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