net/rxrpc/af_rxrpc.c

   1 /* AF_RXRPC implementation
   2  *
   3  * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
   4  * Written by David Howells (dhowells@redhat.com)
   5  *
   6  * This program is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU General Public License
   8  * as published by the Free Software Foundation; either version
   9  * 2 of the License, or (at your option) any later version.
  10  */
  11
  12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  13
  14 #include <linux/module.h>
  15 #include <linux/kernel.h>
  16 #include <linux/net.h>
  17 #include <linux/slab.h>
  18 #include <linux/skbuff.h>
  19 #include <linux/random.h>
  20 #include <linux/poll.h>
  21 #include <linux/proc_fs.h>
  22 #include <linux/key-type.h>
  23 #include <net/net_namespace.h>
  24 #include <net/sock.h>
  25 #include <net/af_rxrpc.h>
  26 #define CREATE_TRACE_POINTS
  27 #include "ar-internal.h"
  28
  29 MODULE_DESCRIPTION("RxRPC network protocol");
  30 MODULE_AUTHOR("Red Hat, Inc.");
  31 MODULE_LICENSE("GPL");
  32 MODULE_ALIAS_NETPROTO(PF_RXRPC);
  33
  34 unsigned int rxrpc_debug; // = RXRPC_DEBUG_KPROTO;
  35 module_param_named(debug, rxrpc_debug, uint, 0644);
  36 MODULE_PARM_DESC(debug, "RxRPC debugging mask");
  37
  38 static struct proto rxrpc_proto;
  39 static const struct proto_ops rxrpc_rpc_ops;
  40
  41 /* current debugging ID */
  42 atomic_t rxrpc_debug_id;
  43 EXPORT_SYMBOL(rxrpc_debug_id);
  44
  45 /* count of skbs currently in use */
  46 atomic_t rxrpc_n_tx_skbs, rxrpc_n_rx_skbs;
  47
  48 struct workqueue_struct *rxrpc_workqueue;
  49
  50 static void rxrpc_sock_destructor(struct sock *);
  51
  52 /*
  53  * see if an RxRPC socket is currently writable
  54  */
  55 static inline int rxrpc_writable(struct sock *sk)
  56 {
  57         return refcount_read(&sk->sk_wmem_alloc) < (size_t) sk->sk_sndbuf;
  58 }
  59
  60 /*
  61  * wait for write bufferage to become available
  62  */
  63 static void rxrpc_write_space(struct sock *sk)
  64 {
  65         _enter("%p", sk);
  66         rcu_read_lock();
  67         if (rxrpc_writable(sk)) {
  68                 struct socket_wq *wq = rcu_dereference(sk->sk_wq);
  69
  70                 if (skwq_has_sleeper(wq))
  71                         wake_up_interruptible(&wq->wait);
  72                 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
  73         }
  74         rcu_read_unlock();
  75 }
  76
  77 /*
  78  * validate an RxRPC address
  79  */
  80 static int rxrpc_validate_address(struct rxrpc_sock *rx,
  81                                   struct sockaddr_rxrpc *srx,
  82                                   int len)
  83 {
  84         unsigned int tail;
  85
  86         if (len < sizeof(struct sockaddr_rxrpc))
  87                 return -EINVAL;
  88
  89         if (srx->srx_family != AF_RXRPC)
  90                 return -EAFNOSUPPORT;
  91
  92         if (srx->transport_type != SOCK_DGRAM)
  93                 return -ESOCKTNOSUPPORT;
  94
  95         len -= offsetof(struct sockaddr_rxrpc, transport);
  96         if (srx->transport_len < sizeof(sa_family_t) ||
  97             srx->transport_len > len)
  98                 return -EINVAL;
  99
 100         if (srx->transport.family != rx->family &&
 101             srx->transport.family == AF_INET && rx->family != AF_INET6)
 102                 return -EAFNOSUPPORT;
 103
 104         switch (srx->transport.family) {
 105         case AF_INET:
 106                 if (srx->transport_len < sizeof(struct sockaddr_in))
 107                         return -EINVAL;
 108                 tail = offsetof(struct sockaddr_rxrpc, transport.sin.__pad);
 109                 break;
 110
 111 #ifdef CONFIG_AF_RXRPC_IPV6
 112         case AF_INET6:
 113                 if (srx->transport_len < sizeof(struct sockaddr_in6))
 114                         return -EINVAL;
 115                 tail = offsetof(struct sockaddr_rxrpc, transport) +
 116                         sizeof(struct sockaddr_in6);
 117                 break;
 118 #endif
 119
 120         default:
 121                 return -EAFNOSUPPORT;
 122         }
 123
 124         if (tail < len)
 125                 memset((void *)srx + tail, 0, len - tail);
 126         _debug("INET: %pISp", &srx->transport);
 127         return 0;
 128 }
 129
 130 /*
 131  * bind a local address to an RxRPC socket
 132  */
 133 static int rxrpc_bind(struct socket *sock, struct sockaddr *saddr, int len)
 134 {
 135         struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)saddr;
 136         struct rxrpc_local *local;
 137         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 138         u16 service_id;
 139         int ret;
 140
 141         _enter("%p,%p,%d", rx, saddr, len);
 142
 143         ret = rxrpc_validate_address(rx, srx, len);
 144         if (ret < 0)
 145                 goto error;
 146         service_id = srx->srx_service;
 147
 148         lock_sock(&rx->sk);
 149
 150         switch (rx->sk.sk_state) {
 151         case RXRPC_UNBOUND:
 152                 rx->srx = *srx;
 153                 local = rxrpc_lookup_local(sock_net(&rx->sk), &rx->srx);
 154                 if (IS_ERR(local)) {
 155                         ret = PTR_ERR(local);
 156                         goto error_unlock;
 157                 }
 158
 159                 if (service_id) {
 160                         write_lock(&local->services_lock);
 161                         if (rcu_access_pointer(local->service))
 162                                 goto service_in_use;
 163                         rx->local = local;
 164                         rcu_assign_pointer(local->service, rx);
 165                         write_unlock(&local->services_lock);
 166
 167                         rx->sk.sk_state = RXRPC_SERVER_BOUND;
 168                 } else {
 169                         rx->local = local;
 170                         rx->sk.sk_state = RXRPC_CLIENT_BOUND;
 171                 }
 172                 break;
 173
 174         case RXRPC_SERVER_BOUND:
 175                 ret = -EINVAL;
 176                 if (service_id == 0)
 177                         goto error_unlock;
 178                 ret = -EADDRINUSE;
 179                 if (service_id == rx->srx.srx_service)
 180                         goto error_unlock;
 181                 ret = -EINVAL;
 182                 srx->srx_service = rx->srx.srx_service;
 183                 if (memcmp(srx, &rx->srx, sizeof(*srx)) != 0)
 184                         goto error_unlock;
 185                 rx->second_service = service_id;
 186                 rx->sk.sk_state = RXRPC_SERVER_BOUND2;
 187                 break;
 188
 189         default:
 190                 ret = -EINVAL;
 191                 goto error_unlock;
 192         }
 193
 194         release_sock(&rx->sk);
 195         _leave(" = 0");
 196         return 0;
 197
 198 service_in_use:
 199         write_unlock(&local->services_lock);
 200         rxrpc_put_local(local);
 201         ret = -EADDRINUSE;
 202 error_unlock:
 203         release_sock(&rx->sk);
 204 error:
 205         _leave(" = %d", ret);
 206         return ret;
 207 }
 208
 209 /*
 210  * set the number of pending calls permitted on a listening socket
 211  */
 212 static int rxrpc_listen(struct socket *sock, int backlog)
 213 {
 214         struct sock *sk = sock->sk;
 215         struct rxrpc_sock *rx = rxrpc_sk(sk);
 216         unsigned int max, old;
 217         int ret;
 218
 219         _enter("%p,%d", rx, backlog);
 220
 221         lock_sock(&rx->sk);
 222
 223         switch (rx->sk.sk_state) {
 224         case RXRPC_UNBOUND:
 225                 ret = -EADDRNOTAVAIL;
 226                 break;
 227         case RXRPC_SERVER_BOUND:
 228         case RXRPC_SERVER_BOUND2:
 229                 ASSERT(rx->local != NULL);
 230                 max = READ_ONCE(rxrpc_max_backlog);
 231                 ret = -EINVAL;
 232                 if (backlog == INT_MAX)
 233                         backlog = max;
 234                 else if (backlog < 0 || backlog > max)
 235                         break;
 236                 old = sk->sk_max_ack_backlog;
 237                 sk->sk_max_ack_backlog = backlog;
 238                 ret = rxrpc_service_prealloc(rx, GFP_KERNEL);
 239                 if (ret == 0)
 240                         rx->sk.sk_state = RXRPC_SERVER_LISTENING;
 241                 else
 242                         sk->sk_max_ack_backlog = old;
 243                 break;
 244         case RXRPC_SERVER_LISTENING:
 245                 if (backlog == 0) {
 246                         rx->sk.sk_state = RXRPC_SERVER_LISTEN_DISABLED;
 247                         sk->sk_max_ack_backlog = 0;
 248                         rxrpc_discard_prealloc(rx);
 249                         ret = 0;
 250                         break;
 251                 }
 252                 /* Fall through */
 253         default:
 254                 ret = -EBUSY;
 255                 break;
 256         }
 257
 258         release_sock(&rx->sk);
 259         _leave(" = %d", ret);
 260         return ret;
 261 }
 262
 263 /**
 264  * rxrpc_kernel_begin_call - Allow a kernel service to begin a call
 265  * @sock: The socket on which to make the call
 266  * @srx: The address of the peer to contact
 267  * @key: The security context to use (defaults to socket setting)
 268  * @user_call_ID: The ID to use
 269  * @tx_total_len: Total length of data to transmit during the call (or -1)
 270  * @gfp: The allocation constraints
 271  * @notify_rx: Where to send notifications instead of socket queue
 272  * @upgrade: Request service upgrade for call
 273  * @debug_id: The debug ID for tracing to be assigned to the call
 274  *
 275  * Allow a kernel service to begin a call on the nominated socket.  This just
 276  * sets up all the internal tracking structures and allocates connection and
 277  * call IDs as appropriate.  The call to be used is returned.
 278  *
 279  * The default socket destination address and security may be overridden by
 280  * supplying @srx and @key.
 281  */
 282 struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
 283                                            struct sockaddr_rxrpc *srx,
 284                                            struct key *key,
 285                                            unsigned long user_call_ID,
 286                                            s64 tx_total_len,
 287                                            gfp_t gfp,
 288                                            rxrpc_notify_rx_t notify_rx,
 289                                            bool upgrade,
 290                                            unsigned int debug_id)
 291 {
 292         struct rxrpc_conn_parameters cp;
 293         struct rxrpc_call_params p;
 294         struct rxrpc_call *call;
 295         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 296         int ret;
 297
 298         _enter(",,%x,%lx", key_serial(key), user_call_ID);
 299
 300         ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
 301         if (ret < 0)
 302                 return ERR_PTR(ret);
 303
 304         lock_sock(&rx->sk);
 305
 306         if (!key)
 307                 key = rx->key;
 308         if (key && !key->payload.data[0])
 309                 key = NULL; /* a no-security key */
 310
 311         memset(&p, 0, sizeof(p));
 312         p.user_call_ID = user_call_ID;
 313         p.tx_total_len = tx_total_len;
 314
 315         memset(&cp, 0, sizeof(cp));
 316         cp.local                = rx->local;
 317         cp.key                  = key;
 318         cp.security_level       = rx->min_sec_level;
 319         cp.exclusive            = false;
 320         cp.upgrade              = upgrade;
 321         cp.service_id           = srx->srx_service;
 322         call = rxrpc_new_client_call(rx, &cp, srx, &p, gfp, debug_id);
 323         /* The socket has been unlocked. */
 324         if (!IS_ERR(call)) {
 325                 call->notify_rx = notify_rx;
 326                 mutex_unlock(&call->user_mutex);
 327         }
 328
 329         rxrpc_put_peer(cp.peer);
 330         _leave(" = %p", call);
 331         return call;
 332 }
 333 EXPORT_SYMBOL(rxrpc_kernel_begin_call);
 334
 335 /*
 336  * Dummy function used to stop the notifier talking to recvmsg().
 337  */
 338 static void rxrpc_dummy_notify_rx(struct sock *sk, struct rxrpc_call *rxcall,
 339                                   unsigned long call_user_ID)
 340 {
 341 }
 342
 343 /**
 344  * rxrpc_kernel_end_call - Allow a kernel service to end a call it was using
 345  * @sock: The socket the call is on
 346  * @call: The call to end
 347  *
 348  * Allow a kernel service to end a call it was using.  The call must be
 349  * complete before this is called (the call should be aborted if necessary).
 350  */
 351 void rxrpc_kernel_end_call(struct socket *sock, struct rxrpc_call *call)
 352 {
 353         _enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
 354
 355         mutex_lock(&call->user_mutex);
 356         rxrpc_release_call(rxrpc_sk(sock->sk), call);
 357
 358         /* Make sure we're not going to call back into a kernel service */
 359         if (call->notify_rx) {
 360                 spin_lock_bh(&call->notify_lock);
 361                 call->notify_rx = rxrpc_dummy_notify_rx;
 362                 spin_unlock_bh(&call->notify_lock);
 363         }
 364
 365         mutex_unlock(&call->user_mutex);
 366         rxrpc_put_call(call, rxrpc_call_put_kernel);
 367 }
 368 EXPORT_SYMBOL(rxrpc_kernel_end_call);
 369
 370 /**
 371  * rxrpc_kernel_check_life - Check to see whether a call is still alive
 372  * @sock: The socket the call is on
 373  * @call: The call to check
 374  * @_life: Where to store the life value
 375  *
 376  * Allow a kernel service to find out whether a call is still alive - ie. we're
 377  * getting ACKs from the server.  Passes back in *_life a number representing
 378  * the life state which can be compared to that returned by a previous call and
 379  * return true if the call is still alive.
 380  *
 381  * If the life state stalls, rxrpc_kernel_probe_life() should be called and
 382  * then 2RTT waited.
 383  */
 384 bool rxrpc_kernel_check_life(const struct socket *sock,
 385                              const struct rxrpc_call *call,
 386                              u32 *_life)
 387 {
 388         *_life = call->acks_latest;
 389         return call->state != RXRPC_CALL_COMPLETE;
 390 }
 391 EXPORT_SYMBOL(rxrpc_kernel_check_life);
 392
 393 /**
 394  * rxrpc_kernel_probe_life - Poke the peer to see if it's still alive
 395  * @sock: The socket the call is on
 396  * @call: The call to check
 397  *
 398  * In conjunction with rxrpc_kernel_check_life(), allow a kernel service to
 399  * find out whether a call is still alive by pinging it.  This should cause the
 400  * life state to be bumped in about 2*RTT.
 401  *
 402  * The must be called in TASK_RUNNING state on pain of might_sleep() objecting.
 403  */
 404 void rxrpc_kernel_probe_life(struct socket *sock, struct rxrpc_call *call)
 405 {
 406         rxrpc_propose_ACK(call, RXRPC_ACK_PING, 0, 0, true, false,
 407                           rxrpc_propose_ack_ping_for_check_life);
 408         rxrpc_send_ack_packet(call, true, NULL);
 409 }
 410 EXPORT_SYMBOL(rxrpc_kernel_probe_life);
 411
 412 /**
 413  * rxrpc_kernel_get_epoch - Retrieve the epoch value from a call.
 414  * @sock: The socket the call is on
 415  * @call: The call to query
 416  *
 417  * Allow a kernel service to retrieve the epoch value from a service call to
 418  * see if the client at the other end rebooted.
 419  */
 420 u32 rxrpc_kernel_get_epoch(struct socket *sock, struct rxrpc_call *call)
 421 {
 422         return call->conn->proto.epoch;
 423 }
 424 EXPORT_SYMBOL(rxrpc_kernel_get_epoch);
 425
 426 /**
 427  * rxrpc_kernel_new_call_notification - Get notifications of new calls
 428  * @sock: The socket to intercept received messages on
 429  * @notify_new_call: Function to be called when new calls appear
 430  * @discard_new_call: Function to discard preallocated calls
 431  *
 432  * Allow a kernel service to be given notifications about new calls.
 433  */
 434 void rxrpc_kernel_new_call_notification(
 435         struct socket *sock,
 436         rxrpc_notify_new_call_t notify_new_call,
 437         rxrpc_discard_new_call_t discard_new_call)
 438 {
 439         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 440
 441         rx->notify_new_call = notify_new_call;
 442         rx->discard_new_call = discard_new_call;
 443 }
 444 EXPORT_SYMBOL(rxrpc_kernel_new_call_notification);
 445
 446 /*
 447  * connect an RxRPC socket
 448  * - this just targets it at a specific destination; no actual connection
 449  *   negotiation takes place
 450  */
 451 static int rxrpc_connect(struct socket *sock, struct sockaddr *addr,
 452                          int addr_len, int flags)
 453 {
 454         struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
 455         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 456         int ret;
 457
 458         _enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
 459
 460         ret = rxrpc_validate_address(rx, srx, addr_len);
 461         if (ret < 0) {
 462                 _leave(" = %d [bad addr]", ret);
 463                 return ret;
 464         }
 465
 466         lock_sock(&rx->sk);
 467
 468         ret = -EISCONN;
 469         if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
 470                 goto error;
 471
 472         switch (rx->sk.sk_state) {
 473         case RXRPC_UNBOUND:
 474                 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
 475         case RXRPC_CLIENT_UNBOUND:
 476         case RXRPC_CLIENT_BOUND:
 477                 break;
 478         default:
 479                 ret = -EBUSY;
 480                 goto error;
 481         }
 482
 483         rx->connect_srx = *srx;
 484         set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
 485         ret = 0;
 486
 487 error:
 488         release_sock(&rx->sk);
 489         return ret;
 490 }
 491
 492 /*
 493  * send a message through an RxRPC socket
 494  * - in a client this does a number of things:
 495  *   - finds/sets up a connection for the security specified (if any)
 496  *   - initiates a call (ID in control data)
 497  *   - ends the request phase of a call (if MSG_MORE is not set)
 498  *   - sends a call data packet
 499  *   - may send an abort (abort code in control data)
 500  */
 501 static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
 502 {
 503         struct rxrpc_local *local;
 504         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 505         int ret;
 506
 507         _enter(",{%d},,%zu", rx->sk.sk_state, len);
 508
 509         if (m->msg_flags & MSG_OOB)
 510                 return -EOPNOTSUPP;
 511
 512         if (m->msg_name) {
 513                 ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
 514                 if (ret < 0) {
 515                         _leave(" = %d [bad addr]", ret);
 516                         return ret;
 517                 }
 518         }
 519
 520         lock_sock(&rx->sk);
 521
 522         switch (rx->sk.sk_state) {
 523         case RXRPC_UNBOUND:
 524                 rx->srx.srx_family = AF_RXRPC;
 525                 rx->srx.srx_service = 0;
 526                 rx->srx.transport_type = SOCK_DGRAM;
 527                 rx->srx.transport.family = rx->family;
 528                 switch (rx->family) {
 529                 case AF_INET:
 530                         rx->srx.transport_len = sizeof(struct sockaddr_in);
 531                         break;
 532 #ifdef CONFIG_AF_RXRPC_IPV6
 533                 case AF_INET6:
 534                         rx->srx.transport_len = sizeof(struct sockaddr_in6);
 535                         break;
 536 #endif
 537                 default:
 538                         ret = -EAFNOSUPPORT;
 539                         goto error_unlock;
 540                 }
 541                 local = rxrpc_lookup_local(sock_net(sock->sk), &rx->srx);
 542                 if (IS_ERR(local)) {
 543                         ret = PTR_ERR(local);
 544                         goto error_unlock;
 545                 }
 546
 547                 rx->local = local;
 548                 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
 549                 /* Fall through */
 550
 551         case RXRPC_CLIENT_UNBOUND:
 552         case RXRPC_CLIENT_BOUND:
 553                 if (!m->msg_name &&
 554                     test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
 555                         m->msg_name = &rx->connect_srx;
 556                         m->msg_namelen = sizeof(rx->connect_srx);
 557                 }
 558                 /* Fall through */
 559         case RXRPC_SERVER_BOUND:
 560         case RXRPC_SERVER_LISTENING:
 561                 ret = rxrpc_do_sendmsg(rx, m, len);
 562                 /* The socket has been unlocked */
 563                 goto out;
 564         default:
 565                 ret = -EINVAL;
 566                 goto error_unlock;
 567         }
 568
 569 error_unlock:
 570         release_sock(&rx->sk);
 571 out:
 572         _leave(" = %d", ret);
 573         return ret;
 574 }
 575
 576 /*
 577  * set RxRPC socket options
 578  */
 579 static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
 580                             char __user *optval, unsigned int optlen)
 581 {
 582         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 583         unsigned int min_sec_level;
 584         u16 service_upgrade[2];
 585         int ret;
 586
 587         _enter(",%d,%d,,%d", level, optname, optlen);
 588
 589         lock_sock(&rx->sk);
 590         ret = -EOPNOTSUPP;
 591
 592         if (level == SOL_RXRPC) {
 593                 switch (optname) {
 594                 case RXRPC_EXCLUSIVE_CONNECTION:
 595                         ret = -EINVAL;
 596                         if (optlen != 0)
 597                                 goto error;
 598                         ret = -EISCONN;
 599                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 600                                 goto error;
 601                         rx->exclusive = true;
 602                         goto success;
 603
 604                 case RXRPC_SECURITY_KEY:
 605                         ret = -EINVAL;
 606                         if (rx->key)
 607                                 goto error;
 608                         ret = -EISCONN;
 609                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 610                                 goto error;
 611                         ret = rxrpc_request_key(rx, optval, optlen);
 612                         goto error;
 613
 614                 case RXRPC_SECURITY_KEYRING:
 615                         ret = -EINVAL;
 616                         if (rx->key)
 617                                 goto error;
 618                         ret = -EISCONN;
 619                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 620                                 goto error;
 621                         ret = rxrpc_server_keyring(rx, optval, optlen);
 622                         goto error;
 623
 624                 case RXRPC_MIN_SECURITY_LEVEL:
 625                         ret = -EINVAL;
 626                         if (optlen != sizeof(unsigned int))
 627                                 goto error;
 628                         ret = -EISCONN;
 629                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 630                                 goto error;
 631                         ret = get_user(min_sec_level,
 632                                        (unsigned int __user *) optval);
 633                         if (ret < 0)
 634                                 goto error;
 635                         ret = -EINVAL;
 636                         if (min_sec_level > RXRPC_SECURITY_MAX)
 637                                 goto error;
 638                         rx->min_sec_level = min_sec_level;
 639                         goto success;
 640
 641                 case RXRPC_UPGRADEABLE_SERVICE:
 642                         ret = -EINVAL;
 643                         if (optlen != sizeof(service_upgrade) ||
 644                             rx->service_upgrade.from != 0)
 645                                 goto error;
 646                         ret = -EISCONN;
 647                         if (rx->sk.sk_state != RXRPC_SERVER_BOUND2)
 648                                 goto error;
 649                         ret = -EFAULT;
 650                         if (copy_from_user(service_upgrade, optval,
 651                                            sizeof(service_upgrade)) != 0)
 652                                 goto error;
 653                         ret = -EINVAL;
 654                         if ((service_upgrade[0] != rx->srx.srx_service ||
 655                              service_upgrade[1] != rx->second_service) &&
 656                             (service_upgrade[0] != rx->second_service ||
 657                              service_upgrade[1] != rx->srx.srx_service))
 658                                 goto error;
 659                         rx->service_upgrade.from = service_upgrade[0];
 660                         rx->service_upgrade.to = service_upgrade[1];
 661                         goto success;
 662
 663                 default:
 664                         break;
 665                 }
 666         }
 667
 668 success:
 669         ret = 0;
 670 error:
 671         release_sock(&rx->sk);
 672         return ret;
 673 }
 674
 675 /*
 676  * Get socket options.
 677  */
 678 static int rxrpc_getsockopt(struct socket *sock, int level, int optname,
 679                             char __user *optval, int __user *_optlen)
 680 {
 681         int optlen;
 682
 683         if (level != SOL_RXRPC)
 684                 return -EOPNOTSUPP;
 685
 686         if (get_user(optlen, _optlen))
 687                 return -EFAULT;
 688
 689         switch (optname) {
 690         case RXRPC_SUPPORTED_CMSG:
 691                 if (optlen < sizeof(int))
 692                         return -ETOOSMALL;
 693                 if (put_user(RXRPC__SUPPORTED - 1, (int __user *)optval) ||
 694                     put_user(sizeof(int), _optlen))
 695                         return -EFAULT;
 696                 return 0;
 697
 698         default:
 699                 return -EOPNOTSUPP;
 700         }
 701 }
 702
 703 /*
 704  * permit an RxRPC socket to be polled
 705  */
 706 static __poll_t rxrpc_poll(struct file *file, struct socket *sock,
 707                                poll_table *wait)
 708 {
 709         struct sock *sk = sock->sk;
 710         struct rxrpc_sock *rx = rxrpc_sk(sk);
 711         __poll_t mask;
 712
 713         sock_poll_wait(file, sock, wait);
 714         mask = 0;
 715
 716         /* the socket is readable if there are any messages waiting on the Rx
 717          * queue */
 718         if (!list_empty(&rx->recvmsg_q))
 719                 mask |= EPOLLIN | EPOLLRDNORM;
 720
 721         /* the socket is writable if there is space to add new data to the
 722          * socket; there is no guarantee that any particular call in progress
 723          * on the socket may have space in the Tx ACK window */
 724         if (rxrpc_writable(sk))
 725                 mask |= EPOLLOUT | EPOLLWRNORM;
 726
 727         return mask;
 728 }
 729
 730 /*
 731  * create an RxRPC socket
 732  */
 733 static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
 734                         int kern)
 735 {
 736         struct rxrpc_net *rxnet;
 737         struct rxrpc_sock *rx;
 738         struct sock *sk;
 739
 740         _enter("%p,%d", sock, protocol);
 741
 742         /* we support transport protocol UDP/UDP6 only */
 743         if (protocol != PF_INET &&
 744             IS_ENABLED(CONFIG_AF_RXRPC_IPV6) && protocol != PF_INET6)
 745                 return -EPROTONOSUPPORT;
 746
 747         if (sock->type != SOCK_DGRAM)
 748                 return -ESOCKTNOSUPPORT;
 749
 750         sock->ops = &rxrpc_rpc_ops;
 751         sock->state = SS_UNCONNECTED;
 752
 753         sk = sk_alloc(net, PF_RXRPC, GFP_KERNEL, &rxrpc_proto, kern);
 754         if (!sk)
 755                 return -ENOMEM;
 756
 757         sock_init_data(sock, sk);
 758         sock_set_flag(sk, SOCK_RCU_FREE);
 759         sk->sk_state            = RXRPC_UNBOUND;
 760         sk->sk_write_space      = rxrpc_write_space;
 761         sk->sk_max_ack_backlog  = 0;
 762         sk->sk_destruct         = rxrpc_sock_destructor;
 763
 764         rx = rxrpc_sk(sk);
 765         rx->family = protocol;
 766         rx->calls = RB_ROOT;
 767
 768         spin_lock_init(&rx->incoming_lock);
 769         INIT_LIST_HEAD(&rx->sock_calls);
 770         INIT_LIST_HEAD(&rx->to_be_accepted);
 771         INIT_LIST_HEAD(&rx->recvmsg_q);
 772         rwlock_init(&rx->recvmsg_lock);
 773         rwlock_init(&rx->call_lock);
 774         memset(&rx->srx, 0, sizeof(rx->srx));
 775
 776         rxnet = rxrpc_net(sock_net(&rx->sk));
 777         timer_reduce(&rxnet->peer_keepalive_timer, jiffies + 1);
 778
 779         _leave(" = 0 [%p]", rx);
 780         return 0;
 781 }
 782
 783 /*
 784  * Kill all the calls on a socket and shut it down.
 785  */
 786 static int rxrpc_shutdown(struct socket *sock, int flags)
 787 {
 788         struct sock *sk = sock->sk;
 789         struct rxrpc_sock *rx = rxrpc_sk(sk);
 790         int ret = 0;
 791
 792         _enter("%p,%d", sk, flags);
 793
 794         if (flags != SHUT_RDWR)
 795                 return -EOPNOTSUPP;
 796         if (sk->sk_state == RXRPC_CLOSE)
 797                 return -ESHUTDOWN;
 798
 799         lock_sock(sk);
 800
 801         spin_lock_bh(&sk->sk_receive_queue.lock);
 802         if (sk->sk_state < RXRPC_CLOSE) {
 803                 sk->sk_state = RXRPC_CLOSE;
 804                 sk->sk_shutdown = SHUTDOWN_MASK;
 805         } else {
 806                 ret = -ESHUTDOWN;
 807         }
 808         spin_unlock_bh(&sk->sk_receive_queue.lock);
 809
 810         rxrpc_discard_prealloc(rx);
 811
 812         release_sock(sk);
 813         return ret;
 814 }
 815
 816 /*
 817  * RxRPC socket destructor
 818  */
 819 static void rxrpc_sock_destructor(struct sock *sk)
 820 {
 821         _enter("%p", sk);
 822
 823         rxrpc_purge_queue(&sk->sk_receive_queue);
 824
 825         WARN_ON(refcount_read(&sk->sk_wmem_alloc));
 826         WARN_ON(!sk_unhashed(sk));
 827         WARN_ON(sk->sk_socket);
 828
 829         if (!sock_flag(sk, SOCK_DEAD)) {
 830                 printk("Attempt to release alive rxrpc socket: %p\n", sk);
 831                 return;
 832         }
 833 }
 834
 835 /*
 836  * release an RxRPC socket
 837  */
 838 static int rxrpc_release_sock(struct sock *sk)
 839 {
 840         struct rxrpc_sock *rx = rxrpc_sk(sk);
 841         struct rxrpc_net *rxnet = rxrpc_net(sock_net(&rx->sk));
 842
 843         _enter("%p{%d,%d}", sk, sk->sk_state, refcount_read(&sk->sk_refcnt));
 844
 845         /* declare the socket closed for business */
 846         sock_orphan(sk);
 847         sk->sk_shutdown = SHUTDOWN_MASK;
 848
 849         /* We want to kill off all connections from a service socket
 850          * as fast as possible because we can't share these; client
 851          * sockets, on the other hand, can share an endpoint.
 852          */
 853         switch (sk->sk_state) {
 854         case RXRPC_SERVER_BOUND:
 855         case RXRPC_SERVER_BOUND2:
 856         case RXRPC_SERVER_LISTENING:
 857         case RXRPC_SERVER_LISTEN_DISABLED:
 858                 rx->local->service_closed = true;
 859                 break;
 860         }
 861
 862         spin_lock_bh(&sk->sk_receive_queue.lock);
 863         sk->sk_state = RXRPC_CLOSE;
 864         spin_unlock_bh(&sk->sk_receive_queue.lock);
 865
 866         if (rx->local && rcu_access_pointer(rx->local->service) == rx) {
 867                 write_lock(&rx->local->services_lock);
 868                 rcu_assign_pointer(rx->local->service, NULL);
 869                 write_unlock(&rx->local->services_lock);
 870         }
 871
 872         /* try to flush out this socket */
 873         rxrpc_discard_prealloc(rx);
 874         rxrpc_release_calls_on_socket(rx);
 875         flush_workqueue(rxrpc_workqueue);
 876         rxrpc_purge_queue(&sk->sk_receive_queue);
 877         rxrpc_queue_work(&rxnet->service_conn_reaper);
 878         rxrpc_queue_work(&rxnet->client_conn_reaper);
 879
 880         rxrpc_put_local(rx->local);
 881         rx->local = NULL;
 882         key_put(rx->key);
 883         rx->key = NULL;
 884         key_put(rx->securities);
 885         rx->securities = NULL;
 886         sock_put(sk);
 887
 888         _leave(" = 0");
 889         return 0;
 890 }
 891
 892 /*
 893  * release an RxRPC BSD socket on close() or equivalent
 894  */
 895 static int rxrpc_release(struct socket *sock)
 896 {
 897         struct sock *sk = sock->sk;
 898
 899         _enter("%p{%p}", sock, sk);
 900
 901         if (!sk)
 902                 return 0;
 903
 904         sock->sk = NULL;
 905
 906         return rxrpc_release_sock(sk);
 907 }
 908
 909 /*
 910  * RxRPC network protocol
 911  */
 912 static const struct proto_ops rxrpc_rpc_ops = {
 913         .family         = PF_RXRPC,
 914         .owner          = THIS_MODULE,
 915         .release        = rxrpc_release,
 916         .bind           = rxrpc_bind,
 917         .connect        = rxrpc_connect,
 918         .socketpair     = sock_no_socketpair,
 919         .accept         = sock_no_accept,
 920         .getname        = sock_no_getname,
 921         .poll           = rxrpc_poll,
 922         .ioctl          = sock_no_ioctl,
 923         .listen         = rxrpc_listen,
 924         .shutdown       = rxrpc_shutdown,
 925         .setsockopt     = rxrpc_setsockopt,
 926         .getsockopt     = rxrpc_getsockopt,
 927         .sendmsg        = rxrpc_sendmsg,
 928         .recvmsg        = rxrpc_recvmsg,
 929         .mmap           = sock_no_mmap,
 930         .sendpage       = sock_no_sendpage,
 931 };
 932
 933 static struct proto rxrpc_proto = {
 934         .name           = "RXRPC",
 935         .owner          = THIS_MODULE,
 936         .obj_size       = sizeof(struct rxrpc_sock),
 937         .max_header     = sizeof(struct rxrpc_wire_header),
 938 };
 939
 940 static const struct net_proto_family rxrpc_family_ops = {
 941         .family = PF_RXRPC,
 942         .create = rxrpc_create,
 943         .owner  = THIS_MODULE,
 944 };
 945
 946 /*
 947  * initialise and register the RxRPC protocol
 948  */
 949 static int __init af_rxrpc_init(void)
 950 {
 951         int ret = -1;
 952         unsigned int tmp;
 953
 954         BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > FIELD_SIZEOF(struct sk_buff, cb));
 955
 956         get_random_bytes(&tmp, sizeof(tmp));
 957         tmp &= 0x3fffffff;
 958         if (tmp == 0)
 959                 tmp = 1;
 960         idr_set_cursor(&rxrpc_client_conn_ids, tmp);
 961
 962         ret = -ENOMEM;
 963         rxrpc_call_jar = kmem_cache_create(
 964                 "rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
 965                 SLAB_HWCACHE_ALIGN, NULL);
 966         if (!rxrpc_call_jar) {
 967                 pr_notice("Failed to allocate call jar\n");
 968                 goto error_call_jar;
 969         }
 970
 971         rxrpc_workqueue = alloc_workqueue("krxrpcd", 0, 1);
 972         if (!rxrpc_workqueue) {
 973                 pr_notice("Failed to allocate work queue\n");
 974                 goto error_work_queue;
 975         }
 976
 977         ret = rxrpc_init_security();
 978         if (ret < 0) {
 979                 pr_crit("Cannot initialise security\n");
 980                 goto error_security;
 981         }
 982
 983         ret = register_pernet_subsys(&rxrpc_net_ops);
 984         if (ret)
 985                 goto error_pernet;
 986
 987         ret = proto_register(&rxrpc_proto, 1);
 988         if (ret < 0) {
 989                 pr_crit("Cannot register protocol\n");
 990                 goto error_proto;
 991         }
 992
 993         ret = sock_register(&rxrpc_family_ops);
 994         if (ret < 0) {
 995                 pr_crit("Cannot register socket family\n");
 996                 goto error_sock;
 997         }
 998
 999         ret = register_key_type(&key_type_rxrpc);
1000         if (ret < 0) {
1001                 pr_crit("Cannot register client key type\n");
1002                 goto error_key_type;
1003         }
1004
1005         ret = register_key_type(&key_type_rxrpc_s);
1006         if (ret < 0) {
1007                 pr_crit("Cannot register server key type\n");
1008                 goto error_key_type_s;
1009         }
1010
1011         ret = rxrpc_sysctl_init();
1012         if (ret < 0) {
1013                 pr_crit("Cannot register sysctls\n");
1014                 goto error_sysctls;
1015         }
1016
1017         return 0;
1018
1019 error_sysctls:
1020         unregister_key_type(&key_type_rxrpc_s);
1021 error_key_type_s:
1022         unregister_key_type(&key_type_rxrpc);
1023 error_key_type:
1024         sock_unregister(PF_RXRPC);
1025 error_sock:
1026         proto_unregister(&rxrpc_proto);
1027 error_proto:
1028         unregister_pernet_subsys(&rxrpc_net_ops);
1029 error_pernet:
1030         rxrpc_exit_security();
1031 error_security:
1032         destroy_workqueue(rxrpc_workqueue);
1033 error_work_queue:
1034         kmem_cache_destroy(rxrpc_call_jar);
1035 error_call_jar:
1036         return ret;
1037 }
1038
1039 /*
1040  * unregister the RxRPC protocol
1041  */
1042 static void __exit af_rxrpc_exit(void)
1043 {
1044         _enter("");
1045         rxrpc_sysctl_exit();
1046         unregister_key_type(&key_type_rxrpc_s);
1047         unregister_key_type(&key_type_rxrpc);
1048         sock_unregister(PF_RXRPC);
1049         proto_unregister(&rxrpc_proto);
1050         unregister_pernet_subsys(&rxrpc_net_ops);
1051         ASSERTCMP(atomic_read(&rxrpc_n_tx_skbs), ==, 0);
1052         ASSERTCMP(atomic_read(&rxrpc_n_rx_skbs), ==, 0);
1053
1054         /* Make sure the local and peer records pinned by any dying connections
1055          * are released.
1056          */
1057         rcu_barrier();
1058         rxrpc_destroy_client_conn_ids();
1059
1060         destroy_workqueue(rxrpc_workqueue);
1061         rxrpc_exit_security();
1062         kmem_cache_destroy(rxrpc_call_jar);
1063         _leave("");
1064 }
1065
1066 module_init(af_rxrpc_init);
1067 module_exit(af_rxrpc_exit);