/net/tipc/link.c
C | 2855 lines | 2148 code | 312 blank | 395 comment | 376 complexity | 8e20a8e3c1b08ee70c2acc96714a19f1 MD5 | raw file
Large files files are truncated, but you can click here to view the full file
1/* 2 * net/tipc/link.c: TIPC link code 3 * 4 * Copyright (c) 1996-2007, 2012-2016, Ericsson AB 5 * Copyright (c) 2004-2007, 2010-2013, Wind River Systems 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions are met: 10 * 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the names of the copyright holders nor the names of its 17 * contributors may be used to endorse or promote products derived from 18 * this software without specific prior written permission. 19 * 20 * Alternatively, this software may be distributed under the terms of the 21 * GNU General Public License ("GPL") version 2 as published by the Free 22 * Software Foundation. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 34 * POSSIBILITY OF SUCH DAMAGE. 35 */ 36 37#include "core.h" 38#include "subscr.h" 39#include "link.h" 40#include "bcast.h" 41#include "socket.h" 42#include "name_distr.h" 43#include "discover.h" 44#include "netlink.h" 45#include "monitor.h" 46#include "trace.h" 47#include "crypto.h" 48 49#include <linux/pkt_sched.h> 50 51struct tipc_stats { 52 u32 sent_pkts; 53 u32 recv_pkts; 54 u32 sent_states; 55 u32 recv_states; 56 u32 sent_probes; 57 u32 recv_probes; 58 u32 sent_nacks; 59 u32 recv_nacks; 60 u32 sent_acks; 61 u32 sent_bundled; 62 u32 sent_bundles; 63 u32 recv_bundled; 64 u32 recv_bundles; 65 u32 retransmitted; 66 u32 sent_fragmented; 67 u32 sent_fragments; 68 u32 recv_fragmented; 69 u32 recv_fragments; 70 u32 link_congs; /* # port sends blocked by congestion */ 71 u32 deferred_recv; 72 u32 duplicates; 73 u32 max_queue_sz; /* send queue size high water mark */ 74 u32 accu_queue_sz; /* used for send queue size profiling */ 75 u32 queue_sz_counts; /* used for send queue size profiling */ 76 u32 msg_length_counts; /* used for message length profiling */ 77 u32 msg_lengths_total; /* used for message length profiling */ 78 u32 msg_length_profile[7]; /* used for msg. length profiling */ 79}; 80 81/** 82 * struct tipc_link - TIPC link data structure 83 * @addr: network address of link's peer node 84 * @name: link name character string 85 * @media_addr: media address to use when sending messages over link 86 * @timer: link timer 87 * @net: pointer to namespace struct 88 * @refcnt: reference counter for permanent references (owner node & timer) 89 * @peer_session: link session # being used by peer end of link 90 * @peer_bearer_id: bearer id used by link's peer endpoint 91 * @bearer_id: local bearer id used by link 92 * @tolerance: minimum link continuity loss needed to reset link [in ms] 93 * @abort_limit: # of unacknowledged continuity probes needed to reset link 94 * @state: current state of link FSM 95 * @peer_caps: bitmap describing capabilities of peer node 96 * @silent_intv_cnt: # of timer intervals without any reception from peer 97 * @proto_msg: template for control messages generated by link 98 * @pmsg: convenience pointer to "proto_msg" field 99 * @priority: current link priority 100 * @net_plane: current link network plane ('A' through 'H') 101 * @mon_state: cookie with information needed by link monitor 102 * @backlog_limit: backlog queue congestion thresholds (indexed by importance) 103 * @exp_msg_count: # of tunnelled messages expected during link changeover 104 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset 105 * @mtu: current maximum packet size for this link 106 * @advertised_mtu: advertised own mtu when link is being established 107 * @transmitq: queue for sent, non-acked messages 108 * @backlogq: queue for messages waiting to be sent 109 * @snt_nxt: next sequence number to use for outbound messages 110 * @ackers: # of peers that needs to ack each packet before it can be released 111 * @acked: # last packet acked by a certain peer. Used for broadcast. 112 * @rcv_nxt: next sequence number to expect for inbound messages 113 * @deferred_queue: deferred queue saved OOS b'cast message received from node 114 * @unacked_window: # of inbound messages rx'd without ack'ing back to peer 115 * @inputq: buffer queue for messages to be delivered upwards 116 * @namedq: buffer queue for name table messages to be delivered upwards 117 * @next_out: ptr to first unsent outbound message in queue 118 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate 119 * @long_msg_seq_no: next identifier to use for outbound fragmented messages 120 * @reasm_buf: head of partially reassembled inbound message fragments 121 * @bc_rcvr: marks that this is a broadcast receiver link 122 * @stats: collects statistics regarding link activity 123 */ 124struct tipc_link { 125 u32 addr; 126 char name[TIPC_MAX_LINK_NAME]; 127 struct net *net; 128 129 /* Management and link supervision data */ 130 u16 peer_session; 131 u16 session; 132 u16 snd_nxt_state; 133 u16 rcv_nxt_state; 134 u32 peer_bearer_id; 135 u32 bearer_id; 136 u32 tolerance; 137 u32 abort_limit; 138 u32 state; 139 u16 peer_caps; 140 bool in_session; 141 bool active; 142 u32 silent_intv_cnt; 143 char if_name[TIPC_MAX_IF_NAME]; 144 u32 priority; 145 char net_plane; 146 struct tipc_mon_state mon_state; 147 u16 rst_cnt; 148 149 /* Failover/synch */ 150 u16 drop_point; 151 struct sk_buff *failover_reasm_skb; 152 struct sk_buff_head failover_deferdq; 153 154 /* Max packet negotiation */ 155 u16 mtu; 156 u16 advertised_mtu; 157 158 /* Sending */ 159 struct sk_buff_head transmq; 160 struct sk_buff_head backlogq; 161 struct { 162 u16 len; 163 u16 limit; 164 struct sk_buff *target_bskb; 165 } backlog[5]; 166 u16 snd_nxt; 167 168 /* Reception */ 169 u16 rcv_nxt; 170 u32 rcv_unacked; 171 struct sk_buff_head deferdq; 172 struct sk_buff_head *inputq; 173 struct sk_buff_head *namedq; 174 175 /* Congestion handling */ 176 struct sk_buff_head wakeupq; 177 u16 window; 178 u16 min_win; 179 u16 ssthresh; 180 u16 max_win; 181 u16 cong_acks; 182 u16 checkpoint; 183 184 /* Fragmentation/reassembly */ 185 struct sk_buff *reasm_buf; 186 struct sk_buff *reasm_tnlmsg; 187 188 /* Broadcast */ 189 u16 ackers; 190 u16 acked; 191 struct tipc_link *bc_rcvlink; 192 struct tipc_link *bc_sndlink; 193 u8 nack_state; 194 bool bc_peer_is_up; 195 196 /* Statistics */ 197 struct tipc_stats stats; 198}; 199 200/* 201 * Error message prefixes 202 */ 203static const char *link_co_err = "Link tunneling error, "; 204static const char *link_rst_msg = "Resetting link "; 205 206/* Send states for broadcast NACKs 207 */ 208enum { 209 BC_NACK_SND_CONDITIONAL, 210 BC_NACK_SND_UNCONDITIONAL, 211 BC_NACK_SND_SUPPRESS, 212}; 213 214#define TIPC_BC_RETR_LIM (jiffies + msecs_to_jiffies(10)) 215#define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1)) 216 217/* 218 * Interval between NACKs when packets arrive out of order 219 */ 220#define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2) 221 222/* Link FSM states: 223 */ 224enum { 225 LINK_ESTABLISHED = 0xe, 226 LINK_ESTABLISHING = 0xe << 4, 227 LINK_RESET = 0x1 << 8, 228 LINK_RESETTING = 0x2 << 12, 229 LINK_PEER_RESET = 0xd << 16, 230 LINK_FAILINGOVER = 0xf << 20, 231 LINK_SYNCHING = 0xc << 24 232}; 233 234/* Link FSM state checking routines 235 */ 236static int link_is_up(struct tipc_link *l) 237{ 238 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING); 239} 240 241static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb, 242 struct sk_buff_head *xmitq); 243static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe, 244 bool probe_reply, u16 rcvgap, 245 int tolerance, int priority, 246 struct sk_buff_head *xmitq); 247static void link_print(struct tipc_link *l, const char *str); 248static int tipc_link_build_nack_msg(struct tipc_link *l, 249 struct sk_buff_head *xmitq); 250static void tipc_link_build_bc_init_msg(struct tipc_link *l, 251 struct sk_buff_head *xmitq); 252static int tipc_link_release_pkts(struct tipc_link *l, u16 to); 253static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data, u16 gap); 254static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap, 255 struct tipc_gap_ack_blks *ga, 256 struct sk_buff_head *xmitq); 257static void tipc_link_update_cwin(struct tipc_link *l, int released, 258 bool retransmitted); 259/* 260 * Simple non-static link routines (i.e. referenced outside this file) 261 */ 262bool tipc_link_is_up(struct tipc_link *l) 263{ 264 return link_is_up(l); 265} 266 267bool tipc_link_peer_is_down(struct tipc_link *l) 268{ 269 return l->state == LINK_PEER_RESET; 270} 271 272bool tipc_link_is_reset(struct tipc_link *l) 273{ 274 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING); 275} 276 277bool tipc_link_is_establishing(struct tipc_link *l) 278{ 279 return l->state == LINK_ESTABLISHING; 280} 281 282bool tipc_link_is_synching(struct tipc_link *l) 283{ 284 return l->state == LINK_SYNCHING; 285} 286 287bool tipc_link_is_failingover(struct tipc_link *l) 288{ 289 return l->state == LINK_FAILINGOVER; 290} 291 292bool tipc_link_is_blocked(struct tipc_link *l) 293{ 294 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER); 295} 296 297static bool link_is_bc_sndlink(struct tipc_link *l) 298{ 299 return !l->bc_sndlink; 300} 301 302static bool link_is_bc_rcvlink(struct tipc_link *l) 303{ 304 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l)); 305} 306 307void tipc_link_set_active(struct tipc_link *l, bool active) 308{ 309 l->active = active; 310} 311 312u32 tipc_link_id(struct tipc_link *l) 313{ 314 return l->peer_bearer_id << 16 | l->bearer_id; 315} 316 317int tipc_link_min_win(struct tipc_link *l) 318{ 319 return l->min_win; 320} 321 322int tipc_link_max_win(struct tipc_link *l) 323{ 324 return l->max_win; 325} 326 327int tipc_link_prio(struct tipc_link *l) 328{ 329 return l->priority; 330} 331 332unsigned long tipc_link_tolerance(struct tipc_link *l) 333{ 334 return l->tolerance; 335} 336 337struct sk_buff_head *tipc_link_inputq(struct tipc_link *l) 338{ 339 return l->inputq; 340} 341 342char tipc_link_plane(struct tipc_link *l) 343{ 344 return l->net_plane; 345} 346 347void tipc_link_update_caps(struct tipc_link *l, u16 capabilities) 348{ 349 l->peer_caps = capabilities; 350} 351 352void tipc_link_add_bc_peer(struct tipc_link *snd_l, 353 struct tipc_link *uc_l, 354 struct sk_buff_head *xmitq) 355{ 356 struct tipc_link *rcv_l = uc_l->bc_rcvlink; 357 358 snd_l->ackers++; 359 rcv_l->acked = snd_l->snd_nxt - 1; 360 snd_l->state = LINK_ESTABLISHED; 361 tipc_link_build_bc_init_msg(uc_l, xmitq); 362} 363 364void tipc_link_remove_bc_peer(struct tipc_link *snd_l, 365 struct tipc_link *rcv_l, 366 struct sk_buff_head *xmitq) 367{ 368 u16 ack = snd_l->snd_nxt - 1; 369 370 snd_l->ackers--; 371 rcv_l->bc_peer_is_up = true; 372 rcv_l->state = LINK_ESTABLISHED; 373 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq); 374 trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!"); 375 tipc_link_reset(rcv_l); 376 rcv_l->state = LINK_RESET; 377 if (!snd_l->ackers) { 378 trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!"); 379 tipc_link_reset(snd_l); 380 snd_l->state = LINK_RESET; 381 __skb_queue_purge(xmitq); 382 } 383} 384 385int tipc_link_bc_peers(struct tipc_link *l) 386{ 387 return l->ackers; 388} 389 390static u16 link_bc_rcv_gap(struct tipc_link *l) 391{ 392 struct sk_buff *skb = skb_peek(&l->deferdq); 393 u16 gap = 0; 394 395 if (more(l->snd_nxt, l->rcv_nxt)) 396 gap = l->snd_nxt - l->rcv_nxt; 397 if (skb) 398 gap = buf_seqno(skb) - l->rcv_nxt; 399 return gap; 400} 401 402void tipc_link_set_mtu(struct tipc_link *l, int mtu) 403{ 404 l->mtu = mtu; 405} 406 407int tipc_link_mtu(struct tipc_link *l) 408{ 409 return l->mtu; 410} 411 412int tipc_link_mss(struct tipc_link *l) 413{ 414#ifdef CONFIG_TIPC_CRYPTO 415 return l->mtu - INT_H_SIZE - EMSG_OVERHEAD; 416#else 417 return l->mtu - INT_H_SIZE; 418#endif 419} 420 421u16 tipc_link_rcv_nxt(struct tipc_link *l) 422{ 423 return l->rcv_nxt; 424} 425 426u16 tipc_link_acked(struct tipc_link *l) 427{ 428 return l->acked; 429} 430 431char *tipc_link_name(struct tipc_link *l) 432{ 433 return l->name; 434} 435 436u32 tipc_link_state(struct tipc_link *l) 437{ 438 return l->state; 439} 440 441/** 442 * tipc_link_create - create a new link 443 * @n: pointer to associated node 444 * @if_name: associated interface name 445 * @bearer_id: id (index) of associated bearer 446 * @tolerance: link tolerance to be used by link 447 * @net_plane: network plane (A,B,c..) this link belongs to 448 * @mtu: mtu to be advertised by link 449 * @priority: priority to be used by link 450 * @min_win: minimal send window to be used by link 451 * @max_win: maximal send window to be used by link 452 * @session: session to be used by link 453 * @ownnode: identity of own node 454 * @peer: node id of peer node 455 * @peer_caps: bitmap describing peer node capabilities 456 * @bc_sndlink: the namespace global link used for broadcast sending 457 * @bc_rcvlink: the peer specific link used for broadcast reception 458 * @inputq: queue to put messages ready for delivery 459 * @namedq: queue to put binding table update messages ready for delivery 460 * @link: return value, pointer to put the created link 461 * 462 * Returns true if link was created, otherwise false 463 */ 464bool tipc_link_create(struct net *net, char *if_name, int bearer_id, 465 int tolerance, char net_plane, u32 mtu, int priority, 466 u32 min_win, u32 max_win, u32 session, u32 self, 467 u32 peer, u8 *peer_id, u16 peer_caps, 468 struct tipc_link *bc_sndlink, 469 struct tipc_link *bc_rcvlink, 470 struct sk_buff_head *inputq, 471 struct sk_buff_head *namedq, 472 struct tipc_link **link) 473{ 474 char peer_str[NODE_ID_STR_LEN] = {0,}; 475 char self_str[NODE_ID_STR_LEN] = {0,}; 476 struct tipc_link *l; 477 478 l = kzalloc(sizeof(*l), GFP_ATOMIC); 479 if (!l) 480 return false; 481 *link = l; 482 l->session = session; 483 484 /* Set link name for unicast links only */ 485 if (peer_id) { 486 tipc_nodeid2string(self_str, tipc_own_id(net)); 487 if (strlen(self_str) > 16) 488 sprintf(self_str, "%x", self); 489 tipc_nodeid2string(peer_str, peer_id); 490 if (strlen(peer_str) > 16) 491 sprintf(peer_str, "%x", peer); 492 } 493 /* Peer i/f name will be completed by reset/activate message */ 494 snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown", 495 self_str, if_name, peer_str); 496 497 strcpy(l->if_name, if_name); 498 l->addr = peer; 499 l->peer_caps = peer_caps; 500 l->net = net; 501 l->in_session = false; 502 l->bearer_id = bearer_id; 503 l->tolerance = tolerance; 504 if (bc_rcvlink) 505 bc_rcvlink->tolerance = tolerance; 506 l->net_plane = net_plane; 507 l->advertised_mtu = mtu; 508 l->mtu = mtu; 509 l->priority = priority; 510 tipc_link_set_queue_limits(l, min_win, max_win); 511 l->ackers = 1; 512 l->bc_sndlink = bc_sndlink; 513 l->bc_rcvlink = bc_rcvlink; 514 l->inputq = inputq; 515 l->namedq = namedq; 516 l->state = LINK_RESETTING; 517 __skb_queue_head_init(&l->transmq); 518 __skb_queue_head_init(&l->backlogq); 519 __skb_queue_head_init(&l->deferdq); 520 __skb_queue_head_init(&l->failover_deferdq); 521 skb_queue_head_init(&l->wakeupq); 522 skb_queue_head_init(l->inputq); 523 return true; 524} 525 526/** 527 * tipc_link_bc_create - create new link to be used for broadcast 528 * @n: pointer to associated node 529 * @mtu: mtu to be used initially if no peers 530 * @window: send window to be used 531 * @inputq: queue to put messages ready for delivery 532 * @namedq: queue to put binding table update messages ready for delivery 533 * @link: return value, pointer to put the created link 534 * 535 * Returns true if link was created, otherwise false 536 */ 537bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer, 538 int mtu, u32 min_win, u32 max_win, u16 peer_caps, 539 struct sk_buff_head *inputq, 540 struct sk_buff_head *namedq, 541 struct tipc_link *bc_sndlink, 542 struct tipc_link **link) 543{ 544 struct tipc_link *l; 545 546 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, min_win, 547 max_win, 0, ownnode, peer, NULL, peer_caps, 548 bc_sndlink, NULL, inputq, namedq, link)) 549 return false; 550 551 l = *link; 552 strcpy(l->name, tipc_bclink_name); 553 trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!"); 554 tipc_link_reset(l); 555 l->state = LINK_RESET; 556 l->ackers = 0; 557 l->bc_rcvlink = l; 558 559 /* Broadcast send link is always up */ 560 if (link_is_bc_sndlink(l)) 561 l->state = LINK_ESTABLISHED; 562 563 /* Disable replicast if even a single peer doesn't support it */ 564 if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST)) 565 tipc_bcast_toggle_rcast(net, false); 566 567 return true; 568} 569 570/** 571 * tipc_link_fsm_evt - link finite state machine 572 * @l: pointer to link 573 * @evt: state machine event to be processed 574 */ 575int tipc_link_fsm_evt(struct tipc_link *l, int evt) 576{ 577 int rc = 0; 578 int old_state = l->state; 579 580 switch (l->state) { 581 case LINK_RESETTING: 582 switch (evt) { 583 case LINK_PEER_RESET_EVT: 584 l->state = LINK_PEER_RESET; 585 break; 586 case LINK_RESET_EVT: 587 l->state = LINK_RESET; 588 break; 589 case LINK_FAILURE_EVT: 590 case LINK_FAILOVER_BEGIN_EVT: 591 case LINK_ESTABLISH_EVT: 592 case LINK_FAILOVER_END_EVT: 593 case LINK_SYNCH_BEGIN_EVT: 594 case LINK_SYNCH_END_EVT: 595 default: 596 goto illegal_evt; 597 } 598 break; 599 case LINK_RESET: 600 switch (evt) { 601 case LINK_PEER_RESET_EVT: 602 l->state = LINK_ESTABLISHING; 603 break; 604 case LINK_FAILOVER_BEGIN_EVT: 605 l->state = LINK_FAILINGOVER; 606 case LINK_FAILURE_EVT: 607 case LINK_RESET_EVT: 608 case LINK_ESTABLISH_EVT: 609 case LINK_FAILOVER_END_EVT: 610 break; 611 case LINK_SYNCH_BEGIN_EVT: 612 case LINK_SYNCH_END_EVT: 613 default: 614 goto illegal_evt; 615 } 616 break; 617 case LINK_PEER_RESET: 618 switch (evt) { 619 case LINK_RESET_EVT: 620 l->state = LINK_ESTABLISHING; 621 break; 622 case LINK_PEER_RESET_EVT: 623 case LINK_ESTABLISH_EVT: 624 case LINK_FAILURE_EVT: 625 break; 626 case LINK_SYNCH_BEGIN_EVT: 627 case LINK_SYNCH_END_EVT: 628 case LINK_FAILOVER_BEGIN_EVT: 629 case LINK_FAILOVER_END_EVT: 630 default: 631 goto illegal_evt; 632 } 633 break; 634 case LINK_FAILINGOVER: 635 switch (evt) { 636 case LINK_FAILOVER_END_EVT: 637 l->state = LINK_RESET; 638 break; 639 case LINK_PEER_RESET_EVT: 640 case LINK_RESET_EVT: 641 case LINK_ESTABLISH_EVT: 642 case LINK_FAILURE_EVT: 643 break; 644 case LINK_FAILOVER_BEGIN_EVT: 645 case LINK_SYNCH_BEGIN_EVT: 646 case LINK_SYNCH_END_EVT: 647 default: 648 goto illegal_evt; 649 } 650 break; 651 case LINK_ESTABLISHING: 652 switch (evt) { 653 case LINK_ESTABLISH_EVT: 654 l->state = LINK_ESTABLISHED; 655 break; 656 case LINK_FAILOVER_BEGIN_EVT: 657 l->state = LINK_FAILINGOVER; 658 break; 659 case LINK_RESET_EVT: 660 l->state = LINK_RESET; 661 break; 662 case LINK_FAILURE_EVT: 663 case LINK_PEER_RESET_EVT: 664 case LINK_SYNCH_BEGIN_EVT: 665 case LINK_FAILOVER_END_EVT: 666 break; 667 case LINK_SYNCH_END_EVT: 668 default: 669 goto illegal_evt; 670 } 671 break; 672 case LINK_ESTABLISHED: 673 switch (evt) { 674 case LINK_PEER_RESET_EVT: 675 l->state = LINK_PEER_RESET; 676 rc |= TIPC_LINK_DOWN_EVT; 677 break; 678 case LINK_FAILURE_EVT: 679 l->state = LINK_RESETTING; 680 rc |= TIPC_LINK_DOWN_EVT; 681 break; 682 case LINK_RESET_EVT: 683 l->state = LINK_RESET; 684 break; 685 case LINK_ESTABLISH_EVT: 686 case LINK_SYNCH_END_EVT: 687 break; 688 case LINK_SYNCH_BEGIN_EVT: 689 l->state = LINK_SYNCHING; 690 break; 691 case LINK_FAILOVER_BEGIN_EVT: 692 case LINK_FAILOVER_END_EVT: 693 default: 694 goto illegal_evt; 695 } 696 break; 697 case LINK_SYNCHING: 698 switch (evt) { 699 case LINK_PEER_RESET_EVT: 700 l->state = LINK_PEER_RESET; 701 rc |= TIPC_LINK_DOWN_EVT; 702 break; 703 case LINK_FAILURE_EVT: 704 l->state = LINK_RESETTING; 705 rc |= TIPC_LINK_DOWN_EVT; 706 break; 707 case LINK_RESET_EVT: 708 l->state = LINK_RESET; 709 break; 710 case LINK_ESTABLISH_EVT: 711 case LINK_SYNCH_BEGIN_EVT: 712 break; 713 case LINK_SYNCH_END_EVT: 714 l->state = LINK_ESTABLISHED; 715 break; 716 case LINK_FAILOVER_BEGIN_EVT: 717 case LINK_FAILOVER_END_EVT: 718 default: 719 goto illegal_evt; 720 } 721 break; 722 default: 723 pr_err("Unknown FSM state %x in %s\n", l->state, l->name); 724 } 725 trace_tipc_link_fsm(l->name, old_state, l->state, evt); 726 return rc; 727illegal_evt: 728 pr_err("Illegal FSM event %x in state %x on link %s\n", 729 evt, l->state, l->name); 730 trace_tipc_link_fsm(l->name, old_state, l->state, evt); 731 return rc; 732} 733 734/* link_profile_stats - update statistical profiling of traffic 735 */ 736static void link_profile_stats(struct tipc_link *l) 737{ 738 struct sk_buff *skb; 739 struct tipc_msg *msg; 740 int length; 741 742 /* Update counters used in statistical profiling of send traffic */ 743 l->stats.accu_queue_sz += skb_queue_len(&l->transmq); 744 l->stats.queue_sz_counts++; 745 746 skb = skb_peek(&l->transmq); 747 if (!skb) 748 return; 749 msg = buf_msg(skb); 750 length = msg_size(msg); 751 752 if (msg_user(msg) == MSG_FRAGMENTER) { 753 if (msg_type(msg) != FIRST_FRAGMENT) 754 return; 755 length = msg_size(msg_inner_hdr(msg)); 756 } 757 l->stats.msg_lengths_total += length; 758 l->stats.msg_length_counts++; 759 if (length <= 64) 760 l->stats.msg_length_profile[0]++; 761 else if (length <= 256) 762 l->stats.msg_length_profile[1]++; 763 else if (length <= 1024) 764 l->stats.msg_length_profile[2]++; 765 else if (length <= 4096) 766 l->stats.msg_length_profile[3]++; 767 else if (length <= 16384) 768 l->stats.msg_length_profile[4]++; 769 else if (length <= 32768) 770 l->stats.msg_length_profile[5]++; 771 else 772 l->stats.msg_length_profile[6]++; 773} 774 775/** 776 * tipc_link_too_silent - check if link is "too silent" 777 * @l: tipc link to be checked 778 * 779 * Returns true if the link 'silent_intv_cnt' is about to reach the 780 * 'abort_limit' value, otherwise false 781 */ 782bool tipc_link_too_silent(struct tipc_link *l) 783{ 784 return (l->silent_intv_cnt + 2 > l->abort_limit); 785} 786 787static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r, 788 u16 from, u16 to, struct sk_buff_head *xmitq); 789/* tipc_link_timeout - perform periodic task as instructed from node timeout 790 */ 791int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq) 792{ 793 int mtyp = 0; 794 int rc = 0; 795 bool state = false; 796 bool probe = false; 797 bool setup = false; 798 u16 bc_snt = l->bc_sndlink->snd_nxt - 1; 799 u16 bc_acked = l->bc_rcvlink->acked; 800 struct tipc_mon_state *mstate = &l->mon_state; 801 802 trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " "); 803 trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " "); 804 switch (l->state) { 805 case LINK_ESTABLISHED: 806 case LINK_SYNCHING: 807 mtyp = STATE_MSG; 808 link_profile_stats(l); 809 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id); 810 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit)) 811 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT); 812 state = bc_acked != bc_snt; 813 state |= l->bc_rcvlink->rcv_unacked; 814 state |= l->rcv_unacked; 815 state |= !skb_queue_empty(&l->transmq); 816 state |= !skb_queue_empty(&l->deferdq); 817 probe = mstate->probing; 818 probe |= l->silent_intv_cnt; 819 if (probe || mstate->monitoring) 820 l->silent_intv_cnt++; 821 if (l->snd_nxt == l->checkpoint) { 822 tipc_link_update_cwin(l, 0, 0); 823 probe = true; 824 } 825 l->checkpoint = l->snd_nxt; 826 break; 827 case LINK_RESET: 828 setup = l->rst_cnt++ <= 4; 829 setup |= !(l->rst_cnt % 16); 830 mtyp = RESET_MSG; 831 break; 832 case LINK_ESTABLISHING: 833 setup = true; 834 mtyp = ACTIVATE_MSG; 835 break; 836 case LINK_PEER_RESET: 837 case LINK_RESETTING: 838 case LINK_FAILINGOVER: 839 break; 840 default: 841 break; 842 } 843 844 if (state || probe || setup) 845 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq); 846 847 return rc; 848} 849 850/** 851 * link_schedule_user - schedule a message sender for wakeup after congestion 852 * @l: congested link 853 * @hdr: header of message that is being sent 854 * Create pseudo msg to send back to user when congestion abates 855 */ 856static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr) 857{ 858 u32 dnode = tipc_own_addr(l->net); 859 u32 dport = msg_origport(hdr); 860 struct sk_buff *skb; 861 862 /* Create and schedule wakeup pseudo message */ 863 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0, 864 dnode, l->addr, dport, 0, 0); 865 if (!skb) 866 return -ENOBUFS; 867 msg_set_dest_droppable(buf_msg(skb), true); 868 TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr); 869 skb_queue_tail(&l->wakeupq, skb); 870 l->stats.link_congs++; 871 trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!"); 872 return -ELINKCONG; 873} 874 875/** 876 * link_prepare_wakeup - prepare users for wakeup after congestion 877 * @l: congested link 878 * Wake up a number of waiting users, as permitted by available space 879 * in the send queue 880 */ 881static void link_prepare_wakeup(struct tipc_link *l) 882{ 883 struct sk_buff_head *wakeupq = &l->wakeupq; 884 struct sk_buff_head *inputq = l->inputq; 885 struct sk_buff *skb, *tmp; 886 struct sk_buff_head tmpq; 887 int avail[5] = {0,}; 888 int imp = 0; 889 890 __skb_queue_head_init(&tmpq); 891 892 for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) 893 avail[imp] = l->backlog[imp].limit - l->backlog[imp].len; 894 895 skb_queue_walk_safe(wakeupq, skb, tmp) { 896 imp = TIPC_SKB_CB(skb)->chain_imp; 897 if (avail[imp] <= 0) 898 continue; 899 avail[imp]--; 900 __skb_unlink(skb, wakeupq); 901 __skb_queue_tail(&tmpq, skb); 902 } 903 904 spin_lock_bh(&inputq->lock); 905 skb_queue_splice_tail(&tmpq, inputq); 906 spin_unlock_bh(&inputq->lock); 907 908} 909 910void tipc_link_reset(struct tipc_link *l) 911{ 912 struct sk_buff_head list; 913 u32 imp; 914 915 __skb_queue_head_init(&list); 916 917 l->in_session = false; 918 /* Force re-synch of peer session number before establishing */ 919 l->peer_session--; 920 l->session++; 921 l->mtu = l->advertised_mtu; 922 923 spin_lock_bh(&l->wakeupq.lock); 924 skb_queue_splice_init(&l->wakeupq, &list); 925 spin_unlock_bh(&l->wakeupq.lock); 926 927 spin_lock_bh(&l->inputq->lock); 928 skb_queue_splice_init(&list, l->inputq); 929 spin_unlock_bh(&l->inputq->lock); 930 931 __skb_queue_purge(&l->transmq); 932 __skb_queue_purge(&l->deferdq); 933 __skb_queue_purge(&l->backlogq); 934 __skb_queue_purge(&l->failover_deferdq); 935 for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) { 936 l->backlog[imp].len = 0; 937 l->backlog[imp].target_bskb = NULL; 938 } 939 kfree_skb(l->reasm_buf); 940 kfree_skb(l->reasm_tnlmsg); 941 kfree_skb(l->failover_reasm_skb); 942 l->reasm_buf = NULL; 943 l->reasm_tnlmsg = NULL; 944 l->failover_reasm_skb = NULL; 945 l->rcv_unacked = 0; 946 l->snd_nxt = 1; 947 l->rcv_nxt = 1; 948 l->snd_nxt_state = 1; 949 l->rcv_nxt_state = 1; 950 l->acked = 0; 951 l->silent_intv_cnt = 0; 952 l->rst_cnt = 0; 953 l->bc_peer_is_up = false; 954 memset(&l->mon_state, 0, sizeof(l->mon_state)); 955 tipc_link_reset_stats(l); 956} 957 958/** 959 * tipc_link_xmit(): enqueue buffer list according to queue situation 960 * @link: link to use 961 * @list: chain of buffers containing message 962 * @xmitq: returned list of packets to be sent by caller 963 * 964 * Consumes the buffer chain. 965 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS 966 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted 967 */ 968int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list, 969 struct sk_buff_head *xmitq) 970{ 971 struct tipc_msg *hdr = buf_msg(skb_peek(list)); 972 struct sk_buff_head *backlogq = &l->backlogq; 973 struct sk_buff_head *transmq = &l->transmq; 974 struct sk_buff *skb, *_skb; 975 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1; 976 u16 ack = l->rcv_nxt - 1; 977 u16 seqno = l->snd_nxt; 978 int pkt_cnt = skb_queue_len(list); 979 int imp = msg_importance(hdr); 980 unsigned int mss = tipc_link_mss(l); 981 unsigned int cwin = l->window; 982 unsigned int mtu = l->mtu; 983 bool new_bundle; 984 int rc = 0; 985 986 if (unlikely(msg_size(hdr) > mtu)) { 987 pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n", 988 skb_queue_len(list), msg_user(hdr), 989 msg_type(hdr), msg_size(hdr), mtu); 990 __skb_queue_purge(list); 991 return -EMSGSIZE; 992 } 993 994 /* Allow oversubscription of one data msg per source at congestion */ 995 if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) { 996 if (imp == TIPC_SYSTEM_IMPORTANCE) { 997 pr_warn("%s<%s>, link overflow", link_rst_msg, l->name); 998 return -ENOBUFS; 999 } 1000 rc = link_schedule_user(l, hdr); 1001 } 1002 1003 if (pkt_cnt > 1) { 1004 l->stats.sent_fragmented++; 1005 l->stats.sent_fragments += pkt_cnt; 1006 } 1007 1008 /* Prepare each packet for sending, and add to relevant queue: */ 1009 while ((skb = __skb_dequeue(list))) { 1010 if (likely(skb_queue_len(transmq) < cwin)) { 1011 hdr = buf_msg(skb); 1012 msg_set_seqno(hdr, seqno); 1013 msg_set_ack(hdr, ack); 1014 msg_set_bcast_ack(hdr, bc_ack); 1015 _skb = skb_clone(skb, GFP_ATOMIC); 1016 if (!_skb) { 1017 kfree_skb(skb); 1018 __skb_queue_purge(list); 1019 return -ENOBUFS; 1020 } 1021 __skb_queue_tail(transmq, skb); 1022 /* next retransmit attempt */ 1023 if (link_is_bc_sndlink(l)) 1024 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM; 1025 __skb_queue_tail(xmitq, _skb); 1026 TIPC_SKB_CB(skb)->ackers = l->ackers; 1027 l->rcv_unacked = 0; 1028 l->stats.sent_pkts++; 1029 seqno++; 1030 continue; 1031 } 1032 if (tipc_msg_try_bundle(l->backlog[imp].target_bskb, &skb, 1033 mss, l->addr, &new_bundle)) { 1034 if (skb) { 1035 /* Keep a ref. to the skb for next try */ 1036 l->backlog[imp].target_bskb = skb; 1037 l->backlog[imp].len++; 1038 __skb_queue_tail(backlogq, skb); 1039 } else { 1040 if (new_bundle) { 1041 l->stats.sent_bundles++; 1042 l->stats.sent_bundled++; 1043 } 1044 l->stats.sent_bundled++; 1045 } 1046 continue; 1047 } 1048 l->backlog[imp].target_bskb = NULL; 1049 l->backlog[imp].len += (1 + skb_queue_len(list)); 1050 __skb_queue_tail(backlogq, skb); 1051 skb_queue_splice_tail_init(list, backlogq); 1052 } 1053 l->snd_nxt = seqno; 1054 return rc; 1055} 1056 1057static void tipc_link_update_cwin(struct tipc_link *l, int released, 1058 bool retransmitted) 1059{ 1060 int bklog_len = skb_queue_len(&l->backlogq); 1061 struct sk_buff_head *txq = &l->transmq; 1062 int txq_len = skb_queue_len(txq); 1063 u16 cwin = l->window; 1064 1065 /* Enter fast recovery */ 1066 if (unlikely(retransmitted)) { 1067 l->ssthresh = max_t(u16, l->window / 2, 300); 1068 l->window = min_t(u16, l->ssthresh, l->window); 1069 return; 1070 } 1071 /* Enter slow start */ 1072 if (unlikely(!released)) { 1073 l->ssthresh = max_t(u16, l->window / 2, 300); 1074 l->window = l->min_win; 1075 return; 1076 } 1077 /* Don't increase window if no pressure on the transmit queue */ 1078 if (txq_len + bklog_len < cwin) 1079 return; 1080 1081 /* Don't increase window if there are holes the transmit queue */ 1082 if (txq_len && l->snd_nxt - buf_seqno(skb_peek(txq)) != txq_len) 1083 return; 1084 1085 l->cong_acks += released; 1086 1087 /* Slow start */ 1088 if (cwin <= l->ssthresh) { 1089 l->window = min_t(u16, cwin + released, l->max_win); 1090 return; 1091 } 1092 /* Congestion avoidance */ 1093 if (l->cong_acks < cwin) 1094 return; 1095 l->window = min_t(u16, ++cwin, l->max_win); 1096 l->cong_acks = 0; 1097} 1098 1099static void tipc_link_advance_backlog(struct tipc_link *l, 1100 struct sk_buff_head *xmitq) 1101{ 1102 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1; 1103 struct sk_buff_head *txq = &l->transmq; 1104 struct sk_buff *skb, *_skb; 1105 u16 ack = l->rcv_nxt - 1; 1106 u16 seqno = l->snd_nxt; 1107 struct tipc_msg *hdr; 1108 u16 cwin = l->window; 1109 u32 imp; 1110 1111 while (skb_queue_len(txq) < cwin) { 1112 skb = skb_peek(&l->backlogq); 1113 if (!skb) 1114 break; 1115 _skb = skb_clone(skb, GFP_ATOMIC); 1116 if (!_skb) 1117 break; 1118 __skb_dequeue(&l->backlogq); 1119 hdr = buf_msg(skb); 1120 imp = msg_importance(hdr); 1121 l->backlog[imp].len--; 1122 if (unlikely(skb == l->backlog[imp].target_bskb)) 1123 l->backlog[imp].target_bskb = NULL; 1124 __skb_queue_tail(&l->transmq, skb); 1125 /* next retransmit attempt */ 1126 if (link_is_bc_sndlink(l)) 1127 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM; 1128 1129 __skb_queue_tail(xmitq, _skb); 1130 TIPC_SKB_CB(skb)->ackers = l->ackers; 1131 msg_set_seqno(hdr, seqno); 1132 msg_set_ack(hdr, ack); 1133 msg_set_bcast_ack(hdr, bc_ack); 1134 l->rcv_unacked = 0; 1135 l->stats.sent_pkts++; 1136 seqno++; 1137 } 1138 l->snd_nxt = seqno; 1139} 1140 1141/** 1142 * link_retransmit_failure() - Detect repeated retransmit failures 1143 * @l: tipc link sender 1144 * @r: tipc link receiver (= l in case of unicast) 1145 * @rc: returned code 1146 * 1147 * Return: true if the repeated retransmit failures happens, otherwise 1148 * false 1149 */ 1150static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r, 1151 int *rc) 1152{ 1153 struct sk_buff *skb = skb_peek(&l->transmq); 1154 struct tipc_msg *hdr; 1155 1156 if (!skb) 1157 return false; 1158 1159 if (!TIPC_SKB_CB(skb)->retr_cnt) 1160 return false; 1161 1162 if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp + 1163 msecs_to_jiffies(r->tolerance * 10))) 1164 return false; 1165 1166 hdr = buf_msg(skb); 1167 if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr))) 1168 return false; 1169 1170 pr_warn("Retransmission failure on link <%s>\n", l->name); 1171 link_print(l, "State of link "); 1172 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n", 1173 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr)); 1174 pr_info("sqno %u, prev: %x, dest: %x\n", 1175 msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr)); 1176 pr_info("retr_stamp %d, retr_cnt %d\n", 1177 jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp), 1178 TIPC_SKB_CB(skb)->retr_cnt); 1179 1180 trace_tipc_list_dump(&l->transmq, true, "retrans failure!"); 1181 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!"); 1182 trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!"); 1183 1184 if (link_is_bc_sndlink(l)) { 1185 r->state = LINK_RESET; 1186 *rc = TIPC_LINK_DOWN_EVT; 1187 } else { 1188 *rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT); 1189 } 1190 1191 return true; 1192} 1193 1194/* tipc_link_bc_retrans() - retransmit zero or more packets 1195 * @l: the link to transmit on 1196 * @r: the receiving link ordering the retransmit. Same as l if unicast 1197 * @from: retransmit from (inclusive) this sequence number 1198 * @to: retransmit to (inclusive) this sequence number 1199 * xmitq: queue for accumulating the retransmitted packets 1200 */ 1201static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r, 1202 u16 from, u16 to, struct sk_buff_head *xmitq) 1203{ 1204 struct sk_buff *_skb, *skb = skb_peek(&l->transmq); 1205 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1; 1206 u16 ack = l->rcv_nxt - 1; 1207 int retransmitted = 0; 1208 struct tipc_msg *hdr; 1209 int rc = 0; 1210 1211 if (!skb) 1212 return 0; 1213 if (less(to, from)) 1214 return 0; 1215 1216 trace_tipc_link_retrans(r, from, to, &l->transmq); 1217 1218 if (link_retransmit_failure(l, r, &rc)) 1219 return rc; 1220 1221 skb_queue_walk(&l->transmq, skb) { 1222 hdr = buf_msg(skb); 1223 if (less(msg_seqno(hdr), from)) 1224 continue; 1225 if (more(msg_seqno(hdr), to)) 1226 break; 1227 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr)) 1228 continue; 1229 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM; 1230 _skb = pskb_copy(skb, GFP_ATOMIC); 1231 if (!_skb) 1232 return 0; 1233 hdr = buf_msg(_skb); 1234 msg_set_ack(hdr, ack); 1235 msg_set_bcast_ack(hdr, bc_ack); 1236 _skb->priority = TC_PRIO_CONTROL; 1237 __skb_queue_tail(xmitq, _skb); 1238 l->stats.retransmitted++; 1239 retransmitted++; 1240 /* Increase actual retrans counter & mark first time */ 1241 if (!TIPC_SKB_CB(skb)->retr_cnt++) 1242 TIPC_SKB_CB(skb)->retr_stamp = jiffies; 1243 } 1244 tipc_link_update_cwin(l, 0, retransmitted); 1245 return 0; 1246} 1247 1248/* tipc_data_input - deliver data and name distr msgs to upper layer 1249 * 1250 * Consumes buffer if message is of right type 1251 * Node lock must be held 1252 */ 1253static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb, 1254 struct sk_buff_head *inputq) 1255{ 1256 struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq; 1257 struct tipc_msg *hdr = buf_msg(skb); 1258 1259 switch (msg_user(hdr)) { 1260 case TIPC_LOW_IMPORTANCE: 1261 case TIPC_MEDIUM_IMPORTANCE: 1262 case TIPC_HIGH_IMPORTANCE: 1263 case TIPC_CRITICAL_IMPORTANCE: 1264 if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) { 1265 skb_queue_tail(mc_inputq, skb); 1266 return true; 1267 } 1268 /* fall through */ 1269 case CONN_MANAGER: 1270 skb_queue_tail(inputq, skb); 1271 return true; 1272 case GROUP_PROTOCOL: 1273 skb_queue_tail(mc_inputq, skb); 1274 return true; 1275 case NAME_DISTRIBUTOR: 1276 l->bc_rcvlink->state = LINK_ESTABLISHED; 1277 skb_queue_tail(l->namedq, skb); 1278 return true; 1279 case MSG_BUNDLER: 1280 case TUNNEL_PROTOCOL: 1281 case MSG_FRAGMENTER: 1282 case BCAST_PROTOCOL: 1283 return false; 1284 default: 1285 pr_warn("Dropping received illegal msg type\n"); 1286 kfree_skb(skb); 1287 return true; 1288 }; 1289} 1290 1291/* tipc_link_input - process packet that has passed link protocol check 1292 * 1293 * Consumes buffer 1294 */ 1295static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb, 1296 struct sk_buff_head *inputq, 1297 struct sk_buff **reasm_skb) 1298{ 1299 struct tipc_msg *hdr = buf_msg(skb); 1300 struct sk_buff *iskb; 1301 struct sk_buff_head tmpq; 1302 int usr = msg_user(hdr); 1303 int pos = 0; 1304 1305 if (usr == MSG_BUNDLER) { 1306 skb_queue_head_init(&tmpq); 1307 l->stats.recv_bundles++; 1308 l->stats.recv_bundled += msg_msgcnt(hdr); 1309 while (tipc_msg_extract(skb, &iskb, &pos)) 1310 tipc_data_input(l, iskb, &tmpq); 1311 tipc_skb_queue_splice_tail(&tmpq, inputq); 1312 return 0; 1313 } else if (usr == MSG_FRAGMENTER) { 1314 l->stats.recv_fragments++; 1315 if (tipc_buf_append(reasm_skb, &skb)) { 1316 l->stats.recv_fragmented++; 1317 tipc_data_input(l, skb, inputq); 1318 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) { 1319 pr_warn_ratelimited("Unable to build fragment list\n"); 1320 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT); 1321 } 1322 return 0; 1323 } else if (usr == BCAST_PROTOCOL) { 1324 tipc_bcast_lock(l->net); 1325 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr); 1326 tipc_bcast_unlock(l->net); 1327 } 1328 1329 kfree_skb(skb); 1330 return 0; 1331} 1332 1333/* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the 1334 * inner message along with the ones in the old link's 1335 * deferdq 1336 * @l: tunnel link 1337 * @skb: TUNNEL_PROTOCOL message 1338 * @inputq: queue to put messages ready for delivery 1339 */ 1340static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb, 1341 struct sk_buff_head *inputq) 1342{ 1343 struct sk_buff **reasm_skb = &l->failover_reasm_skb; 1344 struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg; 1345 struct sk_buff_head *fdefq = &l->failover_deferdq; 1346 struct tipc_msg *hdr = buf_msg(skb); 1347 struct sk_buff *iskb; 1348 int ipos = 0; 1349 int rc = 0; 1350 u16 seqno; 1351 1352 if (msg_type(hdr) == SYNCH_MSG) { 1353 kfree_skb(skb); 1354 return 0; 1355 } 1356 1357 /* Not a fragment? */ 1358 if (likely(!msg_nof_fragms(hdr))) { 1359 if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) { 1360 pr_warn_ratelimited("Unable to extract msg, defq: %d\n", 1361 skb_queue_len(fdefq)); 1362 return 0; 1363 } 1364 kfree_skb(skb); 1365 } else { 1366 /* Set fragment type for buf_append */ 1367 if (msg_fragm_no(hdr) == 1) 1368 msg_set_type(hdr, FIRST_FRAGMENT); 1369 else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr)) 1370 msg_set_type(hdr, FRAGMENT); 1371 else 1372 msg_set_type(hdr, LAST_FRAGMENT); 1373 1374 if (!tipc_buf_append(reasm_tnlmsg, &skb)) { 1375 /* Successful but non-complete reassembly? */ 1376 if (*reasm_tnlmsg || link_is_bc_rcvlink(l)) 1377 return 0; 1378 pr_warn_ratelimited("Unable to reassemble tunnel msg\n"); 1379 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT); 1380 } 1381 iskb = skb; 1382 } 1383 1384 do { 1385 seqno = buf_seqno(iskb); 1386 if (unlikely(less(seqno, l->drop_point))) { 1387 kfree_skb(iskb); 1388 continue; 1389 } 1390 if (unlikely(seqno != l->drop_point)) { 1391 __tipc_skb_queue_sorted(fdefq, seqno, iskb); 1392 continue; 1393 } 1394 1395 l->drop_point++; 1396 if (!tipc_data_input(l, iskb, inputq)) 1397 rc |= tipc_link_input(l, iskb, inputq, reasm_skb); 1398 if (unlikely(rc)) 1399 break; 1400 } while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point))); 1401 1402 return rc; 1403} 1404 1405static int tipc_link_release_pkts(struct tipc_link *l, u16 acked) 1406{ 1407 int released = 0; 1408 struct sk_buff *skb, *tmp; 1409 1410 skb_queue_walk_safe(&l->transmq, skb, tmp) { 1411 if (more(buf_seqno(skb), acked)) 1412 break; 1413 __skb_unlink(skb, &l->transmq); 1414 kfree_skb(skb); 1415 released++; 1416 } 1417 return released; 1418} 1419 1420/* tipc_build_gap_ack_blks - build Gap ACK blocks 1421 * @l: tipc link that data have come with gaps in sequence if any 1422 * @data: data buffer to store the Gap ACK blocks after built 1423 * 1424 * returns the actual allocated memory size 1425 */ 1426static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data, u16 gap) 1427{ 1428 struct sk_buff *skb = skb_peek(&l->deferdq); 1429 struct tipc_gap_ack_blks *ga = data; 1430 u16 len, expect, seqno = 0; 1431 u8 n = 0; 1432 1433 if (!skb || !gap) 1434 goto exit; 1435 1436 expect = buf_seqno(skb); 1437 skb_queue_walk(&l->deferdq, skb) { 1438 seqno = buf_seqno(skb); 1439 if (unlikely(more(seqno, expect))) { 1440 ga->gacks[n].ack = htons(expect - 1); 1441 ga->gacks[n].gap = htons(seqno - expect); 1442 if (++n >= MAX_GAP_ACK_BLKS) { 1443 pr_info_ratelimited("Too few Gap ACK blocks!\n"); 1444 goto exit; 1445 } 1446 } else if (unlikely(less(seqno, expect))) { 1447 pr_warn("Unexpected skb in deferdq!\n"); 1448 continue; 1449 } 1450 expect = seqno + 1; 1451 } 1452 1453 /* last block */ 1454 ga->gacks[n].ack = htons(seqno); 1455 ga->gacks[n].gap = 0; 1456 n++; 1457 1458exit: 1459 len = tipc_gap_ack_blks_sz(n); 1460 ga->len = htons(len); 1461 ga->gack_cnt = n; 1462 return len; 1463} 1464 1465/* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing 1466 * acked packets, also doing retransmissions if 1467 * gaps found 1468 * @l: tipc link with transmq queue to be advanced 1469 * @acked: seqno of last packet acked by peer without any gaps before 1470 * @gap: # of gap packets 1471 * @ga: buffer pointer to Gap ACK blocks from peer 1472 * @xmitq: queue for accumulating the retransmitted packets if any 1473 * 1474 * In case of a repeated retransmit failures, the call will return shortly 1475 * with a returned code (e.g. TIPC_LINK_DOWN_EVT) 1476 */ 1477static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap, 1478 struct tipc_gap_ack_blks *ga, 1479 struct sk_buff_head *xmitq) 1480{ 1481 struct sk_buff *skb, *_skb, *tmp; 1482 struct tipc_msg *hdr; 1483 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1; 1484 bool retransmitted = false; 1485 u16 ack = l->rcv_nxt - 1; 1486 bool passed = false; 1487 u16 released = 0; 1488 u16 seqno, n = 0; 1489 int rc = 0; 1490 1491 skb_queue_walk_safe(&l->transmq, skb, tmp) { 1492 seqno = buf_seqno(skb); 1493 1494next_gap_ack: 1495 if (less_eq(seqno, acked)) { 1496 /* release skb */ 1497 __skb_unlink(skb, &l->transmq); 1498 kfree_skb(skb); 1499 released++; 1500 } else if (less_eq(seqno, acked + gap)) { 1501 /* First, check if repeated retrans failures occurs? */ 1502 if (!passed && link_retransmit_failure(l, l, &rc)) 1503 return rc; 1504 passed = true; 1505 1506 /* retransmit skb if unrestricted*/ 1507 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr)) 1508 continue; 1509 TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME; 1510 _skb = pskb_copy(skb, GFP_ATOMIC); 1511 if (!_skb) 1512 continue; 1513 hdr = buf_msg(_skb); 1514 msg_set_ack(hdr, ack); 1515 msg_set_bcast_ack(hdr, bc_ack); 1516 _skb->priority = TC_PRIO_CONTROL; 1517 __skb_queue_tail(xmitq, _skb); 1518 l->stats.retransmitted++; 1519 retransmitted = true; 1520 /* Increase actual retrans counter & mark first time */ 1521 if (!TIPC_SKB_CB(skb)->retr_cnt++) 1522 TIPC_SKB_CB(skb)->retr_stamp = jiffies; 1523 } else { 1524 /* retry with Gap ACK blocks if any */ 1525 if (!ga || n >= ga->gack_cnt) 1526 break; 1527 acked = ntohs(ga->gacks[n].ack); 1528 gap = ntohs(ga->gacks[n].gap); 1529 n++; 1530 goto next_gap_ack; 1531 } 1532 } 1533 if (released || retransmitted) 1534 tipc_link_update_cwin(l, released, retransmitted); 1535 if (released) 1536 tipc_link_advance_backlog(l, xmitq); 1537 return 0; 1538} 1539 1540/* tipc_link_build_state_msg: prepare link state message for transmission 1541 * 1542 * Note that sending of broadcast ack is coordinated among nodes, to reduce 1543 * risk of ack storms towards the sender 1544 */ 1545int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq) 1546{ 1547 if (!l) 1548 return 0; 1549 1550 /* Broadcast ACK must be sent via a unicast link => defer to caller */ 1551 if (link_is_bc_rcvlink(l)) { 1552 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf) 1553 return 0; 1554 l->rcv_unacked = 0; 1555 1556 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */ 1557 l->snd_nxt = l->rcv_nxt; 1558 return TIPC_LINK_SND_STATE; 1559 } 1560 /* Unicast ACK */ 1561 l->rcv_unacked = 0; 1562 l->stats.sent_acks++; 1563 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq); 1564 return 0; 1565} 1566 1567/* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message 1568 */ 1569void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq) 1570{ 1571 int mtyp = RESET_MSG; 1572 struct sk_buff *skb; 1573 1574 if (l->state == LINK_ESTABLISHING) 1575 mtyp = ACTIVATE_MSG; 1576 1577 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq); 1578 1579 /* Inform peer that this endpoint is going down if applicable */ 1580 skb = skb_peek_tail(xmitq); 1581 if (skb && (l->state == LINK_RESET)) 1582 msg_set_peer_stopping(buf_msg(skb), 1); 1583} 1584 1585/* tipc_link_build_nack_msg: prepare link nack message for transmission 1586 * Note that sending of broadcast NACK is coordinated among nodes, to 1587 * reduce the risk of NACK storms towards the sender 1588 */ 1589static int tipc_link_build_nack_msg(struct tipc_link *l, 1590 struct sk_buff_head *xmitq) 1591{ 1592 u32 def_cnt = ++l->stats.deferred_recv; 1593 struct sk_buff_head *dfq = &l->deferdq; 1594 u32 defq_len = skb_queue_len(dfq); 1595 int match1, match2; 1596 1597 if (link_is_bc_rcvlink(l)) { 1598 match1 = def_cnt & 0xf; 1599 match2 = tipc_own_addr(l->net) & 0xf; 1600 if (match1 == match2) 1601 return TIPC_LINK_SND_STATE; 1602 return 0; 1603 } 1604 1605 if (defq_len >= 3 && !((defq_len - 3) % 16)) { 1606 u16 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt; 1607 1608 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 1609 rcvgap, 0, 0, xmitq); 1610 } 1611 return 0; 1612} 1613 1614/* tipc_link_rcv - process TIPC packets/messages arriving from off-node 1615 * @l: the link that should handle the message 1616 * @skb: TIPC packet 1617 * @xmitq: queue to place packets to be sent after this call 1618 */ 1619int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb, 1620 struct sk_buff_head *xmitq) 1621{ 1622 struct sk_buff_head *defq = &l->deferdq; 1623 struct tipc_msg *hdr = buf_msg(skb); 1624 u16 seqno, rcv_nxt, win_lim; 1625 int released = 0; 1626 int rc = 0; 1627 1628 /* Verify and update link state */ 1629 if (unlikely(msg_user(hdr) == LINK_PROTOCOL)) 1630 return tipc_link_proto_rcv(l, skb, xmitq); 1631 1632 /* Don't send probe at next timeout expiration */ 1633 l->silent_intv_cnt = 0; 1634 1635 do { 1636 hdr = buf_msg(skb); 1637 seqno = msg_seqno(hdr); 1638 rcv_nxt = l->rcv_nxt; 1639 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN; 1640 1641 if (unlikely(!link_is_up(l))) { 1642 if (l->state == LINK_ESTABLISHING) 1643 rc = TIPC_LINK_UP_EVT; 1644 kfree_skb(skb); 1645 break; 1646 } 1647 1648 /* Drop if outside receive window */ 1649 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) { 1650 l->stats.duplicates++; 1651 kfree_skb(skb); 1652 break; 1653 } 1654 released += tipc_link_release_pkts(l, msg_ack(hdr)); 1655 1656 /* Defer delivery if sequence gap */ 1657 if (unlikely(seqno != rcv_nxt)) { 1658 __tipc_skb_queue_sorted(defq, seqno, skb); 1659 rc |= tipc_link_build_nack_msg(l, xmitq); 1660 break; 1661 } 1662 1663 /* Deliver packet */ 1664 l->rcv_nxt++; 1665 l->stats.recv_pkts++; 1666 1667 if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL)) 1668 rc |= tipc_link_tnl_rcv(l, skb, l->inputq); 1669 else if (!tipc_data_input(l, skb, l->inputq)) 1670 rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf); 1671 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN)) 1672 rc |= tipc_link_build_state_msg(l, xmitq); 1673 if (unlikely(rc & ~TIPC_LINK_SND_STATE)) 1674 break; 1675 } while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt))); 1676 1677 /* Forward queues and wake up waiting users */ 1678 if (released) { 1679 tipc_link_update_cwin(l, released, 0); 1680 tipc_link_advance_backlog(l, xmitq); 1681 if (unlikely(!skb_queue_empty(&l->wakeupq))) 1682 link_prepare_wakeup(l); 1683 } 1684 return rc; 1685} 1686 1687static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe, 1688 bool probe_reply, u16 rcvgap, 1689 int tolerance, int priority, 1690 struct sk_buff_head *xmitq) 1691{ 1692 struct tipc_link *bcl = l->bc_rcvlink; 1693 struct sk_buff *skb; 1694 struct tipc_msg *hdr; 1695 struct sk_buff_head *dfq = &l->deferdq; 1696 bool node_up = link_is_up(bcl); 1697 struct tipc_mon_state *mstate = &l->mon_state; 1698 int dlen = 0; 1699 void *data; 1700 u16 glen = 0; 1701 1702 /* Don't send protocol message during reset or link failover */ 1703 if (tipc_link_is_blocked(l)) 1704 return; 1705 1706 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG)) 1707 return; 1708 1709 if ((probe || probe_reply) && !skb_queue_empty(dfq)) 1710 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt; 1711 1712 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE, 1713 tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ, 1714 l->addr, tipc_own_addr(l->net), 0, 0, 0); 1715 if (!skb) 1716 return; 1717 1718 hdr = buf_msg(skb); 1719 data = msg_data(hdr); 1720 msg_set_session(hdr, l->session); 1721 msg_set_bearer_id(hdr, l->bearer_id); 1722 msg_set_net_plane(hdr, l->net_plane); 1723 msg_set_next_sent(hdr, l->snd_nxt); 1724 msg_set_ack(hdr, l->rcv_nxt - 1); 1725 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1); 1726 msg_set_bc_ack_invalid(hdr, !node_up); 1727 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1); 1728 msg_set_link_tolerance(hdr, tolerance); 1729 msg_set_linkprio(hdr, priority); 1730 msg_set_redundant_link(hdr, node_up); 1731 msg_set_seq_gap(hdr, 0); 1732 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2); 1733 1734 if (mtyp == STATE_MSG) { 1735 if (l->peer_caps & TIPC_LINK_PROTO_SEQNO) 1736 msg_set_seqno(hdr, l->snd_nxt_state++); 1737 msg_set_seq_gap(hdr, rcvgap); 1738 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl)); 1739 msg_set_probe(hdr, probe); 1740 msg_set_is_keepalive(hdr, probe || probe_reply); 1741 if (l->peer_caps & TIPC_GAP_ACK_BLOCK) 1742 glen = tipc_build_gap_ack_blks(l, data, rcvgap); 1743 tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id); 1744 msg_set_size(hdr, INT_H_SIZE + glen + dlen); 1745 skb_trim(skb, INT_H_SIZE + glen + dlen); 1746 l->stats.sent_states++; 1747 l->rcv_unacked = 0; 1748 } else { 1749 /* RESET_MSG or ACTIVATE_MSG */ 1750 if (mtyp == ACTIVATE_MSG) { 1751 msg_set_dest_session_valid(hdr, 1); 1752 msg_set_dest_session(hdr, l->peer_session); 1753 } 1754 msg_set_max_pkt(hdr, l->advertised_mtu); 1755 strcpy(data, l->if_name); 1756 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME); 1757 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME); 1758 } 1759 if (probe) 1760 l->stats.sent_probes++; 1761 if (rcvgap) 1762 l->stats.sent_nacks++; 1763 skb->priority = TC_PRIO_CONTROL; 1764 __skb_queue_tail(xmitq, skb); 1765 trace_tipc_proto_build(skb, false, l->name); 1766} 1767 1768void tipc_link_create_dummy_tnl_msg(struct tipc_link *l, 1769 struct sk_buff_head *xmitq) 1770{ 1771 u32 onode = tipc_own_addr(l->net); 1772 struct tipc_msg *hdr, *ihdr; 1773 struct sk_buff_head tnlq; 1774 struct sk_buff *skb; 1775 u32 dnode = l->addr; 1776 1777 __skb_queue_head_init(&tnlq); 1778 skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG, 1779 INT_H_SIZE, BASIC_H_SIZE, 1780 dnode, onode, 0, 0, 0); 1781 if (!skb) { 1782 pr_warn("%sunable to create tunnel packet\n", link_co_err); 1783 return; 1784 } 1785 1786 hdr = buf_msg(skb); 1787 msg_set_msgcnt(hdr, 1); 1788 msg_set_bearer_id(hdr, l->peer_bearer_id); 1789 1790 ihdr = (struct tipc_msg *)msg_data(hdr); 1791 tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG, 1792 BASIC_H_SIZE, dnode); 1793 msg_set_errcode(ihdr, TIPC_ERR_NO_PORT); 1794 __skb_queue_tail(&tnlq, skb); 1795 tipc_link_xmit(l, &tnlq, xmitq); 1796} 1797 1798/* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets 1799 * with contents of the link's transmit and backlog queues. 1800 */ 1801void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl, 1802 int mtyp, struct sk_buff_head *xmitq) 1803{ 1804 struct sk_buff_head *fdefq = &tnl->failover_deferdq; 1805 struct sk_buff *skb, *tnlskb; 1806 struct tipc_msg *hdr, tnlhdr; 1807 struct sk_buff_head *queue = &l->transmq; 1808 struct sk_buff_head tmpxq, tnlq, frags; 1809 u16 pktlen, pktcnt, seqno = l->snd_nxt; 1810 bool pktcnt_need_update = false; 1811 u16 syncpt; 1812 int rc; 1813 1814 if (!tnl) 1815 return; 1816 1817 __skb_queue_head_init(&tnlq); 1818 /* Link Synching: 1819 * From now on, send only one single ("dummy") SYNCH message 1820 * to peer. The SYNCH message does not contain any data, just 1821 * a header conveying the synch point to the peer. 1822 */ 1823 if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) { 1824 tnlskb = tipc_msg_c…
Large files files are truncated, but you can click here to view the full file