net/rxrpc/af_rxrpc.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /* AF_RXRPC implementation
   3  *
   4  * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
   5  * Written by David Howells (dhowells@redhat.com)
   6  */
   7
   8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   9
  10 #include <linux/module.h>
  11 #include <linux/kernel.h>
  12 #include <linux/net.h>
  13 #include <linux/slab.h>
  14 #include <linux/skbuff.h>
  15 #include <linux/random.h>
  16 #include <linux/poll.h>
  17 #include <linux/proc_fs.h>
  18 #include <linux/key-type.h>
  19 #include <net/net_namespace.h>
  20 #include <net/sock.h>
  21 #include <net/af_rxrpc.h>
  22 #define CREATE_TRACE_POINTS
  23 #include "ar-internal.h"
  24
  25 MODULE_DESCRIPTION("RxRPC network protocol");
  26 MODULE_AUTHOR("Red Hat, Inc.");
  27 MODULE_LICENSE("GPL");
  28 MODULE_ALIAS_NETPROTO(PF_RXRPC);
  29
  30 unsigned int rxrpc_debug; // = RXRPC_DEBUG_KPROTO;
  31 module_param_named(debug, rxrpc_debug, uint, 0644);
  32 MODULE_PARM_DESC(debug, "RxRPC debugging mask");
  33
  34 static struct proto rxrpc_proto;
  35 static const struct proto_ops rxrpc_rpc_ops;
  36
  37 /* current debugging ID */
  38 atomic_t rxrpc_debug_id;
  39 EXPORT_SYMBOL(rxrpc_debug_id);
  40
  41 /* count of skbs currently in use */
  42 atomic_t rxrpc_n_tx_skbs, rxrpc_n_rx_skbs;
  43
  44 struct workqueue_struct *rxrpc_workqueue;
  45
  46 static void rxrpc_sock_destructor(struct sock *);
  47
  48 /*
  49  * see if an RxRPC socket is currently writable
  50  */
  51 static inline int rxrpc_writable(struct sock *sk)
  52 {
  53         return refcount_read(&sk->sk_wmem_alloc) < (size_t) sk->sk_sndbuf;
  54 }
  55
  56 /*
  57  * wait for write bufferage to become available
  58  */
  59 static void rxrpc_write_space(struct sock *sk)
  60 {
  61         _enter("%p", sk);
  62         rcu_read_lock();
  63         if (rxrpc_writable(sk)) {
  64                 struct socket_wq *wq = rcu_dereference(sk->sk_wq);
  65
  66                 if (skwq_has_sleeper(wq))
  67                         wake_up_interruptible(&wq->wait);
  68                 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
  69         }
  70         rcu_read_unlock();
  71 }
  72
  73 /*
  74  * validate an RxRPC address
  75  */
  76 static int rxrpc_validate_address(struct rxrpc_sock *rx,
  77                                   struct sockaddr_rxrpc *srx,
  78                                   int len)
  79 {
  80         unsigned int tail;
  81
  82         if (len < sizeof(struct sockaddr_rxrpc))
  83                 return -EINVAL;
  84
  85         if (srx->srx_family != AF_RXRPC)
  86                 return -EAFNOSUPPORT;
  87
  88         if (srx->transport_type != SOCK_DGRAM)
  89                 return -ESOCKTNOSUPPORT;
  90
  91         len -= offsetof(struct sockaddr_rxrpc, transport);
  92         if (srx->transport_len < sizeof(sa_family_t) ||
  93             srx->transport_len > len)
  94                 return -EINVAL;
  95
  96         if (srx->transport.family != rx->family &&
  97             srx->transport.family == AF_INET && rx->family != AF_INET6)
  98                 return -EAFNOSUPPORT;
  99
 100         switch (srx->transport.family) {
 101         case AF_INET:
 102                 if (srx->transport_len < sizeof(struct sockaddr_in))
 103                         return -EINVAL;
 104                 tail = offsetof(struct sockaddr_rxrpc, transport.sin.__pad);
 105                 break;
 106
 107 #ifdef CONFIG_AF_RXRPC_IPV6
 108         case AF_INET6:
 109                 if (srx->transport_len < sizeof(struct sockaddr_in6))
 110                         return -EINVAL;
 111                 tail = offsetof(struct sockaddr_rxrpc, transport) +
 112                         sizeof(struct sockaddr_in6);
 113                 break;
 114 #endif
 115
 116         default:
 117                 return -EAFNOSUPPORT;
 118         }
 119
 120         if (tail < len)
 121                 memset((void *)srx + tail, 0, len - tail);
 122         _debug("INET: %pISp", &srx->transport);
 123         return 0;
 124 }
 125
 126 /*
 127  * bind a local address to an RxRPC socket
 128  */
 129 static int rxrpc_bind(struct socket *sock, struct sockaddr *saddr, int len)
 130 {
 131         struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)saddr;
 132         struct rxrpc_local *local;
 133         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 134         u16 service_id;
 135         int ret;
 136
 137         _enter("%p,%p,%d", rx, saddr, len);
 138
 139         ret = rxrpc_validate_address(rx, srx, len);
 140         if (ret < 0)
 141                 goto error;
 142         service_id = srx->srx_service;
 143
 144         lock_sock(&rx->sk);
 145
 146         switch (rx->sk.sk_state) {
 147         case RXRPC_UNBOUND:
 148                 rx->srx = *srx;
 149                 local = rxrpc_lookup_local(sock_net(&rx->sk), &rx->srx);
 150                 if (IS_ERR(local)) {
 151                         ret = PTR_ERR(local);
 152                         goto error_unlock;
 153                 }
 154
 155                 if (service_id) {
 156                         write_lock(&local->services_lock);
 157                         if (rcu_access_pointer(local->service))
 158                                 goto service_in_use;
 159                         rx->local = local;
 160                         rcu_assign_pointer(local->service, rx);
 161                         write_unlock(&local->services_lock);
 162
 163                         rx->sk.sk_state = RXRPC_SERVER_BOUND;
 164                 } else {
 165                         rx->local = local;
 166                         rx->sk.sk_state = RXRPC_CLIENT_BOUND;
 167                 }
 168                 break;
 169
 170         case RXRPC_SERVER_BOUND:
 171                 ret = -EINVAL;
 172                 if (service_id == 0)
 173                         goto error_unlock;
 174                 ret = -EADDRINUSE;
 175                 if (service_id == rx->srx.srx_service)
 176                         goto error_unlock;
 177                 ret = -EINVAL;
 178                 srx->srx_service = rx->srx.srx_service;
 179                 if (memcmp(srx, &rx->srx, sizeof(*srx)) != 0)
 180                         goto error_unlock;
 181                 rx->second_service = service_id;
 182                 rx->sk.sk_state = RXRPC_SERVER_BOUND2;
 183                 break;
 184
 185         default:
 186                 ret = -EINVAL;
 187                 goto error_unlock;
 188         }
 189
 190         release_sock(&rx->sk);
 191         _leave(" = 0");
 192         return 0;
 193
 194 service_in_use:
 195         write_unlock(&local->services_lock);
 196         rxrpc_unuse_local(local);
 197         rxrpc_put_local(local);
 198         ret = -EADDRINUSE;
 199 error_unlock:
 200         release_sock(&rx->sk);
 201 error:
 202         _leave(" = %d", ret);
 203         return ret;
 204 }
 205
 206 /*
 207  * set the number of pending calls permitted on a listening socket
 208  */
 209 static int rxrpc_listen(struct socket *sock, int backlog)
 210 {
 211         struct sock *sk = sock->sk;
 212         struct rxrpc_sock *rx = rxrpc_sk(sk);
 213         unsigned int max, old;
 214         int ret;
 215
 216         _enter("%p,%d", rx, backlog);
 217
 218         lock_sock(&rx->sk);
 219
 220         switch (rx->sk.sk_state) {
 221         case RXRPC_UNBOUND:
 222                 ret = -EADDRNOTAVAIL;
 223                 break;
 224         case RXRPC_SERVER_BOUND:
 225         case RXRPC_SERVER_BOUND2:
 226                 ASSERT(rx->local != NULL);
 227                 max = READ_ONCE(rxrpc_max_backlog);
 228                 ret = -EINVAL;
 229                 if (backlog == INT_MAX)
 230                         backlog = max;
 231                 else if (backlog < 0 || backlog > max)
 232                         break;
 233                 old = sk->sk_max_ack_backlog;
 234                 sk->sk_max_ack_backlog = backlog;
 235                 ret = rxrpc_service_prealloc(rx, GFP_KERNEL);
 236                 if (ret == 0)
 237                         rx->sk.sk_state = RXRPC_SERVER_LISTENING;
 238                 else
 239                         sk->sk_max_ack_backlog = old;
 240                 break;
 241         case RXRPC_SERVER_LISTENING:
 242                 if (backlog == 0) {
 243                         rx->sk.sk_state = RXRPC_SERVER_LISTEN_DISABLED;
 244                         sk->sk_max_ack_backlog = 0;
 245                         rxrpc_discard_prealloc(rx);
 246                         ret = 0;
 247                         break;
 248                 }
 249                 fallthrough;
 250         default:
 251                 ret = -EBUSY;
 252                 break;
 253         }
 254
 255         release_sock(&rx->sk);
 256         _leave(" = %d", ret);
 257         return ret;
 258 }
 259
 260 /**
 261  * rxrpc_kernel_begin_call - Allow a kernel service to begin a call
 262  * @sock: The socket on which to make the call
 263  * @srx: The address of the peer to contact
 264  * @key: The security context to use (defaults to socket setting)
 265  * @user_call_ID: The ID to use
 266  * @tx_total_len: Total length of data to transmit during the call (or -1)
 267  * @gfp: The allocation constraints
 268  * @notify_rx: Where to send notifications instead of socket queue
 269  * @upgrade: Request service upgrade for call
 270  * @interruptibility: The call is interruptible, or can be canceled.
 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                                            enum rxrpc_interruptibility interruptibility,
 289                                            unsigned int debug_id)
 290 {
 291         struct rxrpc_conn_parameters cp;
 292         struct rxrpc_call_params p;
 293         struct rxrpc_call *call;
 294         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 295         int ret;
 296
 297         _enter(",,%x,%lx", key_serial(key), user_call_ID);
 298
 299         ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
 300         if (ret < 0)
 301                 return ERR_PTR(ret);
 302
 303         lock_sock(&rx->sk);
 304
 305         if (!key)
 306                 key = rx->key;
 307         if (key && !key->payload.data[0])
 308                 key = NULL; /* a no-security key */
 309
 310         memset(&p, 0, sizeof(p));
 311         p.user_call_ID          = user_call_ID;
 312         p.tx_total_len          = tx_total_len;
 313         p.interruptibility      = interruptibility;
 314         p.kernel                = true;
 315
 316         memset(&cp, 0, sizeof(cp));
 317         cp.local                = rx->local;
 318         cp.key                  = key;
 319         cp.security_level       = rx->min_sec_level;
 320         cp.exclusive            = false;
 321         cp.upgrade              = upgrade;
 322         cp.service_id           = srx->srx_service;
 323         call = rxrpc_new_client_call(rx, &cp, srx, &p, gfp, debug_id);
 324         /* The socket has been unlocked. */
 325         if (!IS_ERR(call)) {
 326                 call->notify_rx = notify_rx;
 327                 mutex_unlock(&call->user_mutex);
 328         }
 329
 330         rxrpc_put_peer(cp.peer);
 331         _leave(" = %p", call);
 332         return call;
 333 }
 334 EXPORT_SYMBOL(rxrpc_kernel_begin_call);
 335
 336 /*
 337  * Dummy function used to stop the notifier talking to recvmsg().
 338  */
 339 static void rxrpc_dummy_notify_rx(struct sock *sk, struct rxrpc_call *rxcall,
 340                                   unsigned long call_user_ID)
 341 {
 342 }
 343
 344 /**
 345  * rxrpc_kernel_end_call - Allow a kernel service to end a call it was using
 346  * @sock: The socket the call is on
 347  * @call: The call to end
 348  *
 349  * Allow a kernel service to end a call it was using.  The call must be
 350  * complete before this is called (the call should be aborted if necessary).
 351  */
 352 void rxrpc_kernel_end_call(struct socket *sock, struct rxrpc_call *call)
 353 {
 354         _enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
 355
 356         mutex_lock(&call->user_mutex);
 357         rxrpc_release_call(rxrpc_sk(sock->sk), call);
 358
 359         /* Make sure we're not going to call back into a kernel service */
 360         if (call->notify_rx) {
 361                 spin_lock_bh(&call->notify_lock);
 362                 call->notify_rx = rxrpc_dummy_notify_rx;
 363                 spin_unlock_bh(&call->notify_lock);
 364         }
 365
 366         mutex_unlock(&call->user_mutex);
 367         rxrpc_put_call(call, rxrpc_call_put_kernel);
 368 }
 369 EXPORT_SYMBOL(rxrpc_kernel_end_call);
 370
 371 /**
 372  * rxrpc_kernel_check_life - Check to see whether a call is still alive
 373  * @sock: The socket the call is on
 374  * @call: The call to check
 375  *
 376  * Allow a kernel service to find out whether a call is still alive -
 377  * ie. whether it has completed.
 378  */
 379 bool rxrpc_kernel_check_life(const struct socket *sock,
 380                              const struct rxrpc_call *call)
 381 {
 382         return call->state != RXRPC_CALL_COMPLETE;
 383 }
 384 EXPORT_SYMBOL(rxrpc_kernel_check_life);
 385
 386 /**
 387  * rxrpc_kernel_get_epoch - Retrieve the epoch value from a call.
 388  * @sock: The socket the call is on
 389  * @call: The call to query
 390  *
 391  * Allow a kernel service to retrieve the epoch value from a service call to
 392  * see if the client at the other end rebooted.
 393  */
 394 u32 rxrpc_kernel_get_epoch(struct socket *sock, struct rxrpc_call *call)
 395 {
 396         return call->conn->proto.epoch;
 397 }
 398 EXPORT_SYMBOL(rxrpc_kernel_get_epoch);
 399
 400 /**
 401  * rxrpc_kernel_new_call_notification - Get notifications of new calls
 402  * @sock: The socket to intercept received messages on
 403  * @notify_new_call: Function to be called when new calls appear
 404  * @discard_new_call: Function to discard preallocated calls
 405  *
 406  * Allow a kernel service to be given notifications about new calls.
 407  */
 408 void rxrpc_kernel_new_call_notification(
 409         struct socket *sock,
 410         rxrpc_notify_new_call_t notify_new_call,
 411         rxrpc_discard_new_call_t discard_new_call)
 412 {
 413         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 414
 415         rx->notify_new_call = notify_new_call;
 416         rx->discard_new_call = discard_new_call;
 417 }
 418 EXPORT_SYMBOL(rxrpc_kernel_new_call_notification);
 419
 420 /**
 421  * rxrpc_kernel_set_max_life - Set maximum lifespan on a call
 422  * @sock: The socket the call is on
 423  * @call: The call to configure
 424  * @hard_timeout: The maximum lifespan of the call in jiffies
 425  *
 426  * Set the maximum lifespan of a call.  The call will end with ETIME or
 427  * ETIMEDOUT if it takes longer than this.
 428  */
 429 void rxrpc_kernel_set_max_life(struct socket *sock, struct rxrpc_call *call,
 430                                unsigned long hard_timeout)
 431 {
 432         unsigned long now;
 433
 434         mutex_lock(&call->user_mutex);
 435
 436         now = jiffies;
 437         hard_timeout += now;
 438         WRITE_ONCE(call->expect_term_by, hard_timeout);
 439         rxrpc_reduce_call_timer(call, hard_timeout, now, rxrpc_timer_set_for_hard);
 440
 441         mutex_unlock(&call->user_mutex);
 442 }
 443 EXPORT_SYMBOL(rxrpc_kernel_set_max_life);
 444
 445 /*
 446  * connect an RxRPC socket
 447  * - this just targets it at a specific destination; no actual connection
 448  *   negotiation takes place
 449  */
 450 static int rxrpc_connect(struct socket *sock, struct sockaddr *addr,
 451                          int addr_len, int flags)
 452 {
 453         struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
 454         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 455         int ret;
 456
 457         _enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
 458
 459         ret = rxrpc_validate_address(rx, srx, addr_len);
 460         if (ret < 0) {
 461                 _leave(" = %d [bad addr]", ret);
 462                 return ret;
 463         }
 464
 465         lock_sock(&rx->sk);
 466
 467         ret = -EISCONN;
 468         if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
 469                 goto error;
 470
 471         switch (rx->sk.sk_state) {
 472         case RXRPC_UNBOUND:
 473                 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
 474         case RXRPC_CLIENT_UNBOUND:
 475         case RXRPC_CLIENT_BOUND:
 476                 break;
 477         default:
 478                 ret = -EBUSY;
 479                 goto error;
 480         }
 481
 482         rx->connect_srx = *srx;
 483         set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
 484         ret = 0;
 485
 486 error:
 487         release_sock(&rx->sk);
 488         return ret;
 489 }
 490
 491 /*
 492  * send a message through an RxRPC socket
 493  * - in a client this does a number of things:
 494  *   - finds/sets up a connection for the security specified (if any)
 495  *   - initiates a call (ID in control data)
 496  *   - ends the request phase of a call (if MSG_MORE is not set)
 497  *   - sends a call data packet
 498  *   - may send an abort (abort code in control data)
 499  */
 500 static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
 501 {
 502         struct rxrpc_local *local;
 503         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 504         int ret;
 505
 506         _enter(",{%d},,%zu", rx->sk.sk_state, len);
 507
 508         if (m->msg_flags & MSG_OOB)
 509                 return -EOPNOTSUPP;
 510
 511         if (m->msg_name) {
 512                 ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
 513                 if (ret < 0) {
 514                         _leave(" = %d [bad addr]", ret);
 515                         return ret;
 516                 }
 517         }
 518
 519         lock_sock(&rx->sk);
 520
 521         switch (rx->sk.sk_state) {
 522         case RXRPC_UNBOUND:
 523         case RXRPC_CLIENT_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_BOUND;
 549                 fallthrough;
 550
 551         case RXRPC_CLIENT_BOUND:
 552                 if (!m->msg_name &&
 553                     test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
 554                         m->msg_name = &rx->connect_srx;
 555                         m->msg_namelen = sizeof(rx->connect_srx);
 556                 }
 557                 fallthrough;
 558         case RXRPC_SERVER_BOUND:
 559         case RXRPC_SERVER_LISTENING:
 560                 ret = rxrpc_do_sendmsg(rx, m, len);
 561                 /* The socket has been unlocked */
 562                 goto out;
 563         default:
 564                 ret = -EINVAL;
 565                 goto error_unlock;
 566         }
 567
 568 error_unlock:
 569         release_sock(&rx->sk);
 570 out:
 571         _leave(" = %d", ret);
 572         return ret;
 573 }
 574
 575 int rxrpc_sock_set_min_security_level(struct sock *sk, unsigned int val)
 576 {
 577         if (sk->sk_state != RXRPC_UNBOUND)
 578                 return -EISCONN;
 579         if (val > RXRPC_SECURITY_MAX)
 580                 return -EINVAL;
 581         lock_sock(sk);
 582         rxrpc_sk(sk)->min_sec_level = val;
 583         release_sock(sk);
 584         return 0;
 585 }
 586 EXPORT_SYMBOL(rxrpc_sock_set_min_security_level);
 587
 588 /*
 589  * set RxRPC socket options
 590  */
 591 static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
 592                             sockptr_t optval, unsigned int optlen)
 593 {
 594         struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 595         unsigned int min_sec_level;
 596         u16 service_upgrade[2];
 597         int ret;
 598
 599         _enter(",%d,%d,,%d", level, optname, optlen);
 600
 601         lock_sock(&rx->sk);
 602         ret = -EOPNOTSUPP;
 603
 604         if (level == SOL_RXRPC) {
 605                 switch (optname) {
 606                 case RXRPC_EXCLUSIVE_CONNECTION:
 607                         ret = -EINVAL;
 608                         if (optlen != 0)
 609                                 goto error;
 610                         ret = -EISCONN;
 611                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 612                                 goto error;
 613                         rx->exclusive = true;
 614                         goto success;
 615
 616                 case RXRPC_SECURITY_KEY:
 617                         ret = -EINVAL;
 618                         if (rx->key)
 619                                 goto error;
 620                         ret = -EISCONN;
 621                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 622                                 goto error;
 623                         ret = rxrpc_request_key(rx, optval, optlen);
 624                         goto error;
 625
 626                 case RXRPC_SECURITY_KEYRING:
 627                         ret = -EINVAL;
 628                         if (rx->key)
 629                                 goto error;
 630                         ret = -EISCONN;
 631                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 632                                 goto error;
 633                         ret = rxrpc_server_keyring(rx, optval, optlen);
 634                         goto error;
 635
 636                 case RXRPC_MIN_SECURITY_LEVEL:
 637                         ret = -EINVAL;
 638                         if (optlen != sizeof(unsigned int))
 639                                 goto error;
 640                         ret = -EISCONN;
 641                         if (rx->sk.sk_state != RXRPC_UNBOUND)
 642                                 goto error;
 643                         ret = copy_from_sockptr(&min_sec_level, optval,
 644                                        sizeof(unsigned int));
 645                         if (ret < 0)
 646                                 goto error;
 647                         ret = -EINVAL;
 648                         if (min_sec_level > RXRPC_SECURITY_MAX)
 649                                 goto error;
 650                         rx->min_sec_level = min_sec_level;
 651                         goto success;
 652
 653                 case RXRPC_UPGRADEABLE_SERVICE:
 654                         ret = -EINVAL;
 655                         if (optlen != sizeof(service_upgrade) ||
 656                             rx->service_upgrade.from != 0)
 657                                 goto error;
 658                         ret = -EISCONN;
 659                         if (rx->sk.sk_state != RXRPC_SERVER_BOUND2)
 660                                 goto error;
 661                         ret = -EFAULT;
 662                         if (copy_from_sockptr(service_upgrade, optval,
 663                                            sizeof(service_upgrade)) != 0)
 664                                 goto error;
 665                         ret = -EINVAL;
 666                         if ((service_upgrade[0] != rx->srx.srx_service ||
 667                              service_upgrade[1] != rx->second_service) &&
 668                             (service_upgrade[0] != rx->second_service ||
 669                              service_upgrade[1] != rx->srx.srx_service))
 670                                 goto error;
 671                         rx->service_upgrade.from = service_upgrade[0];
 672                         rx->service_upgrade.to = service_upgrade[1];
 673                         goto success;
 674
 675                 default:
 676                         break;
 677                 }
 678         }
 679
 680 success:
 681         ret = 0;
 682 error:
 683         release_sock(&rx->sk);
 684         return ret;
 685 }
 686
 687 /*
 688  * Get socket options.
 689  */
 690 static int rxrpc_getsockopt(struct socket *sock, int level, int optname,
 691                             char __user *optval, int __user *_optlen)
 692 {
 693         int optlen;
 694
 695         if (level != SOL_RXRPC)
 696                 return -EOPNOTSUPP;
 697
 698         if (get_user(optlen, _optlen))
 699                 return -EFAULT;
 700
 701         switch (optname) {
 702         case RXRPC_SUPPORTED_CMSG:
 703                 if (optlen < sizeof(int))
 704                         return -ETOOSMALL;
 705                 if (put_user(RXRPC__SUPPORTED - 1, (int __user *)optval) ||
 706                     put_user(sizeof(int), _optlen))
 707                         return -EFAULT;
 708                 return 0;
 709
 710         default:
 711                 return -EOPNOTSUPP;
 712         }
 713 }
 714
 715 /*
 716  * permit an RxRPC socket to be polled
 717  */
 718 static __poll_t rxrpc_poll(struct file *file, struct socket *sock,
 719                                poll_table *wait)
 720 {
 721         struct sock *sk = sock->sk;
 722         struct rxrpc_sock *rx = rxrpc_sk(sk);
 723         __poll_t mask;
 724
 725         sock_poll_wait(file, sock, wait);
 726         mask = 0;
 727
 728         /* the socket is readable if there are any messages waiting on the Rx
 729          * queue */
 730         if (!list_empty(&rx->recvmsg_q))
 731                 mask |= EPOLLIN | EPOLLRDNORM;
 732
 733         /* the socket is writable if there is space to add new data to the
 734          * socket; there is no guarantee that any particular call in progress
 735          * on the socket may have space in the Tx ACK window */
 736         if (rxrpc_writable(sk))
 737                 mask |= EPOLLOUT | EPOLLWRNORM;
 738
 739         return mask;
 740 }
 741
 742 /*
 743  * create an RxRPC socket
 744  */
 745 static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
 746                         int kern)
 747 {
 748         struct rxrpc_net *rxnet;
 749         struct rxrpc_sock *rx;
 750         struct sock *sk;
 751
 752         _enter("%p,%d", sock, protocol);
 753
 754         /* we support transport protocol UDP/UDP6 only */
 755         if (protocol != PF_INET &&
 756             IS_ENABLED(CONFIG_AF_RXRPC_IPV6) && protocol != PF_INET6)
 757                 return -EPROTONOSUPPORT;
 758
 759         if (sock->type != SOCK_DGRAM)
 760                 return -ESOCKTNOSUPPORT;
 761
 762         sock->ops = &rxrpc_rpc_ops;
 763         sock->state = SS_UNCONNECTED;
 764
 765         sk = sk_alloc(net, PF_RXRPC, GFP_KERNEL, &rxrpc_proto, kern);
 766         if (!sk)
 767                 return -ENOMEM;
 768
 769         sock_init_data(sock, sk);
 770         sock_set_flag(sk, SOCK_RCU_FREE);
 771         sk->sk_state            = RXRPC_UNBOUND;
 772         sk->sk_write_space      = rxrpc_write_space;
 773         sk->sk_max_ack_backlog  = 0;
 774         sk->sk_destruct         = rxrpc_sock_destructor;
 775
 776         rx = rxrpc_sk(sk);
 777         rx->family = protocol;
 778         rx->calls = RB_ROOT;
 779
 780         spin_lock_init(&rx->incoming_lock);
 781         INIT_LIST_HEAD(&rx->sock_calls);
 782         INIT_LIST_HEAD(&rx->to_be_accepted);
 783         INIT_LIST_HEAD(&rx->recvmsg_q);
 784         rwlock_init(&rx->recvmsg_lock);
 785         rwlock_init(&rx->call_lock);
 786         memset(&rx->srx, 0, sizeof(rx->srx));
 787
 788         rxnet = rxrpc_net(sock_net(&rx->sk));
 789         timer_reduce(&rxnet->peer_keepalive_timer, jiffies + 1);
 790
 791         _leave(" = 0 [%p]", rx);
 792         return 0;
 793 }
 794
 795 /*
 796  * Kill all the calls on a socket and shut it down.
 797  */
 798 static int rxrpc_shutdown(struct socket *sock, int flags)
 799 {
 800         struct sock *sk = sock->sk;
 801         struct rxrpc_sock *rx = rxrpc_sk(sk);
 802         int ret = 0;
 803
 804         _enter("%p,%d", sk, flags);
 805
 806         if (flags != SHUT_RDWR)
 807                 return -EOPNOTSUPP;
 808         if (sk->sk_state == RXRPC_CLOSE)
 809                 return -ESHUTDOWN;
 810
 811         lock_sock(sk);
 812
 813         spin_lock_bh(&sk->sk_receive_queue.lock);
 814         if (sk->sk_state < RXRPC_CLOSE) {
 815                 sk->sk_state = RXRPC_CLOSE;
 816                 sk->sk_shutdown = SHUTDOWN_MASK;
 817         } else {
 818                 ret = -ESHUTDOWN;
 819         }
 820         spin_unlock_bh(&sk->sk_receive_queue.lock);
 821
 822         rxrpc_discard_prealloc(rx);
 823
 824         release_sock(sk);
 825         return ret;
 826 }
 827
 828 /*
 829  * RxRPC socket destructor
 830  */
 831 static void rxrpc_sock_destructor(struct sock *sk)
 832 {
 833         _enter("%p", sk);
 834
 835         rxrpc_purge_queue(&sk->sk_receive_queue);
 836
 837         WARN_ON(refcount_read(&sk->sk_wmem_alloc));
 838         WARN_ON(!sk_unhashed(sk));
 839         WARN_ON(sk->sk_socket);
 840
 841         if (!sock_flag(sk, SOCK_DEAD)) {
 842                 printk("Attempt to release alive rxrpc socket: %p\n", sk);
 843                 return;
 844         }
 845 }
 846
 847 /*
 848  * release an RxRPC socket
 849  */
 850 static int rxrpc_release_sock(struct sock *sk)
 851 {
 852         struct rxrpc_sock *rx = rxrpc_sk(sk);
 853
 854         _enter("%p{%d,%d}", sk, sk->sk_state, refcount_read(&sk->sk_refcnt));
 855
 856         /* declare the socket closed for business */
 857         sock_orphan(sk);
 858         sk->sk_shutdown = SHUTDOWN_MASK;
 859
 860         /* We want to kill off all connections from a service socket
 861          * as fast as possible because we can't share these; client
 862          * sockets, on the other hand, can share an endpoint.
 863          */
 864         switch (sk->sk_state) {
 865         case RXRPC_SERVER_BOUND:
 866         case RXRPC_SERVER_BOUND2:
 867         case RXRPC_SERVER_LISTENING:
 868         case RXRPC_SERVER_LISTEN_DISABLED:
 869                 rx->local->service_closed = true;
 870                 break;
 871         }
 872
 873         spin_lock_bh(&sk->sk_receive_queue.lock);
 874         sk->sk_state = RXRPC_CLOSE;
 875         spin_unlock_bh(&sk->sk_receive_queue.lock);
 876
 877         if (rx->local && rcu_access_pointer(rx->local->service) == rx) {
 878                 write_lock(&rx->local->services_lock);
 879                 rcu_assign_pointer(rx->local->service, NULL);
 880                 write_unlock(&rx->local->services_lock);
 881         }
 882
 883         /* try to flush out this socket */
 884         rxrpc_discard_prealloc(rx);
 885         rxrpc_release_calls_on_socket(rx);
 886         flush_workqueue(rxrpc_workqueue);
 887         rxrpc_purge_queue(&sk->sk_receive_queue);
 888
 889         rxrpc_unuse_local(rx->local);
 890         rxrpc_put_local(rx->local);
 891         rx->local = NULL;
 892         key_put(rx->key);
 893         rx->key = NULL;
 894         key_put(rx->securities);
 895         rx->securities = NULL;
 896         sock_put(sk);
 897
 898         _leave(" = 0");
 899         return 0;
 900 }
 901
 902 /*
 903  * release an RxRPC BSD socket on close() or equivalent
 904  */
 905 static int rxrpc_release(struct socket *sock)
 906 {
 907         struct sock *sk = sock->sk;
 908
 909         _enter("%p{%p}", sock, sk);
 910
 911         if (!sk)
 912                 return 0;
 913
 914         sock->sk = NULL;
 915
 916         return rxrpc_release_sock(sk);
 917 }
 918
 919 /*
 920  * RxRPC network protocol
 921  */
 922 static const struct proto_ops rxrpc_rpc_ops = {
 923         .family         = PF_RXRPC,
 924         .owner          = THIS_MODULE,
 925         .release        = rxrpc_release,
 926         .bind           = rxrpc_bind,
 927         .connect        = rxrpc_connect,
 928         .socketpair     = sock_no_socketpair,
 929         .accept         = sock_no_accept,
 930         .getname        = sock_no_getname,
 931         .poll           = rxrpc_poll,
 932         .ioctl          = sock_no_ioctl,
 933         .listen         = rxrpc_listen,
 934         .shutdown       = rxrpc_shutdown,
 935         .setsockopt     = rxrpc_setsockopt,
 936         .getsockopt     = rxrpc_getsockopt,
 937         .sendmsg        = rxrpc_sendmsg,
 938         .recvmsg        = rxrpc_recvmsg,
 939         .mmap           = sock_no_mmap,
 940         .sendpage       = sock_no_sendpage,
 941 };
 942
 943 static struct proto rxrpc_proto = {
 944         .name           = "RXRPC",
 945         .owner          = THIS_MODULE,
 946         .obj_size       = sizeof(struct rxrpc_sock),
 947         .max_header     = sizeof(struct rxrpc_wire_header),
 948 };
 949
 950 static const struct net_proto_family rxrpc_family_ops = {
 951         .family = PF_RXRPC,
 952         .create = rxrpc_create,
 953         .owner  = THIS_MODULE,
 954 };
 955
 956 /*
 957  * initialise and register the RxRPC protocol
 958  */
 959 static int __init af_rxrpc_init(void)
 960 {
 961         int ret = -1;
 962         unsigned int tmp;
 963
 964         BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > sizeof_field(struct sk_buff, cb));
 965
 966         get_random_bytes(&tmp, sizeof(tmp));
 967         tmp &= 0x3fffffff;
 968         if (tmp == 0)
 969                 tmp = 1;
 970         idr_set_cursor(&rxrpc_client_conn_ids, tmp);
 971
 972         ret = -ENOMEM;
 973         rxrpc_call_jar = kmem_cache_create(
 974                 "rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
 975                 SLAB_HWCACHE_ALIGN, NULL);
 976         if (!rxrpc_call_jar) {
 977                 pr_notice("Failed to allocate call jar\n");
 978                 goto error_call_jar;
 979         }
 980
 981         rxrpc_workqueue = alloc_workqueue("krxrpcd", 0, 1);
 982         if (!rxrpc_workqueue) {
 983                 pr_notice("Failed to allocate work queue\n");
 984                 goto error_work_queue;
 985         }
 986
 987         ret = rxrpc_init_security();
 988         if (ret < 0) {
 989                 pr_crit("Cannot initialise security\n");
 990                 goto error_security;
 991         }
 992
 993         ret = register_pernet_subsys(&rxrpc_net_ops);
 994         if (ret)
 995                 goto error_pernet;
 996
 997         ret = proto_register(&rxrpc_proto, 1);
 998         if (ret < 0) {
 999                 pr_crit("Cannot register protocol\n");
1000                 goto error_proto;
1001         }
1002
1003         ret = sock_register(&rxrpc_family_ops);
1004         if (ret < 0) {
1005                 pr_crit("Cannot register socket family\n");
1006                 goto error_sock;
1007         }
1008
1009         ret = register_key_type(&key_type_rxrpc);
1010         if (ret < 0) {
1011                 pr_crit("Cannot register client key type\n");
1012                 goto error_key_type;
1013         }
1014
1015         ret = register_key_type(&key_type_rxrpc_s);
1016         if (ret < 0) {
1017                 pr_crit("Cannot register server key type\n");
1018                 goto error_key_type_s;
1019         }
1020
1021         ret = rxrpc_sysctl_init();
1022         if (ret < 0) {
1023                 pr_crit("Cannot register sysctls\n");
1024                 goto error_sysctls;
1025         }
1026
1027         return 0;
1028
1029 error_sysctls:
1030         unregister_key_type(&key_type_rxrpc_s);
1031 error_key_type_s:
1032         unregister_key_type(&key_type_rxrpc);
1033 error_key_type:
1034         sock_unregister(PF_RXRPC);
1035 error_sock:
1036         proto_unregister(&rxrpc_proto);
1037 error_proto:
1038         unregister_pernet_subsys(&rxrpc_net_ops);
1039 error_pernet:
1040         rxrpc_exit_security();
1041 error_security:
1042         destroy_workqueue(rxrpc_workqueue);
1043 error_work_queue:
1044         kmem_cache_destroy(rxrpc_call_jar);
1045 error_call_jar:
1046         return ret;
1047 }
1048
1049 /*
1050  * unregister the RxRPC protocol
1051  */
1052 static void __exit af_rxrpc_exit(void)
1053 {
1054         _enter("");
1055         rxrpc_sysctl_exit();
1056         unregister_key_type(&key_type_rxrpc_s);
1057         unregister_key_type(&key_type_rxrpc);
1058         sock_unregister(PF_RXRPC);
1059         proto_unregister(&rxrpc_proto);
1060         unregister_pernet_subsys(&rxrpc_net_ops);
1061         ASSERTCMP(atomic_read(&rxrpc_n_tx_skbs), ==, 0);
1062         ASSERTCMP(atomic_read(&rxrpc_n_rx_skbs), ==, 0);
1063
1064         /* Make sure the local and peer records pinned by any dying connections
1065          * are released.
1066          */
1067         rcu_barrier();
1068         rxrpc_destroy_client_conn_ids();
1069
1070         destroy_workqueue(rxrpc_workqueue);
1071         rxrpc_exit_security();
1072         kmem_cache_destroy(rxrpc_call_jar);
1073         _leave("");
1074 }
1075
1076 module_init(af_rxrpc_init);
1077 module_exit(af_rxrpc_exit);