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
  25MODULE_DESCRIPTION("RxRPC network protocol");
  26MODULE_AUTHOR("Red Hat, Inc.");
  27MODULE_LICENSE("GPL");
  28MODULE_ALIAS_NETPROTO(PF_RXRPC);
  29
  30unsigned int rxrpc_debug; // = RXRPC_DEBUG_KPROTO;
  31module_param_named(debug, rxrpc_debug, uint, 0644);
  32MODULE_PARM_DESC(debug, "RxRPC debugging mask");
  33
  34static struct proto rxrpc_proto;
  35static const struct proto_ops rxrpc_rpc_ops;
  36
  37/* current debugging ID */
  38atomic_t rxrpc_debug_id;
  39EXPORT_SYMBOL(rxrpc_debug_id);
  40
  41/* count of skbs currently in use */
  42atomic_t rxrpc_n_rx_skbs;
  43
  44struct workqueue_struct *rxrpc_workqueue;
  45
  46static void rxrpc_sock_destructor(struct sock *);
  47
  48/*
  49 * see if an RxRPC socket is currently writable
  50 */
  51static 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 */
  59static 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_rcu(sk, SOCK_WAKE_SPACE, POLL_OUT);
  69	}
  70	rcu_read_unlock();
  71}
  72
  73/*
  74 * validate an RxRPC address
  75 */
  76static 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	switch (srx->transport.family) {
  97	case AF_INET:
  98		if (rx->family != AF_INET &&
  99		    rx->family != AF_INET6)
 100			return -EAFNOSUPPORT;
 101		if (srx->transport_len < sizeof(struct sockaddr_in))
 102			return -EINVAL;
 103		tail = offsetof(struct sockaddr_rxrpc, transport.sin.__pad);
 104		break;
 105
 106#ifdef CONFIG_AF_RXRPC_IPV6
 107	case AF_INET6:
 108		if (rx->family != AF_INET6)
 109			return -EAFNOSUPPORT;
 110		if (srx->transport_len < sizeof(struct sockaddr_in6))
 111			return -EINVAL;
 112		tail = offsetof(struct sockaddr_rxrpc, transport) +
 113			sizeof(struct sockaddr_in6);
 114		break;
 115#endif
 116
 117	default:
 118		return -EAFNOSUPPORT;
 119	}
 120
 121	if (tail < len)
 122		memset((void *)srx + tail, 0, len - tail);
 123	_debug("INET: %pISp", &srx->transport);
 124	return 0;
 125}
 126
 127/*
 128 * bind a local address to an RxRPC socket
 129 */
 130static int rxrpc_bind(struct socket *sock, struct sockaddr *saddr, int len)
 131{
 132	struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)saddr;
 133	struct rxrpc_local *local;
 134	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 135	u16 service_id;
 136	int ret;
 137
 138	_enter("%p,%p,%d", rx, saddr, len);
 139
 140	ret = rxrpc_validate_address(rx, srx, len);
 141	if (ret < 0)
 142		goto error;
 143	service_id = srx->srx_service;
 144
 145	lock_sock(&rx->sk);
 146
 147	switch (rx->sk.sk_state) {
 148	case RXRPC_UNBOUND:
 149		rx->srx = *srx;
 150		local = rxrpc_lookup_local(sock_net(&rx->sk), &rx->srx);
 151		if (IS_ERR(local)) {
 152			ret = PTR_ERR(local);
 153			goto error_unlock;
 154		}
 155
 156		if (service_id) {
 157			write_lock(&local->services_lock);
 158			if (local->service)
 159				goto service_in_use;
 160			rx->local = local;
 161			local->service = rx;
 162			write_unlock(&local->services_lock);
 163
 164			rx->sk.sk_state = RXRPC_SERVER_BOUND;
 165		} else {
 166			rx->local = local;
 167			rx->sk.sk_state = RXRPC_CLIENT_BOUND;
 168		}
 169		break;
 170
 171	case RXRPC_SERVER_BOUND:
 172		ret = -EINVAL;
 173		if (service_id == 0)
 174			goto error_unlock;
 175		ret = -EADDRINUSE;
 176		if (service_id == rx->srx.srx_service)
 177			goto error_unlock;
 178		ret = -EINVAL;
 179		srx->srx_service = rx->srx.srx_service;
 180		if (memcmp(srx, &rx->srx, sizeof(*srx)) != 0)
 181			goto error_unlock;
 182		rx->second_service = service_id;
 183		rx->sk.sk_state = RXRPC_SERVER_BOUND2;
 184		break;
 185
 186	default:
 187		ret = -EINVAL;
 188		goto error_unlock;
 189	}
 190
 191	release_sock(&rx->sk);
 192	_leave(" = 0");
 193	return 0;
 194
 195service_in_use:
 196	write_unlock(&local->services_lock);
 197	rxrpc_unuse_local(local, rxrpc_local_unuse_bind);
 198	rxrpc_put_local(local, rxrpc_local_put_bind);
 199	ret = -EADDRINUSE;
 200error_unlock:
 201	release_sock(&rx->sk);
 202error:
 203	_leave(" = %d", ret);
 204	return ret;
 205}
 206
 207/*
 208 * set the number of pending calls permitted on a listening socket
 209 */
 210static 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		fallthrough;
 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_lookup_peer - Obtain remote transport endpoint for an address
 263 * @sock: The socket through which it will be accessed
 264 * @srx: The network address
 265 * @gfp: Allocation flags
 266 *
 267 * Lookup or create a remote transport endpoint record for the specified
 268 * address and return it with a ref held.
 269 */
 270struct rxrpc_peer *rxrpc_kernel_lookup_peer(struct socket *sock,
 271					    struct sockaddr_rxrpc *srx, gfp_t gfp)
 272{
 273	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 274	int ret;
 275
 276	ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
 277	if (ret < 0)
 278		return ERR_PTR(ret);
 279
 280	return rxrpc_lookup_peer(rx->local, srx, gfp);
 281}
 282EXPORT_SYMBOL(rxrpc_kernel_lookup_peer);
 283
 284/**
 285 * rxrpc_kernel_get_peer - Get a reference on a peer
 286 * @peer: The peer to get a reference on.
 287 *
 288 * Get a record for the remote peer in a call.
 289 */
 290struct rxrpc_peer *rxrpc_kernel_get_peer(struct rxrpc_peer *peer)
 291{
 292	return peer ? rxrpc_get_peer(peer, rxrpc_peer_get_application) : NULL;
 293}
 294EXPORT_SYMBOL(rxrpc_kernel_get_peer);
 295
 296/**
 297 * rxrpc_kernel_put_peer - Allow a kernel app to drop a peer reference
 298 * @peer: The peer to drop a ref on
 299 */
 300void rxrpc_kernel_put_peer(struct rxrpc_peer *peer)
 301{
 302	rxrpc_put_peer(peer, rxrpc_peer_put_application);
 303}
 304EXPORT_SYMBOL(rxrpc_kernel_put_peer);
 305
 306/**
 307 * rxrpc_kernel_begin_call - Allow a kernel service to begin a call
 308 * @sock: The socket on which to make the call
 309 * @peer: The peer to contact
 310 * @key: The security context to use (defaults to socket setting)
 311 * @user_call_ID: The ID to use
 312 * @tx_total_len: Total length of data to transmit during the call (or -1)
 313 * @hard_timeout: The maximum lifespan of the call in sec
 314 * @gfp: The allocation constraints
 315 * @notify_rx: Where to send notifications instead of socket queue
 316 * @service_id: The ID of the service to contact
 317 * @upgrade: Request service upgrade for call
 318 * @interruptibility: The call is interruptible, or can be canceled.
 319 * @debug_id: The debug ID for tracing to be assigned to the call
 320 *
 321 * Allow a kernel service to begin a call on the nominated socket.  This just
 322 * sets up all the internal tracking structures and allocates connection and
 323 * call IDs as appropriate.  The call to be used is returned.
 324 *
 325 * The default socket destination address and security may be overridden by
 326 * supplying @srx and @key.
 327 */
 328struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
 329					   struct rxrpc_peer *peer,
 330					   struct key *key,
 331					   unsigned long user_call_ID,
 332					   s64 tx_total_len,
 333					   u32 hard_timeout,
 334					   gfp_t gfp,
 335					   rxrpc_notify_rx_t notify_rx,
 336					   u16 service_id,
 337					   bool upgrade,
 338					   enum rxrpc_interruptibility interruptibility,
 339					   unsigned int debug_id)
 340{
 341	struct rxrpc_conn_parameters cp;
 342	struct rxrpc_call_params p;
 343	struct rxrpc_call *call;
 344	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 345
 346	_enter(",,%x,%lx", key_serial(key), user_call_ID);
 347
 348	if (WARN_ON_ONCE(peer->local != rx->local))
 349		return ERR_PTR(-EIO);
 350
 351	lock_sock(&rx->sk);
 352
 353	if (!key)
 354		key = rx->key;
 355	if (key && !key->payload.data[0])
 356		key = NULL; /* a no-security key */
 357
 358	memset(&p, 0, sizeof(p));
 359	p.user_call_ID		= user_call_ID;
 360	p.tx_total_len		= tx_total_len;
 361	p.interruptibility	= interruptibility;
 362	p.kernel		= true;
 363	p.timeouts.hard		= hard_timeout;
 364
 365	memset(&cp, 0, sizeof(cp));
 366	cp.local		= rx->local;
 367	cp.peer			= peer;
 368	cp.key			= key;
 369	cp.security_level	= rx->min_sec_level;
 370	cp.exclusive		= false;
 371	cp.upgrade		= upgrade;
 372	cp.service_id		= service_id;
 373	call = rxrpc_new_client_call(rx, &cp, &p, gfp, debug_id);
 374	/* The socket has been unlocked. */
 375	if (!IS_ERR(call)) {
 376		call->notify_rx = notify_rx;
 377		mutex_unlock(&call->user_mutex);
 378	}
 379
 380	_leave(" = %p", call);
 381	return call;
 382}
 383EXPORT_SYMBOL(rxrpc_kernel_begin_call);
 384
 385/*
 386 * Dummy function used to stop the notifier talking to recvmsg().
 387 */
 388static void rxrpc_dummy_notify_rx(struct sock *sk, struct rxrpc_call *rxcall,
 389				  unsigned long call_user_ID)
 390{
 391}
 392
 393/**
 394 * rxrpc_kernel_shutdown_call - Allow a kernel service to shut down a call it was using
 395 * @sock: The socket the call is on
 396 * @call: The call to end
 397 *
 398 * Allow a kernel service to shut down a call it was using.  The call must be
 399 * complete before this is called (the call should be aborted if necessary).
 400 */
 401void rxrpc_kernel_shutdown_call(struct socket *sock, struct rxrpc_call *call)
 402{
 403	_enter("%d{%d}", call->debug_id, refcount_read(&call->ref));
 404
 405	mutex_lock(&call->user_mutex);
 406	if (!test_bit(RXRPC_CALL_RELEASED, &call->flags)) {
 407		rxrpc_release_call(rxrpc_sk(sock->sk), call);
 408
 409		/* Make sure we're not going to call back into a kernel service */
 410		if (call->notify_rx) {
 411			spin_lock(&call->notify_lock);
 412			call->notify_rx = rxrpc_dummy_notify_rx;
 413			spin_unlock(&call->notify_lock);
 414		}
 415	}
 416	mutex_unlock(&call->user_mutex);
 417}
 418EXPORT_SYMBOL(rxrpc_kernel_shutdown_call);
 419
 420/**
 421 * rxrpc_kernel_put_call - Release a reference to a call
 422 * @sock: The socket the call is on
 423 * @call: The call to put
 424 *
 425 * Drop the application's ref on an rxrpc call.
 426 */
 427void rxrpc_kernel_put_call(struct socket *sock, struct rxrpc_call *call)
 428{
 429	rxrpc_put_call(call, rxrpc_call_put_kernel);
 430}
 431EXPORT_SYMBOL(rxrpc_kernel_put_call);
 432
 433/**
 434 * rxrpc_kernel_check_life - Check to see whether a call is still alive
 435 * @sock: The socket the call is on
 436 * @call: The call to check
 437 *
 438 * Allow a kernel service to find out whether a call is still alive - whether
 439 * it has completed successfully and all received data has been consumed.
 440 */
 441bool rxrpc_kernel_check_life(const struct socket *sock,
 442			     const struct rxrpc_call *call)
 443{
 444	if (!rxrpc_call_is_complete(call))
 445		return true;
 446	if (call->completion != RXRPC_CALL_SUCCEEDED)
 447		return false;
 448	return !skb_queue_empty(&call->recvmsg_queue);
 449}
 450EXPORT_SYMBOL(rxrpc_kernel_check_life);
 451
 452/**
 453 * rxrpc_kernel_get_epoch - Retrieve the epoch value from a call.
 454 * @sock: The socket the call is on
 455 * @call: The call to query
 456 *
 457 * Allow a kernel service to retrieve the epoch value from a service call to
 458 * see if the client at the other end rebooted.
 459 */
 460u32 rxrpc_kernel_get_epoch(struct socket *sock, struct rxrpc_call *call)
 461{
 462	return call->conn->proto.epoch;
 463}
 464EXPORT_SYMBOL(rxrpc_kernel_get_epoch);
 465
 466/**
 467 * rxrpc_kernel_new_call_notification - Get notifications of new calls
 468 * @sock: The socket to intercept received messages on
 469 * @notify_new_call: Function to be called when new calls appear
 470 * @discard_new_call: Function to discard preallocated calls
 471 *
 472 * Allow a kernel service to be given notifications about new calls.
 473 */
 474void rxrpc_kernel_new_call_notification(
 475	struct socket *sock,
 476	rxrpc_notify_new_call_t notify_new_call,
 477	rxrpc_discard_new_call_t discard_new_call)
 478{
 479	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 480
 481	rx->notify_new_call = notify_new_call;
 482	rx->discard_new_call = discard_new_call;
 483}
 484EXPORT_SYMBOL(rxrpc_kernel_new_call_notification);
 485
 486/**
 487 * rxrpc_kernel_set_max_life - Set maximum lifespan on a call
 488 * @sock: The socket the call is on
 489 * @call: The call to configure
 490 * @hard_timeout: The maximum lifespan of the call in ms
 491 *
 492 * Set the maximum lifespan of a call.  The call will end with ETIME or
 493 * ETIMEDOUT if it takes longer than this.
 494 */
 495void rxrpc_kernel_set_max_life(struct socket *sock, struct rxrpc_call *call,
 496			       unsigned long hard_timeout)
 497{
 498	ktime_t delay = ms_to_ktime(hard_timeout), expect_term_by;
 499
 500	mutex_lock(&call->user_mutex);
 501
 502	expect_term_by = ktime_add(ktime_get_real(), delay);
 503	WRITE_ONCE(call->expect_term_by, expect_term_by);
 504	trace_rxrpc_timer_set(call, delay, rxrpc_timer_trace_hard);
 505	rxrpc_poke_call(call, rxrpc_call_poke_set_timeout);
 506
 507	mutex_unlock(&call->user_mutex);
 508}
 509EXPORT_SYMBOL(rxrpc_kernel_set_max_life);
 510
 511/*
 512 * connect an RxRPC socket
 513 * - this just targets it at a specific destination; no actual connection
 514 *   negotiation takes place
 515 */
 516static int rxrpc_connect(struct socket *sock, struct sockaddr *addr,
 517			 int addr_len, int flags)
 518{
 519	struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
 520	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 521	int ret;
 522
 523	_enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
 524
 525	ret = rxrpc_validate_address(rx, srx, addr_len);
 526	if (ret < 0) {
 527		_leave(" = %d [bad addr]", ret);
 528		return ret;
 529	}
 530
 531	lock_sock(&rx->sk);
 532
 533	ret = -EISCONN;
 534	if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
 535		goto error;
 536
 537	switch (rx->sk.sk_state) {
 538	case RXRPC_UNBOUND:
 539		rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
 540		break;
 541	case RXRPC_CLIENT_UNBOUND:
 542	case RXRPC_CLIENT_BOUND:
 543		break;
 544	default:
 545		ret = -EBUSY;
 546		goto error;
 547	}
 548
 549	rx->connect_srx = *srx;
 550	set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
 551	ret = 0;
 552
 553error:
 554	release_sock(&rx->sk);
 555	return ret;
 556}
 557
 558/*
 559 * send a message through an RxRPC socket
 560 * - in a client this does a number of things:
 561 *   - finds/sets up a connection for the security specified (if any)
 562 *   - initiates a call (ID in control data)
 563 *   - ends the request phase of a call (if MSG_MORE is not set)
 564 *   - sends a call data packet
 565 *   - may send an abort (abort code in control data)
 566 */
 567static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
 568{
 569	struct rxrpc_local *local;
 570	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 571	int ret;
 572
 573	_enter(",{%d},,%zu", rx->sk.sk_state, len);
 574
 575	if (m->msg_flags & MSG_OOB)
 576		return -EOPNOTSUPP;
 577
 578	if (m->msg_name) {
 579		ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
 580		if (ret < 0) {
 581			_leave(" = %d [bad addr]", ret);
 582			return ret;
 583		}
 584	}
 585
 586	lock_sock(&rx->sk);
 587
 588	switch (rx->sk.sk_state) {
 589	case RXRPC_UNBOUND:
 590	case RXRPC_CLIENT_UNBOUND:
 591		rx->srx.srx_family = AF_RXRPC;
 592		rx->srx.srx_service = 0;
 593		rx->srx.transport_type = SOCK_DGRAM;
 594		rx->srx.transport.family = rx->family;
 595		switch (rx->family) {
 596		case AF_INET:
 597			rx->srx.transport_len = sizeof(struct sockaddr_in);
 598			break;
 599#ifdef CONFIG_AF_RXRPC_IPV6
 600		case AF_INET6:
 601			rx->srx.transport_len = sizeof(struct sockaddr_in6);
 602			break;
 603#endif
 604		default:
 605			ret = -EAFNOSUPPORT;
 606			goto error_unlock;
 607		}
 608		local = rxrpc_lookup_local(sock_net(sock->sk), &rx->srx);
 609		if (IS_ERR(local)) {
 610			ret = PTR_ERR(local);
 611			goto error_unlock;
 612		}
 613
 614		rx->local = local;
 615		rx->sk.sk_state = RXRPC_CLIENT_BOUND;
 616		fallthrough;
 617
 618	case RXRPC_CLIENT_BOUND:
 619		if (!m->msg_name &&
 620		    test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
 621			m->msg_name = &rx->connect_srx;
 622			m->msg_namelen = sizeof(rx->connect_srx);
 623		}
 624		fallthrough;
 625	case RXRPC_SERVER_BOUND:
 626	case RXRPC_SERVER_LISTENING:
 627		ret = rxrpc_do_sendmsg(rx, m, len);
 628		/* The socket has been unlocked */
 629		goto out;
 630	default:
 631		ret = -EINVAL;
 632		goto error_unlock;
 633	}
 634
 635error_unlock:
 636	release_sock(&rx->sk);
 637out:
 638	_leave(" = %d", ret);
 639	return ret;
 640}
 641
 642int rxrpc_sock_set_min_security_level(struct sock *sk, unsigned int val)
 643{
 644	if (sk->sk_state != RXRPC_UNBOUND)
 645		return -EISCONN;
 646	if (val > RXRPC_SECURITY_MAX)
 647		return -EINVAL;
 648	lock_sock(sk);
 649	rxrpc_sk(sk)->min_sec_level = val;
 650	release_sock(sk);
 651	return 0;
 652}
 653EXPORT_SYMBOL(rxrpc_sock_set_min_security_level);
 654
 655/*
 656 * set RxRPC socket options
 657 */
 658static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
 659			    sockptr_t optval, unsigned int optlen)
 660{
 661	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 662	unsigned int min_sec_level;
 663	u16 service_upgrade[2];
 664	int ret;
 665
 666	_enter(",%d,%d,,%d", level, optname, optlen);
 667
 668	lock_sock(&rx->sk);
 669	ret = -EOPNOTSUPP;
 670
 671	if (level == SOL_RXRPC) {
 672		switch (optname) {
 673		case RXRPC_EXCLUSIVE_CONNECTION:
 674			ret = -EINVAL;
 675			if (optlen != 0)
 676				goto error;
 677			ret = -EISCONN;
 678			if (rx->sk.sk_state != RXRPC_UNBOUND)
 679				goto error;
 680			rx->exclusive = true;
 681			goto success;
 682
 683		case RXRPC_SECURITY_KEY:
 684			ret = -EINVAL;
 685			if (rx->key)
 686				goto error;
 687			ret = -EISCONN;
 688			if (rx->sk.sk_state != RXRPC_UNBOUND)
 689				goto error;
 690			ret = rxrpc_request_key(rx, optval, optlen);
 691			goto error;
 692
 693		case RXRPC_SECURITY_KEYRING:
 694			ret = -EINVAL;
 695			if (rx->key)
 696				goto error;
 697			ret = -EISCONN;
 698			if (rx->sk.sk_state != RXRPC_UNBOUND)
 699				goto error;
 700			ret = rxrpc_server_keyring(rx, optval, optlen);
 701			goto error;
 702
 703		case RXRPC_MIN_SECURITY_LEVEL:
 704			ret = -EINVAL;
 705			if (optlen != sizeof(unsigned int))
 706				goto error;
 707			ret = -EISCONN;
 708			if (rx->sk.sk_state != RXRPC_UNBOUND)
 709				goto error;
 710			ret = copy_safe_from_sockptr(&min_sec_level,
 711						     sizeof(min_sec_level),
 712						     optval, optlen);
 713			if (ret)
 714				goto error;
 715			ret = -EINVAL;
 716			if (min_sec_level > RXRPC_SECURITY_MAX)
 717				goto error;
 718			rx->min_sec_level = min_sec_level;
 719			goto success;
 720
 721		case RXRPC_UPGRADEABLE_SERVICE:
 722			ret = -EINVAL;
 723			if (optlen != sizeof(service_upgrade) ||
 724			    rx->service_upgrade.from != 0)
 725				goto error;
 726			ret = -EISCONN;
 727			if (rx->sk.sk_state != RXRPC_SERVER_BOUND2)
 728				goto error;
 729			ret = -EFAULT;
 730			if (copy_from_sockptr(service_upgrade, optval,
 731					   sizeof(service_upgrade)) != 0)
 732				goto error;
 733			ret = -EINVAL;
 734			if ((service_upgrade[0] != rx->srx.srx_service ||
 735			     service_upgrade[1] != rx->second_service) &&
 736			    (service_upgrade[0] != rx->second_service ||
 737			     service_upgrade[1] != rx->srx.srx_service))
 738				goto error;
 739			rx->service_upgrade.from = service_upgrade[0];
 740			rx->service_upgrade.to = service_upgrade[1];
 741			goto success;
 742
 743		default:
 744			break;
 745		}
 746	}
 747
 748success:
 749	ret = 0;
 750error:
 751	release_sock(&rx->sk);
 752	return ret;
 753}
 754
 755/*
 756 * Get socket options.
 757 */
 758static int rxrpc_getsockopt(struct socket *sock, int level, int optname,
 759			    char __user *optval, int __user *_optlen)
 760{
 761	int optlen;
 762
 763	if (level != SOL_RXRPC)
 764		return -EOPNOTSUPP;
 765
 766	if (get_user(optlen, _optlen))
 767		return -EFAULT;
 768
 769	switch (optname) {
 770	case RXRPC_SUPPORTED_CMSG:
 771		if (optlen < sizeof(int))
 772			return -ETOOSMALL;
 773		if (put_user(RXRPC__SUPPORTED - 1, (int __user *)optval) ||
 774		    put_user(sizeof(int), _optlen))
 775			return -EFAULT;
 776		return 0;
 777
 778	default:
 779		return -EOPNOTSUPP;
 780	}
 781}
 782
 783/*
 784 * permit an RxRPC socket to be polled
 785 */
 786static __poll_t rxrpc_poll(struct file *file, struct socket *sock,
 787			       poll_table *wait)
 788{
 789	struct sock *sk = sock->sk;
 790	struct rxrpc_sock *rx = rxrpc_sk(sk);
 791	__poll_t mask;
 792
 793	sock_poll_wait(file, sock, wait);
 794	mask = 0;
 795
 796	/* the socket is readable if there are any messages waiting on the Rx
 797	 * queue */
 798	if (!list_empty(&rx->recvmsg_q))
 799		mask |= EPOLLIN | EPOLLRDNORM;
 800
 801	/* the socket is writable if there is space to add new data to the
 802	 * socket; there is no guarantee that any particular call in progress
 803	 * on the socket may have space in the Tx ACK window */
 804	if (rxrpc_writable(sk))
 805		mask |= EPOLLOUT | EPOLLWRNORM;
 806
 807	return mask;
 808}
 809
 810/*
 811 * create an RxRPC socket
 812 */
 813static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
 814			int kern)
 815{
 816	struct rxrpc_net *rxnet;
 817	struct rxrpc_sock *rx;
 818	struct sock *sk;
 819
 820	_enter("%p,%d", sock, protocol);
 821
 822	/* we support transport protocol UDP/UDP6 only */
 823	if (protocol != PF_INET &&
 824	    IS_ENABLED(CONFIG_AF_RXRPC_IPV6) && protocol != PF_INET6)
 825		return -EPROTONOSUPPORT;
 826
 827	if (sock->type != SOCK_DGRAM)
 828		return -ESOCKTNOSUPPORT;
 829
 830	sock->ops = &rxrpc_rpc_ops;
 831	sock->state = SS_UNCONNECTED;
 832
 833	sk = sk_alloc(net, PF_RXRPC, GFP_KERNEL, &rxrpc_proto, kern);
 834	if (!sk)
 835		return -ENOMEM;
 836
 837	sock_init_data(sock, sk);
 838	sock_set_flag(sk, SOCK_RCU_FREE);
 839	sk->sk_state		= RXRPC_UNBOUND;
 840	sk->sk_write_space	= rxrpc_write_space;
 841	sk->sk_max_ack_backlog	= 0;
 842	sk->sk_destruct		= rxrpc_sock_destructor;
 843
 844	rx = rxrpc_sk(sk);
 845	rx->family = protocol;
 846	rx->calls = RB_ROOT;
 847
 848	spin_lock_init(&rx->incoming_lock);
 849	INIT_LIST_HEAD(&rx->sock_calls);
 850	INIT_LIST_HEAD(&rx->to_be_accepted);
 851	INIT_LIST_HEAD(&rx->recvmsg_q);
 852	spin_lock_init(&rx->recvmsg_lock);
 853	rwlock_init(&rx->call_lock);
 854	memset(&rx->srx, 0, sizeof(rx->srx));
 855
 856	rxnet = rxrpc_net(sock_net(&rx->sk));
 857	timer_reduce(&rxnet->peer_keepalive_timer, jiffies + 1);
 858
 859	_leave(" = 0 [%p]", rx);
 860	return 0;
 861}
 862
 863/*
 864 * Kill all the calls on a socket and shut it down.
 865 */
 866static int rxrpc_shutdown(struct socket *sock, int flags)
 867{
 868	struct sock *sk = sock->sk;
 869	struct rxrpc_sock *rx = rxrpc_sk(sk);
 870	int ret = 0;
 871
 872	_enter("%p,%d", sk, flags);
 873
 874	if (flags != SHUT_RDWR)
 875		return -EOPNOTSUPP;
 876	if (sk->sk_state == RXRPC_CLOSE)
 877		return -ESHUTDOWN;
 878
 879	lock_sock(sk);
 880
 881	if (sk->sk_state < RXRPC_CLOSE) {
 882		sk->sk_state = RXRPC_CLOSE;
 883		sk->sk_shutdown = SHUTDOWN_MASK;
 884	} else {
 885		ret = -ESHUTDOWN;
 886	}
 887
 888	rxrpc_discard_prealloc(rx);
 889
 890	release_sock(sk);
 891	return ret;
 892}
 893
 894/*
 895 * RxRPC socket destructor
 896 */
 897static void rxrpc_sock_destructor(struct sock *sk)
 898{
 899	_enter("%p", sk);
 900
 901	rxrpc_purge_queue(&sk->sk_receive_queue);
 902
 903	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
 904	WARN_ON(!sk_unhashed(sk));
 905	WARN_ON(sk->sk_socket);
 906
 907	if (!sock_flag(sk, SOCK_DEAD)) {
 908		printk("Attempt to release alive rxrpc socket: %p\n", sk);
 909		return;
 910	}
 911}
 912
 913/*
 914 * release an RxRPC socket
 915 */
 916static int rxrpc_release_sock(struct sock *sk)
 917{
 918	struct rxrpc_sock *rx = rxrpc_sk(sk);
 919
 920	_enter("%p{%d,%d}", sk, sk->sk_state, refcount_read(&sk->sk_refcnt));
 921
 922	/* declare the socket closed for business */
 923	sock_orphan(sk);
 924	sk->sk_shutdown = SHUTDOWN_MASK;
 925
 926	/* We want to kill off all connections from a service socket
 927	 * as fast as possible because we can't share these; client
 928	 * sockets, on the other hand, can share an endpoint.
 929	 */
 930	switch (sk->sk_state) {
 931	case RXRPC_SERVER_BOUND:
 932	case RXRPC_SERVER_BOUND2:
 933	case RXRPC_SERVER_LISTENING:
 934	case RXRPC_SERVER_LISTEN_DISABLED:
 935		rx->local->service_closed = true;
 936		break;
 937	}
 938
 939	sk->sk_state = RXRPC_CLOSE;
 940
 941	if (rx->local && rx->local->service == rx) {
 942		write_lock(&rx->local->services_lock);
 943		rx->local->service = NULL;
 944		write_unlock(&rx->local->services_lock);
 945	}
 946
 947	/* try to flush out this socket */
 948	rxrpc_discard_prealloc(rx);
 949	rxrpc_release_calls_on_socket(rx);
 950	flush_workqueue(rxrpc_workqueue);
 951	rxrpc_purge_queue(&sk->sk_receive_queue);
 952
 953	rxrpc_unuse_local(rx->local, rxrpc_local_unuse_release_sock);
 954	rxrpc_put_local(rx->local, rxrpc_local_put_release_sock);
 955	rx->local = NULL;
 956	key_put(rx->key);
 957	rx->key = NULL;
 958	key_put(rx->securities);
 959	rx->securities = NULL;
 960	sock_put(sk);
 961
 962	_leave(" = 0");
 963	return 0;
 964}
 965
 966/*
 967 * release an RxRPC BSD socket on close() or equivalent
 968 */
 969static int rxrpc_release(struct socket *sock)
 970{
 971	struct sock *sk = sock->sk;
 972
 973	_enter("%p{%p}", sock, sk);
 974
 975	if (!sk)
 976		return 0;
 977
 978	sock->sk = NULL;
 979
 980	return rxrpc_release_sock(sk);
 981}
 982
 983/*
 984 * RxRPC network protocol
 985 */
 986static const struct proto_ops rxrpc_rpc_ops = {
 987	.family		= PF_RXRPC,
 988	.owner		= THIS_MODULE,
 989	.release	= rxrpc_release,
 990	.bind		= rxrpc_bind,
 991	.connect	= rxrpc_connect,
 992	.socketpair	= sock_no_socketpair,
 993	.accept		= sock_no_accept,
 994	.getname	= sock_no_getname,
 995	.poll		= rxrpc_poll,
 996	.ioctl		= sock_no_ioctl,
 997	.listen		= rxrpc_listen,
 998	.shutdown	= rxrpc_shutdown,
 999	.setsockopt	= rxrpc_setsockopt,
1000	.getsockopt	= rxrpc_getsockopt,
1001	.sendmsg	= rxrpc_sendmsg,
1002	.recvmsg	= rxrpc_recvmsg,
1003	.mmap		= sock_no_mmap,
1004};
1005
1006static struct proto rxrpc_proto = {
1007	.name		= "RXRPC",
1008	.owner		= THIS_MODULE,
1009	.obj_size	= sizeof(struct rxrpc_sock),
1010	.max_header	= sizeof(struct rxrpc_wire_header),
1011};
1012
1013static const struct net_proto_family rxrpc_family_ops = {
1014	.family	= PF_RXRPC,
1015	.create = rxrpc_create,
1016	.owner	= THIS_MODULE,
1017};
1018
1019/*
1020 * initialise and register the RxRPC protocol
1021 */
1022static int __init af_rxrpc_init(void)
1023{
1024	int ret = -1;
1025
1026	BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > sizeof_field(struct sk_buff, cb));
1027
1028	ret = -ENOMEM;
1029	rxrpc_gen_version_string();
1030	rxrpc_call_jar = kmem_cache_create(
1031		"rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
1032		SLAB_HWCACHE_ALIGN, NULL);
1033	if (!rxrpc_call_jar) {
1034		pr_notice("Failed to allocate call jar\n");
1035		goto error_call_jar;
1036	}
1037
1038	rxrpc_workqueue = alloc_ordered_workqueue("krxrpcd", WQ_HIGHPRI | WQ_MEM_RECLAIM);
1039	if (!rxrpc_workqueue) {
1040		pr_notice("Failed to allocate work queue\n");
1041		goto error_work_queue;
1042	}
1043
1044	ret = rxrpc_init_security();
1045	if (ret < 0) {
1046		pr_crit("Cannot initialise security\n");
1047		goto error_security;
1048	}
1049
1050	ret = register_pernet_device(&rxrpc_net_ops);
1051	if (ret)
1052		goto error_pernet;
1053
1054	ret = proto_register(&rxrpc_proto, 1);
1055	if (ret < 0) {
1056		pr_crit("Cannot register protocol\n");
1057		goto error_proto;
1058	}
1059
1060	ret = sock_register(&rxrpc_family_ops);
1061	if (ret < 0) {
1062		pr_crit("Cannot register socket family\n");
1063		goto error_sock;
1064	}
1065
1066	ret = register_key_type(&key_type_rxrpc);
1067	if (ret < 0) {
1068		pr_crit("Cannot register client key type\n");
1069		goto error_key_type;
1070	}
1071
1072	ret = register_key_type(&key_type_rxrpc_s);
1073	if (ret < 0) {
1074		pr_crit("Cannot register server key type\n");
1075		goto error_key_type_s;
1076	}
1077
1078	ret = rxrpc_sysctl_init();
1079	if (ret < 0) {
1080		pr_crit("Cannot register sysctls\n");
1081		goto error_sysctls;
1082	}
1083
1084	return 0;
1085
1086error_sysctls:
1087	unregister_key_type(&key_type_rxrpc_s);
1088error_key_type_s:
1089	unregister_key_type(&key_type_rxrpc);
1090error_key_type:
1091	sock_unregister(PF_RXRPC);
1092error_sock:
1093	proto_unregister(&rxrpc_proto);
1094error_proto:
1095	unregister_pernet_device(&rxrpc_net_ops);
1096error_pernet:
1097	rxrpc_exit_security();
1098error_security:
1099	destroy_workqueue(rxrpc_workqueue);
1100error_work_queue:
1101	kmem_cache_destroy(rxrpc_call_jar);
1102error_call_jar:
1103	return ret;
1104}
1105
1106/*
1107 * unregister the RxRPC protocol
1108 */
1109static void __exit af_rxrpc_exit(void)
1110{
1111	_enter("");
1112	rxrpc_sysctl_exit();
1113	unregister_key_type(&key_type_rxrpc_s);
1114	unregister_key_type(&key_type_rxrpc);
1115	sock_unregister(PF_RXRPC);
1116	proto_unregister(&rxrpc_proto);
1117	unregister_pernet_device(&rxrpc_net_ops);
1118	ASSERTCMP(atomic_read(&rxrpc_n_rx_skbs), ==, 0);
1119
1120	/* Make sure the local and peer records pinned by any dying connections
1121	 * are released.
1122	 */
1123	rcu_barrier();
1124
1125	destroy_workqueue(rxrpc_workqueue);
1126	rxrpc_exit_security();
1127	kmem_cache_destroy(rxrpc_call_jar);
1128	_leave("");
1129}
1130
1131module_init(af_rxrpc_init);
1132module_exit(af_rxrpc_exit);