitq);
2082
2083                 rc |= tipc_link_advance_transmq(l, ack, gap, ga, xmitq);
2084
2085                 /* If NACK, retransmit will now start at right position */
2086                 if (gap)
2087                         l->stats.recv_nacks++;
2088
2089                 tipc_link_advance_backlog(l, xmitq);
2090                 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2091                         link_prepare_wakeup(l);
2092         }
2093 exit:
2094         kfree_skb(skb);
2095         return rc;
2096 }
2097
2098 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2099  */
2100 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2101                                          u16 peers_snd_nxt,
2102                                          struct sk_buff_head *xmitq)
2103 {
2104         struct sk_buff *skb;
2105         struct tipc_msg *hdr;
2106         struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2107         u16 ack = l->rcv_nxt - 1;
2108         u16 gap_to = peers_snd_nxt - 1;
2109
2110         skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2111                               0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2112         if (!skb)
2113                 return false;
2114         hdr = buf_msg(skb);
2115         msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2116         msg_set_bcast_ack(hdr, ack);
2117         msg_set_bcgap_after(hdr, ack);
2118         if (dfrd_skb)
2119                 gap_to = buf_seqno(dfrd_skb) - 1;
2120         msg_set_bcgap_to(hdr, gap_to);
2121         msg_set_non_seq(hdr, bcast);
2122         __skb_queue_tail(xmitq, skb);
2123         return true;
2124 }
2125
2126 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2127  *
2128  * Give a newly added peer node the sequence number where it should
2129  * start receiving and acking broadcast packets.
2130  */
2131 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2132                                         struct sk_buff_head *xmitq)
2133 {
2134         struct sk_buff_head list;
2135
2136         __skb_queue_head_init(&list);
2137         if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2138                 return;
2139         msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2140         tipc_link_xmit(l, &list, xmitq);
2141 }
2142
2143 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2144  */
2145 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2146 {
2147         int mtyp = msg_type(hdr);
2148         u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2149
2150         if (link_is_up(l))
2151                 return;
2152
2153         if (msg_user(hdr) == BCAST_PROTOCOL) {
2154                 l->rcv_nxt = peers_snd_nxt;
2155                 l->state = LINK_ESTABLISHED;
2156                 return;
2157         }
2158
2159         if (l->peer_caps & TIPC_BCAST_SYNCH)
2160                 return;
2161
2162         if (msg_peer_node_is_up(hdr))
2163                 return;
2164
2165         /* Compatibility: accept older, less safe initial synch data */
2166         if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2167                 l->rcv_nxt = peers_snd_nxt;
2168 }
2169
2170 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2171  */
2172 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2173                           struct sk_buff_head *xmitq)
2174 {
2175         struct tipc_link *snd_l = l->bc_sndlink;
2176         u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2177         u16 from = msg_bcast_ack(hdr) + 1;
2178         u16 to = from + msg_bc_gap(hdr) - 1;
2179         int rc = 0;
2180
2181         if (!link_is_up(l))
2182                 return rc;
2183
2184         if (!msg_peer_node_is_up(hdr))
2185                 return rc;
2186
2187         /* Open when peer ackowledges our bcast init msg (pkt #1) */
2188         if (msg_ack(hdr))
2189                 l->bc_peer_is_up = true;
2190
2191         if (!l->bc_peer_is_up)
2192                 return rc;
2193
2194         l->stats.recv_nacks++;
2195
2196         /* Ignore if peers_snd_nxt goes beyond receive window */
2197         if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2198                 return rc;
2199
2200         rc = tipc_link_bc_retrans(snd_l, l, from, to, xmitq);
2201
2202         l->snd_nxt = peers_snd_nxt;
2203         if (link_bc_rcv_gap(l))
2204                 rc |= TIPC_LINK_SND_STATE;
2205
2206         /* Return now if sender supports nack via STATE messages */
2207         if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2208                 return rc;
2209
2210         /* Otherwise, be backwards compatible */
2211
2212         if (!more(peers_snd_nxt, l->rcv_nxt)) {
2213                 l->nack_state = BC_NACK_SND_CONDITIONAL;
2214                 return 0;
2215         }
2216
2217         /* Don't NACK if one was recently sent or peeked */
2218         if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2219                 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2220                 return 0;
2221         }
2222
2223         /* Conditionally delay NACK sending until next synch rcv */
2224         if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2225                 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2226                 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2227                         return 0;
2228         }
2229
2230         /* Send NACK now but suppress next one */
2231         tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2232         l->nack_state = BC_NACK_SND_SUPPRESS;
2233         return 0;
2234 }
2235
2236 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
2237                           struct sk_buff_head *xmitq)
2238 {
2239         struct sk_buff *skb, *tmp;
2240         struct tipc_link *snd_l = l->bc_sndlink;
2241
2242         if (!link_is_up(l) || !l->bc_peer_is_up)
2243                 return;
2244
2245         if (!more(acked, l->acked))
2246                 return;
2247
2248         trace_tipc_link_bc_ack(l, l->acked, acked, &snd_l->transmq);
2249         /* Skip over packets peer has already acked */
2250         skb_queue_walk(&snd_l->transmq, skb) {
2251                 if (more(buf_seqno(skb), l->acked))
2252                         break;
2253         }
2254
2255         /* Update/release the packets peer is acking now */
2256         skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
2257                 if (more(buf_seqno(skb), acked))
2258                         break;
2259                 if (!--TIPC_SKB_CB(skb)->ackers) {
2260                         __skb_unlink(skb, &snd_l->transmq);
2261                         kfree_skb(skb);
2262                 }
2263         }
2264         l->acked = acked;
2265         tipc_link_advance_backlog(snd_l, xmitq);
2266         if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
2267                 link_prepare_wakeup(snd_l);
2268 }
2269
2270 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2271  * This function is here for backwards compatibility, since
2272  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2273  */
2274 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2275                           struct sk_buff_head *xmitq)
2276 {
2277         struct tipc_msg *hdr = buf_msg(skb);
2278         u32 dnode = msg_destnode(hdr);
2279         int mtyp = msg_type(hdr);
2280         u16 acked = msg_bcast_ack(hdr);
2281         u16 from = acked + 1;
2282         u16 to = msg_bcgap_to(hdr);
2283         u16 peers_snd_nxt = to + 1;
2284         int rc = 0;
2285
2286         kfree_skb(skb);
2287
2288         if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2289                 return 0;
2290
2291         if (mtyp != STATE_MSG)
2292                 return 0;
2293
2294         if (dnode == tipc_own_addr(l->net)) {
2295                 tipc_link_bc_ack_rcv(l, acked, xmitq);
2296                 rc = tipc_link_bc_retrans(l->bc_sndlink, l, from, to, xmitq);
2297                 l->stats.recv_nacks++;
2298                 return rc;
2299         }
2300
2301         /* Msg for other node => suppress own NACK at next sync if applicable */
2302         if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2303                 l->nack_state = BC_NACK_SND_SUPPRESS;
2304
2305         return 0;
2306 }
2307
2308 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
2309 {
2310         int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2311
2312         l->window = win;
2313         l->backlog[TIPC_LOW_IMPORTANCE].limit      = max_t(u16, 50, win);
2314         l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = max_t(u16, 100, win * 2);
2315         l->backlog[TIPC_HIGH_IMPORTANCE].limit     = max_t(u16, 150, win * 3);
2316         l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
2317         l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
2318 }
2319
2320 /**
2321  * link_reset_stats - reset link statistics
2322  * @l: pointer to link
2323  */
2324 void tipc_link_reset_stats(struct tipc_link *l)
2325 {
2326         memset(&l->stats, 0, sizeof(l->stats));
2327 }
2328
2329 static void link_print(struct tipc_link *l, const char *str)
2330 {
2331         struct sk_buff *hskb = skb_peek(&l->transmq);
2332         u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2333         u16 tail = l->snd_nxt - 1;
2334
2335         pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2336         pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2337                 skb_queue_len(&l->transmq), head, tail,
2338                 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2339 }
2340
2341 /* Parse and validate nested (link) properties valid for media, bearer and link
2342  */
2343 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2344 {
2345         int err;
2346
2347         err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2348                                           tipc_nl_prop_policy, NULL);
2349         if (err)
2350                 return err;
2351
2352         if (props[TIPC_NLA_PROP_PRIO]) {
2353                 u32 prio;
2354
2355                 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2356                 if (prio > TIPC_MAX_LINK_PRI)
2357                         return -EINVAL;
2358         }
2359
2360         if (props[TIPC_NLA_PROP_TOL]) {
2361                 u32 tol;
2362
2363                 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2364                 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2365                         return -EINVAL;
2366         }
2367
2368         if (props[TIPC_NLA_PROP_WIN]) {
2369                 u32 win;
2370
2371                 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2372                 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
2373                         return -EINVAL;
2374         }
2375
2376         return 0;
2377 }
2378
2379 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2380 {
2381         int i;
2382         struct nlattr *stats;
2383
2384         struct nla_map {
2385                 u32 key;
2386                 u32 val;
2387         };
2388
2389         struct nla_map map[] = {
2390                 {TIPC_NLA_STATS_RX_INFO, 0},
2391                 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2392                 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2393                 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2394                 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2395                 {TIPC_NLA_STATS_TX_INFO, 0},
2396                 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2397                 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2398                 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2399                 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2400                 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2401                         s->msg_length_counts : 1},
2402                 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2403                 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2404                 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2405                 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2406                 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2407                 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2408                 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2409                 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2410                 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2411                 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
2412                 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2413                 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2414                 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2415                 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
2416                 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2417                 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2418                 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2419                 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2420                 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2421                 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2422                 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2423                 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2424                         (s->accu_queue_sz / s->queue_sz_counts) : 0}
2425         };
2426
2427         stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2428         if (!stats)
2429                 return -EMSGSIZE;
2430
2431         for (i = 0; i <  ARRAY_SIZE(map); i++)
2432                 if (nla_put_u32(skb, map[i].key, map[i].val))
2433                         goto msg_full;
2434
2435         nla_nest_end(skb, stats);
2436
2437         return 0;
2438 msg_full:
2439         nla_nest_cancel(skb, stats);
2440
2441         return -EMSGSIZE;
2442 }
2443
2444 /* Caller should hold appropriate locks to protect the link */
2445 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2446                        struct tipc_link *link, int nlflags)
2447 {
2448         u32 self = tipc_own_addr(net);
2449         struct nlattr *attrs;
2450         struct nlattr *prop;
2451         void *hdr;
2452         int err;
2453
2454         hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2455                           nlflags, TIPC_NL_LINK_GET);
2456         if (!hdr)
2457                 return -EMSGSIZE;
2458
2459         attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2460         if (!attrs)
2461                 goto msg_full;
2462
2463         if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2464                 goto attr_msg_full;
2465         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2466                 goto attr_msg_full;
2467         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2468                 goto attr_msg_full;
2469         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2470                 goto attr_msg_full;
2471         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2472                 goto attr_msg_full;
2473
2474         if (tipc_link_is_up(link))
2475                 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2476                         goto attr_msg_full;
2477         if (link->active)
2478                 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2479                         goto attr_msg_full;
2480
2481         prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2482         if (!prop)
2483                 goto attr_msg_full;
2484         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2485                 goto prop_msg_full;
2486         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2487                 goto prop_msg_full;
2488         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2489                         link->window))
2490                 goto prop_msg_full;
2491         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2492                 goto prop_msg_full;
2493         nla_nest_end(msg->skb, prop);
2494
2495         err = __tipc_nl_add_stats(msg->skb, &link->stats);
2496         if (err)
2497                 goto attr_msg_full;
2498
2499         nla_nest_end(msg->skb, attrs);
2500         genlmsg_end(msg->skb, hdr);
2501
2502         return 0;
2503
2504 prop_msg_full:
2505         nla_nest_cancel(msg->skb, prop);
2506 attr_msg_full:
2507         nla_nest_cancel(msg->skb, attrs);
2508 msg_full:
2509         genlmsg_cancel(msg->skb, hdr);
2510
2511         return -EMSGSIZE;
2512 }
2513
2514 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2515                                       struct tipc_stats *stats)
2516 {
2517         int i;
2518         struct nlattr *nest;
2519
2520         struct nla_map {
2521                 __u32 key;
2522                 __u32 val;
2523         };
2524
2525         struct nla_map map[] = {
2526                 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2527                 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2528                 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2529                 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2530                 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2531                 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2532                 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2533                 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2534                 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2535                 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2536                 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2537                 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2538                 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2539                 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2540                 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2541                 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2542                 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2543                 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2544                 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2545                         (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2546         };
2547
2548         nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2549         if (!nest)
2550                 return -EMSGSIZE;
2551
2552         for (i = 0; i <  ARRAY_SIZE(map); i++)
2553                 if (nla_put_u32(skb, map[i].key, map[i].val))
2554                         goto msg_full;
2555
2556         nla_nest_end(skb, nest);
2557
2558         return 0;
2559 msg_full:
2560         nla_nest_cancel(skb, nest);
2561
2562         return -EMSGSIZE;
2563 }
2564
2565 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2566 {
2567         int err;
2568         void *hdr;
2569         struct nlattr *attrs;
2570         struct nlattr *prop;
2571         struct tipc_net *tn = net_generic(net, tipc_net_id);
2572         u32 bc_mode = tipc_bcast_get_broadcast_mode(net);
2573         u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2574         struct tipc_link *bcl = tn->bcl;
2575
2576         if (!bcl)
2577                 return 0;
2578
2579         tipc_bcast_lock(net);
2580
2581         hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2582                           NLM_F_MULTI, TIPC_NL_LINK_GET);
2583         if (!hdr) {
2584                 tipc_bcast_unlock(net);
2585                 return -EMSGSIZE;
2586         }
2587
2588         attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2589         if (!attrs)
2590                 goto msg_full;
2591
2592         /* The broadcast link is always up */
2593         if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2594                 goto attr_msg_full;
2595
2596         if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2597                 goto attr_msg_full;
2598         if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2599                 goto attr_msg_full;
2600         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2601                 goto attr_msg_full;
2602         if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2603                 goto attr_msg_full;
2604
2605         prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2606         if (!prop)
2607                 goto attr_msg_full;
2608         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2609                 goto prop_msg_full;
2610         if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2611                 goto prop_msg_full;
2612         if (bc_mode & BCLINK_MODE_SEL)
2613                 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2614                                 bc_ratio))
2615                         goto prop_msg_full;
2616         nla_nest_end(msg->skb, prop);
2617
2618         err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2619         if (err)
2620                 goto attr_msg_full;
2621
2622         tipc_bcast_unlock(net);
2623         nla_nest_end(msg->skb, attrs);
2624         genlmsg_end(msg->skb, hdr);
2625
2626         return 0;
2627
2628 prop_msg_full:
2629         nla_nest_cancel(msg->skb, prop);
2630 attr_msg_full:
2631         nla_nest_cancel(msg->skb, attrs);
2632 msg_full:
2633         tipc_bcast_unlock(net);
2634         genlmsg_cancel(msg->skb, hdr);
2635
2636         return -EMSGSIZE;
2637 }
2638
2639 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2640                              struct sk_buff_head *xmitq)
2641 {
2642         l->tolerance = tol;
2643         if (l->bc_rcvlink)
2644                 l->bc_rcvlink->tolerance = tol;
2645         if (link_is_up(l))
2646                 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2647 }
2648
2649 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2650                         struct sk_buff_head *xmitq)
2651 {
2652         l->priority = prio;
2653         tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2654 }
2655
2656 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2657 {
2658         l->abort_limit = limit;
2659 }
2660
2661 char *tipc_link_name_ext(struct tipc_link *l, char *buf)
2662 {
2663         if (!l)
2664                 scnprintf(buf, TIPC_MAX_LINK_NAME, "null");
2665         else if (link_is_bc_sndlink(l))
2666                 scnprintf(buf, TIPC_MAX_LINK_NAME, "broadcast-sender");
2667         else if (link_is_bc_rcvlink(l))
2668                 scnprintf(buf, TIPC_MAX_LINK_NAME,
2669                           "broadcast-receiver, peer %x", l->addr);
2670         else
2671                 memcpy(buf, l->name, TIPC_MAX_LINK_NAME);
2672
2673         return buf;
2674 }
2675
2676 /**
2677  * tipc_link_dump - dump TIPC link data
2678  * @l: tipc link to be dumped
2679  * @dqueues: bitmask to decide if any link queue to be dumped?
2680  *           - TIPC_DUMP_NONE: don't dump link queues
2681  *           - TIPC_DUMP_TRANSMQ: dump link transmq queue
2682  *           - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2683  *           - TIPC_DUMP_DEFERDQ: dump link deferd queue
2684  *           - TIPC_DUMP_INPUTQ: dump link input queue
2685  *           - TIPC_DUMP_WAKEUP: dump link wakeup queue
2686  *           - TIPC_DUMP_ALL: dump all the link queues above
2687  * @buf: returned buffer of dump data in format
2688  */
2689 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2690 {
2691         int i = 0;
2692         size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2693         struct sk_buff_head *list;
2694         struct sk_buff *hskb, *tskb;
2695         u32 len;
2696
2697         if (!l) {
2698                 i += scnprintf(buf, sz, "link data: (null)\n");
2699                 return i;
2700         }
2701
2702         i += scnprintf(buf, sz, "link data: %x", l->addr);
2703         i += scnprintf(buf + i, sz - i, " %x", l->state);
2704         i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2705         i += scnprintf(buf + i, sz - i, " %u", l->session);
2706         i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2707         i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2708         i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2709         i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2710         i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2711         i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2712         i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2713         i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2714         i += scnprintf(buf + i, sz - i, " %u", 0);
2715         i += scnprintf(buf + i, sz - i, " %u", 0);
2716         i += scnprintf(buf + i, sz - i, " %u", l->acked);
2717
2718         list = &l->transmq;
2719         len = skb_queue_len(list);
2720         hskb = skb_peek(list);
2721         tskb = skb_peek_tail(list);
2722         i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2723                        (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2724                        (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2725
2726         list = &l->deferdq;
2727         len = skb_queue_len(list);
2728         hskb = skb_peek(list);
2729         tskb = skb_peek_tail(list);
2730         i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2731                        (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2732                        (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2733
2734         list = &l->backlogq;
2735         len = skb_queue_len(list);
2736         hskb = skb_peek(list);
2737         tskb = skb_peek_tail(list);
2738         i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2739                        (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2740                        (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2741
2742         list = l->inputq;
2743         len = skb_queue_len(list);
2744         hskb = skb_peek(list);
2745         tskb = skb_peek_tail(list);
2746         i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2747                        (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2748                        (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2749
2750         if (dqueues & TIPC_DUMP_TRANSMQ) {
2751                 i += scnprintf(buf + i, sz - i, "transmq: ");
2752                 i += tipc_list_dump(&l->transmq, false, buf + i);
2753         }
2754         if (dqueues & TIPC_DUMP_BACKLOGQ) {
2755                 i += scnprintf(buf + i, sz - i,
2756                                "backlogq: <%u %u %u %u %u>, ",
2757                                l->backlog[TIPC_LOW_IMPORTANCE].len,
2758                                l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2759                                l->backlog[TIPC_HIGH_IMPORTANCE].len,
2760                                l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2761                                l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2762                 i += tipc_list_dump(&l->backlogq, false, buf + i);
2763         }
2764         if (dqueues & TIPC_DUMP_DEFERDQ) {
2765                 i += scnprintf(buf + i, sz - i, "deferdq: ");
2766                 i += tipc_list_dump(&l->deferdq, false, buf + i);
2767         }
2768         if (dqueues & TIPC_DUMP_INPUTQ) {
2769                 i += scnprintf(buf + i, sz - i, "inputq: ");
2770                 i += tipc_list_dump(l->inputq, false, buf + i);
2771         }
2772         if (dqueues & TIPC_DUMP_WAKEUP) {
2773                 i += scnprintf(buf + i, sz - i, "wakeup: ");
2774                 i += tipc_list_dump(&l->wakeupq, false, buf + i);
2775         }
2776
2777         return i;
2778 }