net/rxrpc/conn_client.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /* Client connection-specific management code.
   3  *
   4  * Copyright (C) 2016, 2020 Red Hat, Inc. All Rights Reserved.
   5  * Written by David Howells (dhowells@redhat.com)
   6  *
   7  * Client connections need to be cached for a little while after they've made a
   8  * call so as to handle retransmitted DATA packets in case the server didn't
   9  * receive the final ACK or terminating ABORT we sent it.
  10  *
  11  * There are flags of relevance to the cache:
  12  *
  13  *  (2) DONT_REUSE - The connection should be discarded as soon as possible and
  14  *      should not be reused.  This is set when an exclusive connection is used
  15  *      or a call ID counter overflows.
  16  *
  17  * The caching state may only be changed if the cache lock is held.
  18  *
  19  * There are two idle client connection expiry durations.  If the total number
  20  * of connections is below the reap threshold, we use the normal duration; if
  21  * it's above, we use the fast duration.
  22  */
  23
  24 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  25
  26 #include <linux/slab.h>
  27 #include <linux/idr.h>
  28 #include <linux/timer.h>
  29 #include <linux/sched/signal.h>
  30
  31 #include "ar-internal.h"
  32
  33 __read_mostly unsigned int rxrpc_reap_client_connections = 900;
  34 __read_mostly unsigned long rxrpc_conn_idle_client_expiry = 2 * 60 * HZ;
  35 __read_mostly unsigned long rxrpc_conn_idle_client_fast_expiry = 2 * HZ;
  36
  37 /*
  38  * We use machine-unique IDs for our client connections.
  39  */
  40 DEFINE_IDR(rxrpc_client_conn_ids);
  41 static DEFINE_SPINLOCK(rxrpc_conn_id_lock);
  42
  43 static void rxrpc_deactivate_bundle(struct rxrpc_bundle *bundle);
  44
  45 /*
  46  * Get a connection ID and epoch for a client connection from the global pool.
  47  * The connection struct pointer is then recorded in the idr radix tree.  The
  48  * epoch doesn't change until the client is rebooted (or, at least, unless the
  49  * module is unloaded).
  50  */
  51 static int rxrpc_get_client_connection_id(struct rxrpc_connection *conn,
  52                                           gfp_t gfp)
  53 {
  54         struct rxrpc_net *rxnet = conn->params.local->rxnet;
  55         int id;
  56
  57         _enter("");
  58
  59         idr_preload(gfp);
  60         spin_lock(&rxrpc_conn_id_lock);
  61
  62         id = idr_alloc_cyclic(&rxrpc_client_conn_ids, conn,
  63                               1, 0x40000000, GFP_NOWAIT);
  64         if (id < 0)
  65                 goto error;
  66
  67         spin_unlock(&rxrpc_conn_id_lock);
  68         idr_preload_end();
  69
  70         conn->proto.epoch = rxnet->epoch;
  71         conn->proto.cid = id << RXRPC_CIDSHIFT;
  72         set_bit(RXRPC_CONN_HAS_IDR, &conn->flags);
  73         _leave(" [CID %x]", conn->proto.cid);
  74         return 0;
  75
  76 error:
  77         spin_unlock(&rxrpc_conn_id_lock);
  78         idr_preload_end();
  79         _leave(" = %d", id);
  80         return id;
  81 }
  82
  83 /*
  84  * Release a connection ID for a client connection from the global pool.
  85  */
  86 static void rxrpc_put_client_connection_id(struct rxrpc_connection *conn)
  87 {
  88         if (test_bit(RXRPC_CONN_HAS_IDR, &conn->flags)) {
  89                 spin_lock(&rxrpc_conn_id_lock);
  90                 idr_remove(&rxrpc_client_conn_ids,
  91                            conn->proto.cid >> RXRPC_CIDSHIFT);
  92                 spin_unlock(&rxrpc_conn_id_lock);
  93         }
  94 }
  95
  96 /*
  97  * Destroy the client connection ID tree.
  98  */
  99 void rxrpc_destroy_client_conn_ids(void)
 100 {
 101         struct rxrpc_connection *conn;
 102         int id;
 103
 104         if (!idr_is_empty(&rxrpc_client_conn_ids)) {
 105                 idr_for_each_entry(&rxrpc_client_conn_ids, conn, id) {
 106                         pr_err("AF_RXRPC: Leaked client conn %p {%d}\n",
 107                                conn, refcount_read(&conn->ref));
 108                 }
 109                 BUG();
 110         }
 111
 112         idr_destroy(&rxrpc_client_conn_ids);
 113 }
 114
 115 /*
 116  * Allocate a connection bundle.
 117  */
 118 static struct rxrpc_bundle *rxrpc_alloc_bundle(struct rxrpc_conn_parameters *cp,
 119                                                gfp_t gfp)
 120 {
 121         struct rxrpc_bundle *bundle;
 122
 123         bundle = kzalloc(sizeof(*bundle), gfp);
 124         if (bundle) {
 125                 bundle->params = *cp;
 126                 rxrpc_get_peer(bundle->params.peer);
 127                 refcount_set(&bundle->ref, 1);
 128                 atomic_set(&bundle->active, 1);
 129                 spin_lock_init(&bundle->channel_lock);
 130                 INIT_LIST_HEAD(&bundle->waiting_calls);
 131         }
 132         return bundle;
 133 }
 134
 135 struct rxrpc_bundle *rxrpc_get_bundle(struct rxrpc_bundle *bundle)
 136 {
 137         refcount_inc(&bundle->ref);
 138         return bundle;
 139 }
 140
 141 static void rxrpc_free_bundle(struct rxrpc_bundle *bundle)
 142 {
 143         rxrpc_put_peer(bundle->params.peer);
 144         kfree(bundle);
 145 }
 146
 147 void rxrpc_put_bundle(struct rxrpc_bundle *bundle)
 148 {
 149         unsigned int d = bundle->debug_id;
 150         bool dead;
 151         int r;
 152
 153         dead = __refcount_dec_and_test(&bundle->ref, &r);
 154
 155         _debug("PUT B=%x %d", d, r - 1);
 156         if (dead)
 157                 rxrpc_free_bundle(bundle);
 158 }
 159
 160 /*
 161  * Allocate a client connection.
 162  */
 163 static struct rxrpc_connection *
 164 rxrpc_alloc_client_connection(struct rxrpc_bundle *bundle, gfp_t gfp)
 165 {
 166         struct rxrpc_connection *conn;
 167         struct rxrpc_net *rxnet = bundle->params.local->rxnet;
 168         int ret;
 169
 170         _enter("");
 171
 172         conn = rxrpc_alloc_connection(gfp);
 173         if (!conn) {
 174                 _leave(" = -ENOMEM");
 175                 return ERR_PTR(-ENOMEM);
 176         }
 177
 178         refcount_set(&conn->ref, 1);
 179         conn->bundle            = bundle;
 180         conn->params            = bundle->params;
 181         conn->out_clientflag    = RXRPC_CLIENT_INITIATED;
 182         conn->state             = RXRPC_CONN_CLIENT;
 183         conn->service_id        = conn->params.service_id;
 184
 185         ret = rxrpc_get_client_connection_id(conn, gfp);
 186         if (ret < 0)
 187                 goto error_0;
 188
 189         ret = rxrpc_init_client_conn_security(conn);
 190         if (ret < 0)
 191                 goto error_1;
 192
 193         ret = conn->security->prime_packet_security(conn);
 194         if (ret < 0)
 195                 goto error_2;
 196
 197         atomic_inc(&rxnet->nr_conns);
 198         write_lock(&rxnet->conn_lock);
 199         list_add_tail(&conn->proc_link, &rxnet->conn_proc_list);
 200         write_unlock(&rxnet->conn_lock);
 201
 202         rxrpc_get_bundle(bundle);
 203         rxrpc_get_peer(conn->params.peer);
 204         rxrpc_get_local(conn->params.local);
 205         key_get(conn->params.key);
 206
 207         trace_rxrpc_conn(conn->debug_id, rxrpc_conn_new_client,
 208                          refcount_read(&conn->ref),
 209                          __builtin_return_address(0));
 210
 211         atomic_inc(&rxnet->nr_client_conns);
 212         trace_rxrpc_client(conn, -1, rxrpc_client_alloc);
 213         _leave(" = %p", conn);
 214         return conn;
 215
 216 error_2:
 217         conn->security->clear(conn);
 218 error_1:
 219         rxrpc_put_client_connection_id(conn);
 220 error_0:
 221         kfree(conn);
 222         _leave(" = %d", ret);
 223         return ERR_PTR(ret);
 224 }
 225
 226 /*
 227  * Determine if a connection may be reused.
 228  */
 229 static bool rxrpc_may_reuse_conn(struct rxrpc_connection *conn)
 230 {
 231         struct rxrpc_net *rxnet;
 232         int id_cursor, id, distance, limit;
 233
 234         if (!conn)
 235                 goto dont_reuse;
 236
 237         rxnet = conn->params.local->rxnet;
 238         if (test_bit(RXRPC_CONN_DONT_REUSE, &conn->flags))
 239                 goto dont_reuse;
 240
 241         if (conn->state != RXRPC_CONN_CLIENT ||
 242             conn->proto.epoch != rxnet->epoch)
 243                 goto mark_dont_reuse;
 244
 245         /* The IDR tree gets very expensive on memory if the connection IDs are
 246          * widely scattered throughout the number space, so we shall want to
 247          * kill off connections that, say, have an ID more than about four
 248          * times the maximum number of client conns away from the current
 249          * allocation point to try and keep the IDs concentrated.
 250          */
 251         id_cursor = idr_get_cursor(&rxrpc_client_conn_ids);
 252         id = conn->proto.cid >> RXRPC_CIDSHIFT;
 253         distance = id - id_cursor;
 254         if (distance < 0)
 255                 distance = -distance;
 256         limit = max_t(unsigned long, atomic_read(&rxnet->nr_conns) * 4, 1024);
 257         if (distance > limit)
 258                 goto mark_dont_reuse;
 259
 260         return true;
 261
 262 mark_dont_reuse:
 263         set_bit(RXRPC_CONN_DONT_REUSE, &conn->flags);
 264 dont_reuse:
 265         return false;
 266 }
 267
 268 /*
 269  * Look up the conn bundle that matches the connection parameters, adding it if
 270  * it doesn't yet exist.
 271  */
 272 static struct rxrpc_bundle *rxrpc_look_up_bundle(struct rxrpc_conn_parameters *cp,
 273                                                  gfp_t gfp)
 274 {
 275         static atomic_t rxrpc_bundle_id;
 276         struct rxrpc_bundle *bundle, *candidate;
 277         struct rxrpc_local *local = cp->local;
 278         struct rb_node *p, **pp, *parent;
 279         long diff;
 280
 281         _enter("{%px,%x,%u,%u}",
 282                cp->peer, key_serial(cp->key), cp->security_level, cp->upgrade);
 283
 284         if (cp->exclusive)
 285                 return rxrpc_alloc_bundle(cp, gfp);
 286
 287         /* First, see if the bundle is already there. */
 288         _debug("search 1");
 289         spin_lock(&local->client_bundles_lock);
 290         p = local->client_bundles.rb_node;
 291         while (p) {
 292                 bundle = rb_entry(p, struct rxrpc_bundle, local_node);
 293
 294 #define cmp(X) ((long)bundle->params.X - (long)cp->X)
 295                 diff = (cmp(peer) ?:
 296                         cmp(key) ?:
 297                         cmp(security_level) ?:
 298                         cmp(upgrade));
 299 #undef cmp
 300                 if (diff < 0)
 301                         p = p->rb_left;
 302                 else if (diff > 0)
 303                         p = p->rb_right;
 304                 else
 305                         goto found_bundle;
 306         }
 307         spin_unlock(&local->client_bundles_lock);
 308         _debug("not found");
 309
 310         /* It wasn't.  We need to add one. */
 311         candidate = rxrpc_alloc_bundle(cp, gfp);
 312         if (!candidate)
 313                 return NULL;
 314
 315         _debug("search 2");
 316         spin_lock(&local->client_bundles_lock);
 317         pp = &local->client_bundles.rb_node;
 318         parent = NULL;
 319         while (*pp) {
 320                 parent = *pp;
 321                 bundle = rb_entry(parent, struct rxrpc_bundle, local_node);
 322
 323 #define cmp(X) ((long)bundle->params.X - (long)cp->X)
 324                 diff = (cmp(peer) ?:
 325                         cmp(key) ?:
 326                         cmp(security_level) ?:
 327                         cmp(upgrade));
 328 #undef cmp
 329                 if (diff < 0)
 330                         pp = &(*pp)->rb_left;
 331                 else if (diff > 0)
 332                         pp = &(*pp)->rb_right;
 333                 else
 334                         goto found_bundle_free;
 335         }
 336
 337         _debug("new bundle");
 338         candidate->debug_id = atomic_inc_return(&rxrpc_bundle_id);
 339         rb_link_node(&candidate->local_node, parent, pp);
 340         rb_insert_color(&candidate->local_node, &local->client_bundles);
 341         rxrpc_get_bundle(candidate);
 342         spin_unlock(&local->client_bundles_lock);
 343         _leave(" = %u [new]", candidate->debug_id);
 344         return candidate;
 345
 346 found_bundle_free:
 347         rxrpc_free_bundle(candidate);
 348 found_bundle:
 349         rxrpc_get_bundle(bundle);
 350         atomic_inc(&bundle->active);
 351         spin_unlock(&local->client_bundles_lock);
 352         _leave(" = %u [found]", bundle->debug_id);
 353         return bundle;
 354 }
 355
 356 /*
 357  * Create or find a client bundle to use for a call.
 358  *
 359  * If we return with a connection, the call will be on its waiting list.  It's
 360  * left to the caller to assign a channel and wake up the call.
 361  */
 362 static struct rxrpc_bundle *rxrpc_prep_call(struct rxrpc_sock *rx,
 363                                             struct rxrpc_call *call,
 364                                             struct rxrpc_conn_parameters *cp,
 365                                             struct sockaddr_rxrpc *srx,
 366                                             gfp_t gfp)
 367 {
 368         struct rxrpc_bundle *bundle;
 369
 370         _enter("{%d,%lx},", call->debug_id, call->user_call_ID);
 371
 372         cp->peer = rxrpc_lookup_peer(rx, cp->local, srx, gfp);
 373         if (!cp->peer)
 374                 goto error;
 375
 376         call->cong_cwnd = cp->peer->cong_cwnd;
 377         if (call->cong_cwnd >= call->cong_ssthresh)
 378                 call->cong_mode = RXRPC_CALL_CONGEST_AVOIDANCE;
 379         else
 380                 call->cong_mode = RXRPC_CALL_SLOW_START;
 381         if (cp->upgrade)
 382                 __set_bit(RXRPC_CALL_UPGRADE, &call->flags);
 383
 384         /* Find the client connection bundle. */
 385         bundle = rxrpc_look_up_bundle(cp, gfp);
 386         if (!bundle)
 387                 goto error;
 388
 389         /* Get this call queued.  Someone else may activate it whilst we're
 390          * lining up a new connection, but that's fine.
 391          */
 392         spin_lock(&bundle->channel_lock);
 393         list_add_tail(&call->chan_wait_link, &bundle->waiting_calls);
 394         spin_unlock(&bundle->channel_lock);
 395
 396         _leave(" = [B=%x]", bundle->debug_id);
 397         return bundle;
 398
 399 error:
 400         _leave(" = -ENOMEM");
 401         return ERR_PTR(-ENOMEM);
 402 }
 403
 404 /*
 405  * Allocate a new connection and add it into a bundle.
 406  */
 407 static void rxrpc_add_conn_to_bundle(struct rxrpc_bundle *bundle, gfp_t gfp)
 408         __releases(bundle->channel_lock)
 409 {
 410         struct rxrpc_connection *candidate = NULL, *old = NULL;
 411         bool conflict;
 412         int i;
 413
 414         _enter("");
 415
 416         conflict = bundle->alloc_conn;
 417         if (!conflict)
 418                 bundle->alloc_conn = true;
 419         spin_unlock(&bundle->channel_lock);
 420         if (conflict) {
 421                 _leave(" [conf]");
 422                 return;
 423         }
 424
 425         candidate = rxrpc_alloc_client_connection(bundle, gfp);
 426
 427         spin_lock(&bundle->channel_lock);
 428         bundle->alloc_conn = false;
 429
 430         if (IS_ERR(candidate)) {
 431                 bundle->alloc_error = PTR_ERR(candidate);
 432                 spin_unlock(&bundle->channel_lock);
 433                 _leave(" [err %ld]", PTR_ERR(candidate));
 434                 return;
 435         }
 436
 437         bundle->alloc_error = 0;
 438
 439         for (i = 0; i < ARRAY_SIZE(bundle->conns); i++) {
 440                 unsigned int shift = i * RXRPC_MAXCALLS;
 441                 int j;
 442
 443                 old = bundle->conns[i];
 444                 if (!rxrpc_may_reuse_conn(old)) {
 445                         if (old)
 446                                 trace_rxrpc_client(old, -1, rxrpc_client_replace);
 447                         candidate->bundle_shift = shift;
 448                         atomic_inc(&bundle->active);
 449                         bundle->conns[i] = candidate;
 450                         for (j = 0; j < RXRPC_MAXCALLS; j++)
 451                                 set_bit(shift + j, &bundle->avail_chans);
 452                         candidate = NULL;
 453                         break;
 454                 }
 455
 456                 old = NULL;
 457         }
 458
 459         spin_unlock(&bundle->channel_lock);
 460
 461         if (candidate) {
 462                 _debug("discard C=%x", candidate->debug_id);
 463                 trace_rxrpc_client(candidate, -1, rxrpc_client_duplicate);
 464                 rxrpc_put_connection(candidate);
 465         }
 466
 467         rxrpc_put_connection(old);
 468         _leave("");
 469 }
 470
 471 /*
 472  * Add a connection to a bundle if there are no usable connections or we have
 473  * connections waiting for extra capacity.
 474  */
 475 static void rxrpc_maybe_add_conn(struct rxrpc_bundle *bundle, gfp_t gfp)
 476 {
 477         struct rxrpc_call *call;
 478         int i, usable;
 479
 480         _enter("");
 481
 482         spin_lock(&bundle->channel_lock);
 483
 484         /* See if there are any usable connections. */
 485         usable = 0;
 486         for (i = 0; i < ARRAY_SIZE(bundle->conns); i++)
 487                 if (rxrpc_may_reuse_conn(bundle->conns[i]))
 488                         usable++;
 489
 490         if (!usable && !list_empty(&bundle->waiting_calls)) {
 491                 call = list_first_entry(&bundle->waiting_calls,
 492                                         struct rxrpc_call, chan_wait_link);
 493                 if (test_bit(RXRPC_CALL_UPGRADE, &call->flags))
 494                         bundle->try_upgrade = true;
 495         }
 496
 497         if (!usable)
 498                 goto alloc_conn;
 499
 500         if (!bundle->avail_chans &&
 501             !bundle->try_upgrade &&
 502             !list_empty(&bundle->waiting_calls) &&
 503             usable < ARRAY_SIZE(bundle->conns))
 504                 goto alloc_conn;
 505
 506         spin_unlock(&bundle->channel_lock);
 507         _leave("");
 508         return;
 509
 510 alloc_conn:
 511         return rxrpc_add_conn_to_bundle(bundle, gfp);
 512 }
 513
 514 /*
 515  * Assign a channel to the call at the front of the queue and wake the call up.
 516  * We don't increment the callNumber counter until this number has been exposed
 517  * to the world.
 518  */
 519 static void rxrpc_activate_one_channel(struct rxrpc_connection *conn,
 520                                        unsigned int channel)
 521 {
 522         struct rxrpc_channel *chan = &conn->channels[channel];
 523         struct rxrpc_bundle *bundle = conn->bundle;
 524         struct rxrpc_call *call = list_entry(bundle->waiting_calls.next,
 525                                              struct rxrpc_call, chan_wait_link);
 526         u32 call_id = chan->call_counter + 1;
 527
 528         _enter("C=%x,%u", conn->debug_id, channel);
 529
 530         trace_rxrpc_client(conn, channel, rxrpc_client_chan_activate);
 531
 532         /* Cancel the final ACK on the previous call if it hasn't been sent yet
 533          * as the DATA packet will implicitly ACK it.
 534          */
 535         clear_bit(RXRPC_CONN_FINAL_ACK_0 + channel, &conn->flags);
 536         clear_bit(conn->bundle_shift + channel, &bundle->avail_chans);
 537
 538         rxrpc_see_call(call);
 539         list_del_init(&call->chan_wait_link);
 540         call->peer      = rxrpc_get_peer(conn->params.peer);
 541         call->conn      = rxrpc_get_connection(conn);
 542         call->cid       = conn->proto.cid | channel;
 543         call->call_id   = call_id;
 544         call->security  = conn->security;
 545         call->security_ix = conn->security_ix;
 546         call->service_id = conn->service_id;
 547
 548         trace_rxrpc_connect_call(call);
 549         _net("CONNECT call %08x:%08x as call %d on conn %d",
 550              call->cid, call->call_id, call->debug_id, conn->debug_id);
 551
 552         write_lock_bh(&call->state_lock);
 553         call->state = RXRPC_CALL_CLIENT_SEND_REQUEST;
 554         write_unlock_bh(&call->state_lock);
 555
 556         /* Paired with the read barrier in rxrpc_connect_call().  This orders
 557          * cid and epoch in the connection wrt to call_id without the need to
 558          * take the channel_lock.
 559          *
 560          * We provisionally assign a callNumber at this point, but we don't
 561          * confirm it until the call is about to be exposed.
 562          *
 563          * TODO: Pair with a barrier in the data_ready handler when that looks
 564          * at the call ID through a connection channel.
 565          */
 566         smp_wmb();
 567
 568         chan->call_id           = call_id;
 569         chan->call_debug_id     = call->debug_id;
 570         rcu_assign_pointer(chan->call, call);
 571         wake_up(&call->waitq);
 572 }
 573
 574 /*
 575  * Remove a connection from the idle list if it's on it.
 576  */
 577 static void rxrpc_unidle_conn(struct rxrpc_bundle *bundle, struct rxrpc_connection *conn)
 578 {
 579         struct rxrpc_net *rxnet = bundle->params.local->rxnet;
 580         bool drop_ref;
 581
 582         if (!list_empty(&conn->cache_link)) {
 583                 drop_ref = false;
 584                 spin_lock(&rxnet->client_conn_cache_lock);
 585                 if (!list_empty(&conn->cache_link)) {
 586                         list_del_init(&conn->cache_link);
 587                         drop_ref = true;
 588                 }
 589                 spin_unlock(&rxnet->client_conn_cache_lock);
 590                 if (drop_ref)
 591                         rxrpc_put_connection(conn);
 592         }
 593 }
 594
 595 /*
 596  * Assign channels and callNumbers to waiting calls with channel_lock
 597  * held by caller.
 598  */
 599 static void rxrpc_activate_channels_locked(struct rxrpc_bundle *bundle)
 600 {
 601         struct rxrpc_connection *conn;
 602         unsigned long avail, mask;
 603         unsigned int channel, slot;
 604
 605         if (bundle->try_upgrade)
 606                 mask = 1;
 607         else
 608                 mask = ULONG_MAX;
 609
 610         while (!list_empty(&bundle->waiting_calls)) {
 611                 avail = bundle->avail_chans & mask;
 612                 if (!avail)
 613                         break;
 614                 channel = __ffs(avail);
 615                 clear_bit(channel, &bundle->avail_chans);
 616
 617                 slot = channel / RXRPC_MAXCALLS;
 618                 conn = bundle->conns[slot];
 619                 if (!conn)
 620                         break;
 621
 622                 if (bundle->try_upgrade)
 623                         set_bit(RXRPC_CONN_PROBING_FOR_UPGRADE, &conn->flags);
 624                 rxrpc_unidle_conn(bundle, conn);
 625
 626                 channel &= (RXRPC_MAXCALLS - 1);
 627                 conn->act_chans |= 1 << channel;
 628                 rxrpc_activate_one_channel(conn, channel);
 629         }
 630 }
 631
 632 /*
 633  * Assign channels and callNumbers to waiting calls.
 634  */
 635 static void rxrpc_activate_channels(struct rxrpc_bundle *bundle)
 636 {
 637         _enter("B=%x", bundle->debug_id);
 638
 639         trace_rxrpc_client(NULL, -1, rxrpc_client_activate_chans);
 640
 641         if (!bundle->avail_chans)
 642                 return;
 643
 644         spin_lock(&bundle->channel_lock);
 645         rxrpc_activate_channels_locked(bundle);
 646         spin_unlock(&bundle->channel_lock);
 647         _leave("");
 648 }
 649
 650 /*
 651  * Wait for a callNumber and a channel to be granted to a call.
 652  */
 653 static int rxrpc_wait_for_channel(struct rxrpc_bundle *bundle,
 654                                   struct rxrpc_call *call, gfp_t gfp)
 655 {
 656         DECLARE_WAITQUEUE(myself, current);
 657         int ret = 0;
 658
 659         _enter("%d", call->debug_id);
 660
 661         if (!gfpflags_allow_blocking(gfp)) {
 662                 rxrpc_maybe_add_conn(bundle, gfp);
 663                 rxrpc_activate_channels(bundle);
 664                 ret = bundle->alloc_error ?: -EAGAIN;
 665                 goto out;
 666         }
 667
 668         add_wait_queue_exclusive(&call->waitq, &myself);
 669         for (;;) {
 670                 rxrpc_maybe_add_conn(bundle, gfp);
 671                 rxrpc_activate_channels(bundle);
 672                 ret = bundle->alloc_error;
 673                 if (ret < 0)
 674                         break;
 675
 676                 switch (call->interruptibility) {
 677                 case RXRPC_INTERRUPTIBLE:
 678                 case RXRPC_PREINTERRUPTIBLE:
 679                         set_current_state(TASK_INTERRUPTIBLE);
 680                         break;
 681                 case RXRPC_UNINTERRUPTIBLE:
 682                 default:
 683                         set_current_state(TASK_UNINTERRUPTIBLE);
 684                         break;
 685                 }
 686                 if (READ_ONCE(call->state) != RXRPC_CALL_CLIENT_AWAIT_CONN)
 687                         break;
 688                 if ((call->interruptibility == RXRPC_INTERRUPTIBLE ||
 689                      call->interruptibility == RXRPC_PREINTERRUPTIBLE) &&
 690                     signal_pending(current)) {
 691                         ret = -ERESTARTSYS;
 692                         break;
 693                 }
 694                 schedule();
 695         }
 696         remove_wait_queue(&call->waitq, &myself);
 697         __set_current_state(TASK_RUNNING);
 698
 699 out:
 700         _leave(" = %d", ret);
 701         return ret;
 702 }
 703
 704 /*
 705  * find a connection for a call
 706  * - called in process context with IRQs enabled
 707  */
 708 int rxrpc_connect_call(struct rxrpc_sock *rx,
 709                        struct rxrpc_call *call,
 710                        struct rxrpc_conn_parameters *cp,
 711                        struct sockaddr_rxrpc *srx,
 712                        gfp_t gfp)
 713 {
 714         struct rxrpc_bundle *bundle;
 715         struct rxrpc_net *rxnet = cp->local->rxnet;
 716         int ret = 0;
 717
 718         _enter("{%d,%lx},", call->debug_id, call->user_call_ID);
 719
 720         rxrpc_discard_expired_client_conns(&rxnet->client_conn_reaper);
 721
 722         bundle = rxrpc_prep_call(rx, call, cp, srx, gfp);
 723         if (IS_ERR(bundle)) {
 724                 ret = PTR_ERR(bundle);
 725                 goto out;
 726         }
 727
 728         if (call->state == RXRPC_CALL_CLIENT_AWAIT_CONN) {
 729                 ret = rxrpc_wait_for_channel(bundle, call, gfp);
 730                 if (ret < 0)
 731                         goto wait_failed;
 732         }
 733
 734 granted_channel:
 735         /* Paired with the write barrier in rxrpc_activate_one_channel(). */
 736         smp_rmb();
 737
 738 out_put_bundle:
 739         rxrpc_deactivate_bundle(bundle);
 740         rxrpc_put_bundle(bundle);
 741 out:
 742         _leave(" = %d", ret);
 743         return ret;
 744
 745 wait_failed:
 746         spin_lock(&bundle->channel_lock);
 747         list_del_init(&call->chan_wait_link);
 748         spin_unlock(&bundle->channel_lock);
 749
 750         if (call->state != RXRPC_CALL_CLIENT_AWAIT_CONN) {
 751                 ret = 0;
 752                 goto granted_channel;
 753         }
 754
 755         trace_rxrpc_client(call->conn, ret, rxrpc_client_chan_wait_failed);
 756         rxrpc_set_call_completion(call, RXRPC_CALL_LOCAL_ERROR, 0, ret);
 757         rxrpc_disconnect_client_call(bundle, call);
 758         goto out_put_bundle;
 759 }
 760
 761 /*
 762  * Note that a call, and thus a connection, is about to be exposed to the
 763  * world.
 764  */
 765 void rxrpc_expose_client_call(struct rxrpc_call *call)
 766 {
 767         unsigned int channel = call->cid & RXRPC_CHANNELMASK;
 768         struct rxrpc_connection *conn = call->conn;
 769         struct rxrpc_channel *chan = &conn->channels[channel];
 770
 771         if (!test_and_set_bit(RXRPC_CALL_EXPOSED, &call->flags)) {
 772                 /* Mark the call ID as being used.  If the callNumber counter
 773                  * exceeds ~2 billion, we kill the connection after its
 774                  * outstanding calls have finished so that the counter doesn't
 775                  * wrap.
 776                  */
 777                 chan->call_counter++;
 778                 if (chan->call_counter >= INT_MAX)
 779                         set_bit(RXRPC_CONN_DONT_REUSE, &conn->flags);
 780                 trace_rxrpc_client(conn, channel, rxrpc_client_exposed);
 781         }
 782 }
 783
 784 /*
 785  * Set the reap timer.
 786  */
 787 static void rxrpc_set_client_reap_timer(struct rxrpc_net *rxnet)
 788 {
 789         if (!rxnet->kill_all_client_conns) {
 790                 unsigned long now = jiffies;
 791                 unsigned long reap_at = now + rxrpc_conn_idle_client_expiry;
 792
 793                 if (rxnet->live)
 794                         timer_reduce(&rxnet->client_conn_reap_timer, reap_at);
 795         }
 796 }
 797
 798 /*
 799  * Disconnect a client call.
 800  */
 801 void rxrpc_disconnect_client_call(struct rxrpc_bundle *bundle, struct rxrpc_call *call)
 802 {
 803         struct rxrpc_connection *conn;
 804         struct rxrpc_channel *chan = NULL;
 805         struct rxrpc_net *rxnet = bundle->params.local->rxnet;
 806         unsigned int channel;
 807         bool may_reuse;
 808         u32 cid;
 809
 810         _enter("c=%x", call->debug_id);
 811
 812         spin_lock(&bundle->channel_lock);
 813         set_bit(RXRPC_CALL_DISCONNECTED, &call->flags);
 814
 815         /* Calls that have never actually been assigned a channel can simply be
 816          * discarded.
 817          */
 818         conn = call->conn;
 819         if (!conn) {
 820                 _debug("call is waiting");
 821                 ASSERTCMP(call->call_id, ==, 0);
 822                 ASSERT(!test_bit(RXRPC_CALL_EXPOSED, &call->flags));
 823                 list_del_init(&call->chan_wait_link);
 824                 goto out;
 825         }
 826
 827         cid = call->cid;
 828         channel = cid & RXRPC_CHANNELMASK;
 829         chan = &conn->channels[channel];
 830         trace_rxrpc_client(conn, channel, rxrpc_client_chan_disconnect);
 831
 832         if (rcu_access_pointer(chan->call) != call) {
 833                 spin_unlock(&bundle->channel_lock);
 834                 BUG();
 835         }
 836
 837         may_reuse = rxrpc_may_reuse_conn(conn);
 838
 839         /* If a client call was exposed to the world, we save the result for
 840          * retransmission.
 841          *
 842          * We use a barrier here so that the call number and abort code can be
 843          * read without needing to take a lock.
 844          *
 845          * TODO: Make the incoming packet handler check this and handle
 846          * terminal retransmission without requiring access to the call.
 847          */
 848         if (test_bit(RXRPC_CALL_EXPOSED, &call->flags)) {
 849                 _debug("exposed %u,%u", call->call_id, call->abort_code);
 850                 __rxrpc_disconnect_call(conn, call);
 851
 852                 if (test_and_clear_bit(RXRPC_CONN_PROBING_FOR_UPGRADE, &conn->flags)) {
 853                         trace_rxrpc_client(conn, channel, rxrpc_client_to_active);
 854                         bundle->try_upgrade = false;
 855                         if (may_reuse)
 856                                 rxrpc_activate_channels_locked(bundle);
 857                 }
 858
 859         }
 860
 861         /* See if we can pass the channel directly to another call. */
 862         if (may_reuse && !list_empty(&bundle->waiting_calls)) {
 863                 trace_rxrpc_client(conn, channel, rxrpc_client_chan_pass);
 864                 rxrpc_activate_one_channel(conn, channel);
 865                 goto out;
 866         }
 867
 868         /* Schedule the final ACK to be transmitted in a short while so that it
 869          * can be skipped if we find a follow-on call.  The first DATA packet
 870          * of the follow on call will implicitly ACK this call.
 871          */
 872         if (call->completion == RXRPC_CALL_SUCCEEDED &&
 873             test_bit(RXRPC_CALL_EXPOSED, &call->flags)) {
 874                 unsigned long final_ack_at = jiffies + 2;
 875
 876                 WRITE_ONCE(chan->final_ack_at, final_ack_at);
 877                 smp_wmb(); /* vs rxrpc_process_delayed_final_acks() */
 878                 set_bit(RXRPC_CONN_FINAL_ACK_0 + channel, &conn->flags);
 879                 rxrpc_reduce_conn_timer(conn, final_ack_at);
 880         }
 881
 882         /* Deactivate the channel. */
 883         rcu_assign_pointer(chan->call, NULL);
 884         set_bit(conn->bundle_shift + channel, &conn->bundle->avail_chans);
 885         conn->act_chans &= ~(1 << channel);
 886
 887         /* If no channels remain active, then put the connection on the idle
 888          * list for a short while.  Give it a ref to stop it going away if it
 889          * becomes unbundled.
 890          */
 891         if (!conn->act_chans) {
 892                 trace_rxrpc_client(conn, channel, rxrpc_client_to_idle);
 893                 conn->idle_timestamp = jiffies;
 894
 895                 rxrpc_get_connection(conn);
 896                 spin_lock(&rxnet->client_conn_cache_lock);
 897                 list_move_tail(&conn->cache_link, &rxnet->idle_client_conns);
 898                 spin_unlock(&rxnet->client_conn_cache_lock);
 899
 900                 rxrpc_set_client_reap_timer(rxnet);
 901         }
 902
 903 out:
 904         spin_unlock(&bundle->channel_lock);
 905         _leave("");
 906         return;
 907 }
 908
 909 /*
 910  * Remove a connection from a bundle.
 911  */
 912 static void rxrpc_unbundle_conn(struct rxrpc_connection *conn)
 913 {
 914         struct rxrpc_bundle *bundle = conn->bundle;
 915         unsigned int bindex;
 916         bool need_drop = false;
 917         int i;
 918
 919         _enter("C=%x", conn->debug_id);
 920
 921         if (conn->flags & RXRPC_CONN_FINAL_ACK_MASK)
 922                 rxrpc_process_delayed_final_acks(conn, true);
 923
 924         spin_lock(&bundle->channel_lock);
 925         bindex = conn->bundle_shift / RXRPC_MAXCALLS;
 926         if (bundle->conns[bindex] == conn) {
 927                 _debug("clear slot %u", bindex);
 928                 bundle->conns[bindex] = NULL;
 929                 for (i = 0; i < RXRPC_MAXCALLS; i++)
 930                         clear_bit(conn->bundle_shift + i, &bundle->avail_chans);
 931                 need_drop = true;
 932         }
 933         spin_unlock(&bundle->channel_lock);
 934
 935         if (need_drop) {
 936                 rxrpc_deactivate_bundle(bundle);
 937                 rxrpc_put_connection(conn);
 938         }
 939 }
 940
 941 /*
 942  * Drop the active count on a bundle.
 943  */
 944 static void rxrpc_deactivate_bundle(struct rxrpc_bundle *bundle)
 945 {
 946         struct rxrpc_local *local = bundle->params.local;
 947         bool need_put = false;
 948
 949         if (atomic_dec_and_lock(&bundle->active, &local->client_bundles_lock)) {
 950                 if (!bundle->params.exclusive) {
 951                         _debug("erase bundle");
 952                         rb_erase(&bundle->local_node, &local->client_bundles);
 953                         need_put = true;
 954                 }
 955
 956                 spin_unlock(&local->client_bundles_lock);
 957                 if (need_put)
 958                         rxrpc_put_bundle(bundle);
 959         }
 960 }
 961
 962 /*
 963  * Clean up a dead client connection.
 964  */
 965 static void rxrpc_kill_client_conn(struct rxrpc_connection *conn)
 966 {
 967         struct rxrpc_local *local = conn->params.local;
 968         struct rxrpc_net *rxnet = local->rxnet;
 969
 970         _enter("C=%x", conn->debug_id);
 971
 972         trace_rxrpc_client(conn, -1, rxrpc_client_cleanup);
 973         atomic_dec(&rxnet->nr_client_conns);
 974
 975         rxrpc_put_client_connection_id(conn);
 976         rxrpc_kill_connection(conn);
 977 }
 978
 979 /*
 980  * Clean up a dead client connections.
 981  */
 982 void rxrpc_put_client_conn(struct rxrpc_connection *conn)
 983 {
 984         const void *here = __builtin_return_address(0);
 985         unsigned int debug_id = conn->debug_id;
 986         bool dead;
 987         int r;
 988
 989         dead = __refcount_dec_and_test(&conn->ref, &r);
 990         trace_rxrpc_conn(debug_id, rxrpc_conn_put_client, r - 1, here);
 991         if (dead)
 992                 rxrpc_kill_client_conn(conn);
 993 }
 994
 995 /*
 996  * Discard expired client connections from the idle list.  Each conn in the
 997  * idle list has been exposed and holds an extra ref because of that.
 998  *
 999  * This may be called from conn setup or from a work item so cannot be
1000  * considered non-reentrant.
1001  */
1002 void rxrpc_discard_expired_client_conns(struct work_struct *work)
1003 {
1004         struct rxrpc_connection *conn;
1005         struct rxrpc_net *rxnet =
1006                 container_of(work, struct rxrpc_net, client_conn_reaper);
1007         unsigned long expiry, conn_expires_at, now;
1008         unsigned int nr_conns;
1009
1010         _enter("");
1011
1012         if (list_empty(&rxnet->idle_client_conns)) {
1013                 _leave(" [empty]");
1014                 return;
1015         }
1016
1017         /* Don't double up on the discarding */
1018         if (!spin_trylock(&rxnet->client_conn_discard_lock)) {
1019                 _leave(" [already]");
1020                 return;
1021         }
1022
1023         /* We keep an estimate of what the number of conns ought to be after
1024          * we've discarded some so that we don't overdo the discarding.
1025          */
1026         nr_conns = atomic_read(&rxnet->nr_client_conns);
1027
1028 next:
1029         spin_lock(&rxnet->client_conn_cache_lock);
1030
1031         if (list_empty(&rxnet->idle_client_conns))
1032                 goto out;
1033
1034         conn = list_entry(rxnet->idle_client_conns.next,
1035                           struct rxrpc_connection, cache_link);
1036
1037         if (!rxnet->kill_all_client_conns) {
1038                 /* If the number of connections is over the reap limit, we
1039                  * expedite discard by reducing the expiry timeout.  We must,
1040                  * however, have at least a short grace period to be able to do
1041                  * final-ACK or ABORT retransmission.
1042                  */
1043                 expiry = rxrpc_conn_idle_client_expiry;
1044                 if (nr_conns > rxrpc_reap_client_connections)
1045                         expiry = rxrpc_conn_idle_client_fast_expiry;
1046                 if (conn->params.local->service_closed)
1047                         expiry = rxrpc_closed_conn_expiry * HZ;
1048
1049                 conn_expires_at = conn->idle_timestamp + expiry;
1050
1051                 now = READ_ONCE(jiffies);
1052                 if (time_after(conn_expires_at, now))
1053                         goto not_yet_expired;
1054         }
1055
1056         trace_rxrpc_client(conn, -1, rxrpc_client_discard);
1057         list_del_init(&conn->cache_link);
1058
1059         spin_unlock(&rxnet->client_conn_cache_lock);
1060
1061         rxrpc_unbundle_conn(conn);
1062         rxrpc_put_connection(conn); /* Drop the ->cache_link ref */
1063
1064         nr_conns--;
1065         goto next;
1066
1067 not_yet_expired:
1068         /* The connection at the front of the queue hasn't yet expired, so
1069          * schedule the work item for that point if we discarded something.
1070          *
1071          * We don't worry if the work item is already scheduled - it can look
1072          * after rescheduling itself at a later time.  We could cancel it, but
1073          * then things get messier.
1074          */
1075         _debug("not yet");
1076         if (!rxnet->kill_all_client_conns)
1077                 timer_reduce(&rxnet->client_conn_reap_timer, conn_expires_at);
1078
1079 out:
1080         spin_unlock(&rxnet->client_conn_cache_lock);
1081         spin_unlock(&rxnet->client_conn_discard_lock);
1082         _leave("");
1083 }
1084
1085 /*
1086  * Preemptively destroy all the client connection records rather than waiting
1087  * for them to time out
1088  */
1089 void rxrpc_destroy_all_client_connections(struct rxrpc_net *rxnet)
1090 {
1091         _enter("");
1092
1093         spin_lock(&rxnet->client_conn_cache_lock);
1094         rxnet->kill_all_client_conns = true;
1095         spin_unlock(&rxnet->client_conn_cache_lock);
1096
1097         del_timer_sync(&rxnet->client_conn_reap_timer);
1098
1099         if (!rxrpc_queue_work(&rxnet->client_conn_reaper))
1100                 _debug("destroy: queue failed");
1101
1102         _leave("");
1103 }
1104
1105 /*
1106  * Clean up the client connections on a local endpoint.
1107  */
1108 void rxrpc_clean_up_local_conns(struct rxrpc_local *local)
1109 {
1110         struct rxrpc_connection *conn, *tmp;
1111         struct rxrpc_net *rxnet = local->rxnet;
1112         LIST_HEAD(graveyard);
1113
1114         _enter("");
1115
1116         spin_lock(&rxnet->client_conn_cache_lock);
1117
1118         list_for_each_entry_safe(conn, tmp, &rxnet->idle_client_conns,
1119                                  cache_link) {
1120                 if (conn->params.local == local) {
1121                         trace_rxrpc_client(conn, -1, rxrpc_client_discard);
1122                         list_move(&conn->cache_link, &graveyard);
1123                 }
1124         }
1125
1126         spin_unlock(&rxnet->client_conn_cache_lock);
1127
1128         while (!list_empty(&graveyard)) {
1129                 conn = list_entry(graveyard.next,
1130                                   struct rxrpc_connection, cache_link);
1131                 list_del_init(&conn->cache_link);
1132                 rxrpc_unbundle_conn(conn);
1133                 rxrpc_put_connection(conn);
1134         }
1135
1136         _leave(" [culled]");
1137 }