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