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                 return -EAFNOSUPPORT;
 102
 103         switch (srx->transport.family) {
 104         case AF_INET:
 105                 if (srx->transport_len < sizeof(struct sockaddr_in))
 106                         return -EINVAL;
 107                 tail = offsetof(struct sockaddr_rxrpc, transport.sin.__pad);
 108                 break;
 109
 110 #ifdef CONFIG_AF_RXRPC_IPV6
 111         case AF_INET6:
 112                 if (srx->transport_len < sizeof(struct sockaddr_in6))
 113                         return -EINVAL;
 114                 tail = offsetof(struct sockaddr_rxrpc, transport) +
 115                         sizeof(struct sockaddr_in6);
 116                 break;
 117 #endif
 118
 119         default:
 120                 return -EAFNOSUPPORT;
 121         }
 122
 123         if (tail < len)
 124                 memset((void *)srx + tail, 0, len - tail);
 125         _debug("INET: %pISp", &srx->transport);
 126         return 0;
 127 }
 128
 129 /*
 130  * bind a local address to an RxRPC socket
 131  */
 132 static int rxrpc_bind(struct socket *sock, struct sockaddr *saddr, int len)
 133 {
 134         struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)saddr;
 135         struct rxrpc_local *local;
 136         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 137         u16 service_id = srx->srx_service;
 138         int ret;
 139
 140         _enter("%p,%p,%d", rx, saddr, len);
 141
 142         ret = rxrpc_validate_address(rx, srx, len);
 143         if (ret < 0)
 144                 goto error;
 145
 146         lock_sock(&rx->sk);
 147
 148         switch (rx->sk.sk_state) {
 149         case RXRPC_UNBOUND:
 150                 rx->srx = *srx;
 151                 local = rxrpc_lookup_local(sock_net(&rx->sk), &rx->srx);
 152                 if (IS_ERR(local)) {
 153                         ret = PTR_ERR(local);
 154                         goto error_unlock;
 155                 }
 156
 157                 if (service_id) {
 158                         write_lock(&local->services_lock);
 159                         if (rcu_access_pointer(local->service))
 160                                 goto service_in_use;
 161                         rx->local = local;
 162                         rcu_assign_pointer(local->service, rx);
 163                         write_unlock(&local->services_lock);
 164
 165                         rx->sk.sk_state = RXRPC_SERVER_BOUND;
 166                 } else {
 167                         rx->local = local;
 168                         rx->sk.sk_state = RXRPC_CLIENT_BOUND;
 169                 }
 170                 break;
 171
 172         case RXRPC_SERVER_BOUND:
 173                 ret = -EINVAL;
 174                 if (service_id == 0)
 175                         goto error_unlock;
 176                 ret = -EADDRINUSE;
 177                 if (service_id == rx->srx.srx_service)
 178                         goto error_unlock;
 179                 ret = -EINVAL;
 180                 srx->srx_service = rx->srx.srx_service;
 181                 if (memcmp(srx, &rx->srx, sizeof(*srx)) != 0)
 182                         goto error_unlock;
 183                 rx->second_service = service_id;
 184                 rx->sk.sk_state = RXRPC_SERVER_BOUND2;
 185                 break;
 186
 187         default:
 188                 ret = -EINVAL;
 189                 goto error_unlock;
 190         }
 191
 192         release_sock(&rx->sk);
 193         _leave(" = 0");
 194         return 0;
 195
 196 service_in_use:
 197         write_unlock(&local->services_lock);
 198         rxrpc_put_local(local);
 199         ret = -EADDRINUSE;
 200 error_unlock:
 201         release_sock(&rx->sk);
 202 error:
 203         _leave(" = %d", ret);
 204         return ret;
 205 }
 206
 207 /*
 208  * set the number of pending calls permitted on a listening socket
 209  */
 210 static int rxrpc_listen(struct socket *sock, int backlog)
 211 {
 212         struct sock *sk = sock->sk;
 213         struct rxrpc_sock *rx = rxrpc_sk(sk);
 214         unsigned int max, old;
 215         int ret;
 216
 217         _enter("%p,%d", rx, backlog);
 218
 219         lock_sock(&rx->sk);
 220
 221         switch (rx->sk.sk_state) {
 222         case RXRPC_UNBOUND:
 223                 ret = -EADDRNOTAVAIL;
 224                 break;
 225         case RXRPC_SERVER_BOUND:
 226         case RXRPC_SERVER_BOUND2:
 227                 ASSERT(rx->local != NULL);
 228                 max = READ_ONCE(rxrpc_max_backlog);
 229                 ret = -EINVAL;
 230                 if (backlog == INT_MAX)
 231                         backlog = max;
 232                 else if (backlog < 0 || backlog > max)
 233                         break;
 234                 old = sk->sk_max_ack_backlog;
 235                 sk->sk_max_ack_backlog = backlog;
 236                 ret = rxrpc_service_prealloc(rx, GFP_KERNEL);
 237                 if (ret == 0)
 238                         rx->sk.sk_state = RXRPC_SERVER_LISTENING;
 239                 else
 240                         sk->sk_max_ack_backlog = old;
 241                 break;
 242         case RXRPC_SERVER_LISTENING:
 243                 if (backlog == 0) {
 244                         rx->sk.sk_state = RXRPC_SERVER_LISTEN_DISABLED;
 245                         sk->sk_max_ack_backlog = 0;
 246                         rxrpc_discard_prealloc(rx);
 247                         ret = 0;
 248                         break;
 249                 }
 250                 /* Fall through */
 251         default:
 252                 ret = -EBUSY;
 253                 break;
 254         }
 255
 256         release_sock(&rx->sk);
 257         _leave(" = %d", ret);
 258         return ret;
 259 }
 260
 261 /**
 262  * rxrpc_kernel_begin_call - Allow a kernel service to begin a call
 263  * @sock: The socket on which to make the call
 264  * @srx: The address of the peer to contact
 265  * @key: The security context to use (defaults to socket setting)
 266  * @user_call_ID: The ID to use
 267  * @tx_total_len: Total length of data to transmit during the call (or -1)
 268  * @gfp: The allocation constraints
 269  * @notify_rx: Where to send notifications instead of socket queue
 270  * @upgrade: Request service upgrade for call
 271  * @debug_id: The debug ID for tracing to be assigned to the call
 272  *
 273  * Allow a kernel service to begin a call on the nominated socket.  This just
 274  * sets up all the internal tracking structures and allocates connection and
 275  * call IDs as appropriate.  The call to be used is returned.
 276  *
 277  * The default socket destination address and security may be overridden by
 278  * supplying @srx and @key.
 279  */
 280 struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
 281                                            struct sockaddr_rxrpc *srx,
 282                                            struct key *key,
 283                                            unsigned long user_call_ID,
 284                                            s64 tx_total_len,
 285                                            gfp_t gfp,
 286                                            rxrpc_notify_rx_t notify_rx,
 287                                            bool upgrade,
 288                                            unsigned int debug_id)
 289 {
 290         struct rxrpc_conn_parameters cp;
 291         struct rxrpc_call_params p;
 292         struct rxrpc_call *call;
 293         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 294         int ret;
 295
 296         _enter(",,%x,%lx", key_serial(key), user_call_ID);
 297
 298         ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
 299         if (ret < 0)
 300                 return ERR_PTR(ret);
 301
 302         lock_sock(&rx->sk);
 303
 304         if (!key)
 305                 key = rx->key;
 306         if (key && !key->payload.data[0])
 307                 key = NULL; /* a no-security key */
 308
 309         memset(&p, 0, sizeof(p));
 310         p.user_call_ID = user_call_ID;
 311         p.tx_total_len = tx_total_len;
 312
 313         memset(&cp, 0, sizeof(cp));
 314         cp.local                = rx->local;
 315         cp.key                  = key;
 316         cp.security_level       = rx->min_sec_level;
 317         cp.exclusive            = false;
 318         cp.upgrade              = upgrade;
 319         cp.service_id           = srx->srx_service;
 320         call = rxrpc_new_client_call(rx, &cp, srx, &p, gfp, debug_id);
 321         /* The socket has been unlocked. */
 322         if (!IS_ERR(call)) {
 323                 call->notify_rx = notify_rx;
 324                 mutex_unlock(&call->user_mutex);
 325         }
 326
 327         rxrpc_put_peer(cp.peer);
 328         _leave(" = %p", call);
 329         return call;
 330 }
 331 EXPORT_SYMBOL(rxrpc_kernel_begin_call);
 332
 333 /*
 334  * Dummy function used to stop the notifier talking to recvmsg().
 335  */
 336 static void rxrpc_dummy_notify_rx(struct sock *sk, struct rxrpc_call *rxcall,
 337                                   unsigned long call_user_ID)
 338 {
 339 }
 340
 341 /**
 342  * rxrpc_kernel_end_call - Allow a kernel service to end a call it was using
 343  * @sock: The socket the call is on
 344  * @call: The call to end
 345  *
 346  * Allow a kernel service to end a call it was using.  The call must be
 347  * complete before this is called (the call should be aborted if necessary).
 348  */
 349 void rxrpc_kernel_end_call(struct socket *sock, struct rxrpc_call *call)
 350 {
 351         _enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
 352
 353         mutex_lock(&call->user_mutex);
 354         rxrpc_release_call(rxrpc_sk(sock->sk), call);
 355
 356         /* Make sure we're not going to call back into a kernel service */
 357         if (call->notify_rx) {
 358                 spin_lock_bh(&call->notify_lock);
 359                 call->notify_rx = rxrpc_dummy_notify_rx;
 360                 spin_unlock_bh(&call->notify_lock);
 361         }
 362
 363         mutex_unlock(&call->user_mutex);
 364         rxrpc_put_call(call, rxrpc_call_put_kernel);
 365 }
 366 EXPORT_SYMBOL(rxrpc_kernel_end_call);
 367
 368 /**
 369  * rxrpc_kernel_check_life - Check to see whether a call is still alive
 370  * @sock: The socket the call is on
 371  * @call: The call to check
 372  *
 373  * Allow a kernel service to find out whether a call is still alive - ie. we're
 374  * getting ACKs from the server.  Returns a number representing the life state
 375  * which can be compared to that returned by a previous call.
 376  *
 377  * If this is a client call, ping ACKs will be sent to the server to find out
 378  * whether it's still responsive and whether the call is still alive on the
 379  * server.
 380  */
 381 u32 rxrpc_kernel_check_life(struct socket *sock, struct rxrpc_call *call)
 382 {
 383         return call->acks_latest;
 384 }
 385 EXPORT_SYMBOL(rxrpc_kernel_check_life);
 386
 387 /**
 388  * rxrpc_kernel_check_call - Check a call's state
 389  * @sock: The socket the call is on
 390  * @call: The call to check
 391  * @_compl: Where to store the completion state
 392  * @_abort_code: Where to store any abort code
 393  *
 394  * Allow a kernel service to query the state of a call and find out the manner
 395  * of its termination if it has completed.  Returns -EINPROGRESS if the call is
 396  * still going, 0 if the call finished successfully, -ECONNABORTED if the call
 397  * was aborted and an appropriate error if the call failed in some other way.
 398  */
 399 int rxrpc_kernel_check_call(struct socket *sock, struct rxrpc_call *call,
 400                             enum rxrpc_call_completion *_compl, u32 *_abort_code)
 401 {
 402         if (call->state != RXRPC_CALL_COMPLETE)
 403                 return -EINPROGRESS;
 404         smp_rmb();
 405         *_compl = call->completion;
 406         *_abort_code = call->abort_code;
 407         return call->error;
 408 }
 409 EXPORT_SYMBOL(rxrpc_kernel_check_call);
 410
 411 /**
 412  * rxrpc_kernel_retry_call - Allow a kernel service to retry a call
 413  * @sock: The socket the call is on
 414  * @call: The call to retry
 415  * @srx: The address of the peer to contact
 416  * @key: The security context to use (defaults to socket setting)
 417  *
 418  * Allow a kernel service to try resending a client call that failed due to a
 419  * network error to a new address.  The Tx queue is maintained intact, thereby
 420  * relieving the need to re-encrypt any request data that has already been
 421  * buffered.
 422  */
 423 int rxrpc_kernel_retry_call(struct socket *sock, struct rxrpc_call *call,
 424                             struct sockaddr_rxrpc *srx, struct key *key)
 425 {
 426         struct rxrpc_conn_parameters cp;
 427         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 428         int ret;
 429
 430         _enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
 431
 432         if (!key)
 433                 key = rx->key;
 434         if (key && !key->payload.data[0])
 435                 key = NULL; /* a no-security key */
 436
 437         memset(&cp, 0, sizeof(cp));
 438         cp.local                = rx->local;
 439         cp.key                  = key;
 440         cp.security_level       = 0;
 441         cp.exclusive            = false;
 442         cp.service_id           = srx->srx_service;
 443
 444         mutex_lock(&call->user_mutex);
 445
 446         ret = rxrpc_prepare_call_for_retry(rx, call);
 447         if (ret == 0)
 448                 ret = rxrpc_retry_client_call(rx, call, &cp, srx, GFP_KERNEL);
 449
 450         mutex_unlock(&call->user_mutex);
 451         rxrpc_put_peer(cp.peer);
 452         _leave(" = %d", ret);
 453         return ret;
 454 }
 455 EXPORT_SYMBOL(rxrpc_kernel_retry_call);
 456
 457 /**
 458  * rxrpc_kernel_new_call_notification - Get notifications of new calls
 459  * @sock: The socket to intercept received messages on
 460  * @notify_new_call: Function to be called when new calls appear
 461  * @discard_new_call: Function to discard preallocated calls
 462  *
 463  * Allow a kernel service to be given notifications about new calls.
 464  */
 465 void rxrpc_kernel_new_call_notification(
 466         struct socket *sock,
 467         rxrpc_notify_new_call_t notify_new_call,
 468         rxrpc_discard_new_call_t discard_new_call)
 469 {
 470         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 471
 472         rx->notify_new_call = notify_new_call;
 473         rx->discard_new_call = discard_new_call;
 474 }
 475 EXPORT_SYMBOL(rxrpc_kernel_new_call_notification);
 476
 477 /*
 478  * connect an RxRPC socket
 479  * - this just targets it at a specific destination; no actual connection
 480  *   negotiation takes place
 481  */
 482 static int rxrpc_connect(struct socket *sock, struct sockaddr *addr,
 483                          int addr_len, int flags)
 484 {
 485         struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
 486         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 487         int ret;
 488
 489         _enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
 490
 491         ret = rxrpc_validate_address(rx, srx, addr_len);
 492         if (ret < 0) {
 493                 _leave(" = %d [bad addr]", ret);
 494                 return ret;
 495         }
 496
 497         lock_sock(&rx->sk);
 498
 499         ret = -EISCONN;
 500         if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
 501                 goto error;
 502
 503         switch (rx->sk.sk_state) {
 504         case RXRPC_UNBOUND:
 505                 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
 506         case RXRPC_CLIENT_UNBOUND:
 507         case RXRPC_CLIENT_BOUND:
 508                 break;
 509         default:
 510                 ret = -EBUSY;
 511                 goto error;
 512         }
 513
 514         rx->connect_srx = *srx;
 515         set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
 516         ret = 0;
 517
 518 error:
 519         release_sock(&rx->sk);
 520         return ret;
 521 }
 522
 523 /*
 524  * send a message through an RxRPC socket
 525  * - in a client this does a number of things:
 526  *   - finds/sets up a connection for the security specified (if any)
 527  *   - initiates a call (ID in control data)
 528  *   - ends the request phase of a call (if MSG_MORE is not set)
 529  *   - sends a call data packet
 530  *   - may send an abort (abort code in control data)
 531  */
 532 static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
 533 {
 534         struct rxrpc_local *local;
 535         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 536         int ret;
 537
 538         _enter(",{%d},,%zu", rx->sk.sk_state, len);
 539
 540         if (m->msg_flags & MSG_OOB)
 541                 return -EOPNOTSUPP;
 542
 543         if (m->msg_name) {
 544                 ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
 545                 if (ret < 0) {
 546                         _leave(" = %d [bad addr]", ret);
 547                         return ret;
 548                 }
 549         }
 550
 551         lock_sock(&rx->sk);
 552
 553         switch (rx->sk.sk_state) {
 554         case RXRPC_UNBOUND:
 555                 rx->srx.srx_family = AF_RXRPC;
 556                 rx->srx.srx_service = 0;
 557                 rx->srx.transport_type = SOCK_DGRAM;
 558                 rx->srx.transport.family = rx->family;
 559                 switch (rx->family) {
 560                 case AF_INET:
 561                         rx->srx.transport_len = sizeof(struct sockaddr_in);
 562                         break;
 563 #ifdef CONFIG_AF_RXRPC_IPV6
 564                 case AF_INET6:
 565                         rx->srx.transport_len = sizeof(struct sockaddr_in6);
 566                         break;
 567 #endif
 568                 default:
 569                         ret = -EAFNOSUPPORT;
 570                         goto error_unlock;
 571                 }
 572                 local = rxrpc_lookup_local(sock_net(sock->sk), &rx->srx);
 573                 if (IS_ERR(local)) {
 574                         ret = PTR_ERR(local);
 575                         goto error_unlock;
 576                 }
 577
 578                 rx->local = local;
 579                 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
 580                 /* Fall through */
 581
 582         case RXRPC_CLIENT_UNBOUND:
 583         case RXRPC_CLIENT_BOUND:
 584                 if (!m->msg_name &&
 585                     test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
 586                         m->msg_name = &rx->connect_srx;
 587                         m->msg_namelen = sizeof(rx->connect_srx);
 588                 }
 589                 /* Fall through */
 590         case RXRPC_SERVER_BOUND:
 591         case RXRPC_SERVER_LISTENING:
 592                 ret = rxrpc_do_sendmsg(rx, m, len);
 593                 /* The socket has been unlocked */
 594                 goto out;
 595         default:
 596                 ret = -EINVAL;
 597                 goto error_unlock;
 598         }
 599
 600 error_unlock:
 601         release_sock(&rx->sk);
 602 out:
 603         _leave(" = %d", ret);
 604         return ret;
 605 }
 606
 607 /*
 608  * set RxRPC socket options
 609  */
 610 static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
 611                             char __user *optval, unsigned int optlen)
 612 {
 613         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 614         unsigned int min_sec_level;
 615         u16 service_upgrade[2];
 616         int ret;
 617
 618         _enter(",%d,%d,,%d", level, optname, optlen);
 619
 620         lock_sock(&rx->sk);
 621         ret = -EOPNOTSUPP;
 622
 623         if (level == SOL_RXRPC) {
 624                 switch (optname) {
 625                 case RXRPC_EXCLUSIVE_CONNECTION:
 626                         ret = -EINVAL;
 627                         if (optlen != 0)
 628                                 goto error;
 629                         ret = -EISCONN;
 630                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 631                                 goto error;
 632                         rx->exclusive = true;
 633                         goto success;
 634
 635                 case RXRPC_SECURITY_KEY:
 636                         ret = -EINVAL;
 637                         if (rx->key)
 638                                 goto error;
 639                         ret = -EISCONN;
 640                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 641                                 goto error;
 642                         ret = rxrpc_request_key(rx, optval, optlen);
 643                         goto error;
 644
 645                 case RXRPC_SECURITY_KEYRING:
 646                         ret = -EINVAL;
 647                         if (rx->key)
 648                                 goto error;
 649                         ret = -EISCONN;
 650                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 651                                 goto error;
 652                         ret = rxrpc_server_keyring(rx, optval, optlen);
 653                         goto error;
 654
 655                 case RXRPC_MIN_SECURITY_LEVEL:
 656                         ret = -EINVAL;
 657                         if (optlen != sizeof(unsigned int))
 658                                 goto error;
 659                         ret = -EISCONN;
 660                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 661                                 goto error;
 662                         ret = get_user(min_sec_level,
 663                                        (unsigned int __user *) optval);
 664                         if (ret < 0)
 665                                 goto error;
 666                         ret = -EINVAL;
 667                         if (min_sec_level > RXRPC_SECURITY_MAX)
 668                                 goto error;
 669                         rx->min_sec_level = min_sec_level;
 670                         goto success;
 671
 672                 case RXRPC_UPGRADEABLE_SERVICE:
 673                         ret = -EINVAL;
 674                         if (optlen != sizeof(service_upgrade) ||
 675                             rx->service_upgrade.from != 0)
 676                                 goto error;
 677                         ret = -EISCONN;
 678                         if (rx->sk.sk_state != RXRPC_SERVER_BOUND2)
 679                                 goto error;
 680                         ret = -EFAULT;
 681                         if (copy_from_user(service_upgrade, optval,
 682                                            sizeof(service_upgrade)) != 0)
 683                                 goto error;
 684                         ret = -EINVAL;
 685                         if ((service_upgrade[0] != rx->srx.srx_service ||
 686                              service_upgrade[1] != rx->second_service) &&
 687                             (service_upgrade[0] != rx->second_service ||
 688                              service_upgrade[1] != rx->srx.srx_service))
 689                                 goto error;
 690                         rx->service_upgrade.from = service_upgrade[0];
 691                         rx->service_upgrade.to = service_upgrade[1];
 692                         goto success;
 693
 694                 default:
 695                         break;
 696                 }
 697         }
 698
 699 success:
 700         ret = 0;
 701 error:
 702         release_sock(&rx->sk);
 703         return ret;
 704 }
 705
 706 /*
 707  * Get socket options.
 708  */
 709 static int rxrpc_getsockopt(struct socket *sock, int level, int optname,
 710                             char __user *optval, int __user *_optlen)
 711 {
 712         int optlen;
 713
 714         if (level != SOL_RXRPC)
 715                 return -EOPNOTSUPP;
 716
 717         if (get_user(optlen, _optlen))
 718                 return -EFAULT;
 719
 720         switch (optname) {
 721         case RXRPC_SUPPORTED_CMSG:
 722                 if (optlen < sizeof(int))
 723                         return -ETOOSMALL;
 724                 if (put_user(RXRPC__SUPPORTED - 1, (int __user *)optval) ||
 725                     put_user(sizeof(int), _optlen))
 726                         return -EFAULT;
 727                 return 0;
 728
 729         default:
 730                 return -EOPNOTSUPP;
 731         }
 732 }
 733
 734 /*
 735  * permit an RxRPC socket to be polled
 736  */
 737 static __poll_t rxrpc_poll(struct file *file, struct socket *sock,
 738                                poll_table *wait)
 739 {
 740         struct sock *sk = sock->sk;
 741         struct rxrpc_sock *rx = rxrpc_sk(sk);
 742         __poll_t mask;
 743
 744         sock_poll_wait(file, wait);
 745         mask = 0;
 746
 747         /* the socket is readable if there are any messages waiting on the Rx
 748          * queue */
 749         if (!list_empty(&rx->recvmsg_q))
 750                 mask |= EPOLLIN | EPOLLRDNORM;
 751
 752         /* the socket is writable if there is space to add new data to the
 753          * socket; there is no guarantee that any particular call in progress
 754          * on the socket may have space in the Tx ACK window */
 755         if (rxrpc_writable(sk))
 756                 mask |= EPOLLOUT | EPOLLWRNORM;
 757
 758         return mask;
 759 }
 760
 761 /*
 762  * create an RxRPC socket
 763  */
 764 static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
 765                         int kern)
 766 {
 767         struct rxrpc_net *rxnet;
 768         struct rxrpc_sock *rx;
 769         struct sock *sk;
 770
 771         _enter("%p,%d", sock, protocol);
 772
 773         /* we support transport protocol UDP/UDP6 only */
 774         if (protocol != PF_INET &&
 775             IS_ENABLED(CONFIG_AF_RXRPC_IPV6) && protocol != PF_INET6)
 776                 return -EPROTONOSUPPORT;
 777
 778         if (sock->type != SOCK_DGRAM)
 779                 return -ESOCKTNOSUPPORT;
 780
 781         sock->ops = &rxrpc_rpc_ops;
 782         sock->state = SS_UNCONNECTED;
 783
 784         sk = sk_alloc(net, PF_RXRPC, GFP_KERNEL, &rxrpc_proto, kern);
 785         if (!sk)
 786                 return -ENOMEM;
 787
 788         sock_init_data(sock, sk);
 789         sock_set_flag(sk, SOCK_RCU_FREE);
 790         sk->sk_state            = RXRPC_UNBOUND;
 791         sk->sk_write_space      = rxrpc_write_space;
 792         sk->sk_max_ack_backlog  = 0;
 793         sk->sk_destruct         = rxrpc_sock_destructor;
 794
 795         rx = rxrpc_sk(sk);
 796         rx->family = protocol;
 797         rx->calls = RB_ROOT;
 798
 799         spin_lock_init(&rx->incoming_lock);
 800         INIT_LIST_HEAD(&rx->sock_calls);
 801         INIT_LIST_HEAD(&rx->to_be_accepted);
 802         INIT_LIST_HEAD(&rx->recvmsg_q);
 803         rwlock_init(&rx->recvmsg_lock);
 804         rwlock_init(&rx->call_lock);
 805         memset(&rx->srx, 0, sizeof(rx->srx));
 806
 807         rxnet = rxrpc_net(sock_net(&rx->sk));
 808         timer_reduce(&rxnet->peer_keepalive_timer, jiffies + 1);
 809
 810         _leave(" = 0 [%p]", rx);
 811         return 0;
 812 }
 813
 814 /*
 815  * Kill all the calls on a socket and shut it down.
 816  */
 817 static int rxrpc_shutdown(struct socket *sock, int flags)
 818 {
 819         struct sock *sk = sock->sk;
 820         struct rxrpc_sock *rx = rxrpc_sk(sk);
 821         int ret = 0;
 822
 823         _enter("%p,%d", sk, flags);
 824
 825         if (flags != SHUT_RDWR)
 826                 return -EOPNOTSUPP;
 827         if (sk->sk_state == RXRPC_CLOSE)
 828                 return -ESHUTDOWN;
 829
 830         lock_sock(sk);
 831
 832         spin_lock_bh(&sk->sk_receive_queue.lock);
 833         if (sk->sk_state < RXRPC_CLOSE) {
 834                 sk->sk_state = RXRPC_CLOSE;
 835                 sk->sk_shutdown = SHUTDOWN_MASK;
 836         } else {
 837                 ret = -ESHUTDOWN;
 838         }
 839         spin_unlock_bh(&sk->sk_receive_queue.lock);
 840
 841         rxrpc_discard_prealloc(rx);
 842
 843         release_sock(sk);
 844         return ret;
 845 }
 846
 847 /*
 848  * RxRPC socket destructor
 849  */
 850 static void rxrpc_sock_destructor(struct sock *sk)
 851 {
 852         _enter("%p", sk);
 853
 854         rxrpc_purge_queue(&sk->sk_receive_queue);
 855
 856         WARN_ON(refcount_read(&sk->sk_wmem_alloc));
 857         WARN_ON(!sk_unhashed(sk));
 858         WARN_ON(sk->sk_socket);
 859
 860         if (!sock_flag(sk, SOCK_DEAD)) {
 861                 printk("Attempt to release alive rxrpc socket: %p\n", sk);
 862                 return;
 863         }
 864 }
 865
 866 /*
 867  * release an RxRPC socket
 868  */
 869 static int rxrpc_release_sock(struct sock *sk)
 870 {
 871         struct rxrpc_sock *rx = rxrpc_sk(sk);
 872         struct rxrpc_net *rxnet = rxrpc_net(sock_net(&rx->sk));
 873
 874         _enter("%p{%d,%d}", sk, sk->sk_state, refcount_read(&sk->sk_refcnt));
 875
 876         /* declare the socket closed for business */
 877         sock_orphan(sk);
 878         sk->sk_shutdown = SHUTDOWN_MASK;
 879
 880         /* We want to kill off all connections from a service socket
 881          * as fast as possible because we can't share these; client
 882          * sockets, on the other hand, can share an endpoint.
 883          */
 884         switch (sk->sk_state) {
 885         case RXRPC_SERVER_BOUND:
 886         case RXRPC_SERVER_BOUND2:
 887         case RXRPC_SERVER_LISTENING:
 888         case RXRPC_SERVER_LISTEN_DISABLED:
 889                 rx->local->service_closed = true;
 890                 break;
 891         }
 892
 893         spin_lock_bh(&sk->sk_receive_queue.lock);
 894         sk->sk_state = RXRPC_CLOSE;
 895         spin_unlock_bh(&sk->sk_receive_queue.lock);
 896
 897         if (rx->local && rcu_access_pointer(rx->local->service) == rx) {
 898                 write_lock(&rx->local->services_lock);
 899                 rcu_assign_pointer(rx->local->service, NULL);
 900                 write_unlock(&rx->local->services_lock);
 901         }
 902
 903         /* try to flush out this socket */
 904         rxrpc_discard_prealloc(rx);
 905         rxrpc_release_calls_on_socket(rx);
 906         flush_workqueue(rxrpc_workqueue);
 907         rxrpc_purge_queue(&sk->sk_receive_queue);
 908         rxrpc_queue_work(&rxnet->service_conn_reaper);
 909         rxrpc_queue_work(&rxnet->client_conn_reaper);
 910
 911         rxrpc_put_local(rx->local);
 912         rx->local = NULL;
 913         key_put(rx->key);
 914         rx->key = NULL;
 915         key_put(rx->securities);
 916         rx->securities = NULL;
 917         sock_put(sk);
 918
 919         _leave(" = 0");
 920         return 0;
 921 }
 922
 923 /*
 924  * release an RxRPC BSD socket on close() or equivalent
 925  */
 926 static int rxrpc_release(struct socket *sock)
 927 {
 928         struct sock *sk = sock->sk;
 929
 930         _enter("%p{%p}", sock, sk);
 931
 932         if (!sk)
 933                 return 0;
 934
 935         sock->sk = NULL;
 936
 937         return rxrpc_release_sock(sk);
 938 }
 939
 940 /*
 941  * RxRPC network protocol
 942  */
 943 static const struct proto_ops rxrpc_rpc_ops = {
 944         .family         = PF_RXRPC,
 945         .owner          = THIS_MODULE,
 946         .release        = rxrpc_release,
 947         .bind           = rxrpc_bind,
 948         .connect        = rxrpc_connect,
 949         .socketpair     = sock_no_socketpair,
 950         .accept         = sock_no_accept,
 951         .getname        = sock_no_getname,
 952         .poll           = rxrpc_poll,
 953         .ioctl          = sock_no_ioctl,
 954         .listen         = rxrpc_listen,
 955         .shutdown       = rxrpc_shutdown,
 956         .setsockopt     = rxrpc_setsockopt,
 957         .getsockopt     = rxrpc_getsockopt,
 958         .sendmsg        = rxrpc_sendmsg,
 959         .recvmsg        = rxrpc_recvmsg,
 960         .mmap           = sock_no_mmap,
 961         .sendpage       = sock_no_sendpage,
 962 };
 963
 964 static struct proto rxrpc_proto = {
 965         .name           = "RXRPC",
 966         .owner          = THIS_MODULE,
 967         .obj_size       = sizeof(struct rxrpc_sock),
 968         .max_header     = sizeof(struct rxrpc_wire_header),
 969 };
 970
 971 static const struct net_proto_family rxrpc_family_ops = {
 972         .family = PF_RXRPC,
 973         .create = rxrpc_create,
 974         .owner  = THIS_MODULE,
 975 };
 976
 977 /*
 978  * initialise and register the RxRPC protocol
 979  */
 980 static int __init af_rxrpc_init(void)
 981 {
 982         int ret = -1;
 983         unsigned int tmp;
 984
 985         BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > FIELD_SIZEOF(struct sk_buff, cb));
 986
 987         get_random_bytes(&tmp, sizeof(tmp));
 988         tmp &= 0x3fffffff;
 989         if (tmp == 0)
 990                 tmp = 1;
 991         idr_set_cursor(&rxrpc_client_conn_ids, tmp);
 992
 993         ret = -ENOMEM;
 994         rxrpc_call_jar = kmem_cache_create(
 995                 "rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
 996                 SLAB_HWCACHE_ALIGN, NULL);
 997         if (!rxrpc_call_jar) {
 998                 pr_notice("Failed to allocate call jar\n");
 999                 goto error_call_jar;
1000         }
1001
1002         rxrpc_workqueue = alloc_workqueue("krxrpcd", 0, 1);
1003         if (!rxrpc_workqueue) {
1004                 pr_notice("Failed to allocate work queue\n");
1005                 goto error_work_queue;
1006         }
1007
1008         ret = rxrpc_init_security();
1009         if (ret < 0) {
1010                 pr_crit("Cannot initialise security\n");
1011                 goto error_security;
1012         }
1013
1014         ret = register_pernet_subsys(&rxrpc_net_ops);
1015         if (ret)
1016                 goto error_pernet;
1017
1018         ret = proto_register(&rxrpc_proto, 1);
1019         if (ret < 0) {
1020                 pr_crit("Cannot register protocol\n");
1021                 goto error_proto;
1022         }
1023
1024         ret = sock_register(&rxrpc_family_ops);
1025         if (ret < 0) {
1026                 pr_crit("Cannot register socket family\n");
1027                 goto error_sock;
1028         }
1029
1030         ret = register_key_type(&key_type_rxrpc);
1031         if (ret < 0) {
1032                 pr_crit("Cannot register client key type\n");
1033                 goto error_key_type;
1034         }
1035
1036         ret = register_key_type(&key_type_rxrpc_s);
1037         if (ret < 0) {
1038                 pr_crit("Cannot register server key type\n");
1039                 goto error_key_type_s;
1040         }
1041
1042         ret = rxrpc_sysctl_init();
1043         if (ret < 0) {
1044                 pr_crit("Cannot register sysctls\n");
1045                 goto error_sysctls;
1046         }
1047
1048         return 0;
1049
1050 error_sysctls:
1051         unregister_key_type(&key_type_rxrpc_s);
1052 error_key_type_s:
1053         unregister_key_type(&key_type_rxrpc);
1054 error_key_type:
1055         sock_unregister(PF_RXRPC);
1056 error_sock:
1057         proto_unregister(&rxrpc_proto);
1058 error_proto:
1059         unregister_pernet_subsys(&rxrpc_net_ops);
1060 error_pernet:
1061         rxrpc_exit_security();
1062 error_security:
1063         destroy_workqueue(rxrpc_workqueue);
1064 error_work_queue:
1065         kmem_cache_destroy(rxrpc_call_jar);
1066 error_call_jar:
1067         return ret;
1068 }
1069
1070 /*
1071  * unregister the RxRPC protocol
1072  */
1073 static void __exit af_rxrpc_exit(void)
1074 {
1075         _enter("");
1076         rxrpc_sysctl_exit();
1077         unregister_key_type(&key_type_rxrpc_s);
1078         unregister_key_type(&key_type_rxrpc);
1079         sock_unregister(PF_RXRPC);
1080         proto_unregister(&rxrpc_proto);
1081         unregister_pernet_subsys(&rxrpc_net_ops);
1082         ASSERTCMP(atomic_read(&rxrpc_n_tx_skbs), ==, 0);
1083         ASSERTCMP(atomic_read(&rxrpc_n_rx_skbs), ==, 0);
1084
1085         /* Make sure the local and peer records pinned by any dying connections
1086          * are released.
1087          */
1088         rcu_barrier();
1089         rxrpc_destroy_client_conn_ids();
1090
1091         destroy_workqueue(rxrpc_workqueue);
1092         rxrpc_exit_security();
1093         kmem_cache_destroy(rxrpc_call_jar);
1094         _leave("");
1095 }
1096
1097 module_init(af_rxrpc_init);
1098 module_exit(af_rxrpc_exit);