xref: /linux/net/rxrpc/af_rxrpc.c (revision f788bb2fb915a79eae0a160580766ba52984d61f)
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 		/* Fall through */
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  * @intr: The call is interruptible
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 
315 	memset(&cp, 0, sizeof(cp));
316 	cp.local		= rx->local;
317 	cp.key			= key;
318 	cp.security_level	= rx->min_sec_level;
319 	cp.exclusive		= false;
320 	cp.upgrade		= upgrade;
321 	cp.service_id		= srx->srx_service;
322 	call = rxrpc_new_client_call(rx, &cp, srx, &p, gfp, debug_id);
323 	/* The socket has been unlocked. */
324 	if (!IS_ERR(call)) {
325 		call->notify_rx = notify_rx;
326 		mutex_unlock(&call->user_mutex);
327 	}
328 
329 	rxrpc_put_peer(cp.peer);
330 	_leave(" = %p", call);
331 	return call;
332 }
333 EXPORT_SYMBOL(rxrpc_kernel_begin_call);
334 
335 /*
336  * Dummy function used to stop the notifier talking to recvmsg().
337  */
338 static void rxrpc_dummy_notify_rx(struct sock *sk, struct rxrpc_call *rxcall,
339 				  unsigned long call_user_ID)
340 {
341 }
342 
343 /**
344  * rxrpc_kernel_end_call - Allow a kernel service to end a call it was using
345  * @sock: The socket the call is on
346  * @call: The call to end
347  *
348  * Allow a kernel service to end a call it was using.  The call must be
349  * complete before this is called (the call should be aborted if necessary).
350  */
351 void rxrpc_kernel_end_call(struct socket *sock, struct rxrpc_call *call)
352 {
353 	_enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
354 
355 	mutex_lock(&call->user_mutex);
356 	rxrpc_release_call(rxrpc_sk(sock->sk), call);
357 
358 	/* Make sure we're not going to call back into a kernel service */
359 	if (call->notify_rx) {
360 		spin_lock_bh(&call->notify_lock);
361 		call->notify_rx = rxrpc_dummy_notify_rx;
362 		spin_unlock_bh(&call->notify_lock);
363 	}
364 
365 	mutex_unlock(&call->user_mutex);
366 	rxrpc_put_call(call, rxrpc_call_put_kernel);
367 }
368 EXPORT_SYMBOL(rxrpc_kernel_end_call);
369 
370 /**
371  * rxrpc_kernel_check_life - Check to see whether a call is still alive
372  * @sock: The socket the call is on
373  * @call: The call to check
374  *
375  * Allow a kernel service to find out whether a call is still alive -
376  * ie. whether it has completed.
377  */
378 bool rxrpc_kernel_check_life(const struct socket *sock,
379 			     const struct rxrpc_call *call)
380 {
381 	return call->state != RXRPC_CALL_COMPLETE;
382 }
383 EXPORT_SYMBOL(rxrpc_kernel_check_life);
384 
385 /**
386  * rxrpc_kernel_get_epoch - Retrieve the epoch value from a call.
387  * @sock: The socket the call is on
388  * @call: The call to query
389  *
390  * Allow a kernel service to retrieve the epoch value from a service call to
391  * see if the client at the other end rebooted.
392  */
393 u32 rxrpc_kernel_get_epoch(struct socket *sock, struct rxrpc_call *call)
394 {
395 	return call->conn->proto.epoch;
396 }
397 EXPORT_SYMBOL(rxrpc_kernel_get_epoch);
398 
399 /**
400  * rxrpc_kernel_new_call_notification - Get notifications of new calls
401  * @sock: The socket to intercept received messages on
402  * @notify_new_call: Function to be called when new calls appear
403  * @discard_new_call: Function to discard preallocated calls
404  *
405  * Allow a kernel service to be given notifications about new calls.
406  */
407 void rxrpc_kernel_new_call_notification(
408 	struct socket *sock,
409 	rxrpc_notify_new_call_t notify_new_call,
410 	rxrpc_discard_new_call_t discard_new_call)
411 {
412 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
413 
414 	rx->notify_new_call = notify_new_call;
415 	rx->discard_new_call = discard_new_call;
416 }
417 EXPORT_SYMBOL(rxrpc_kernel_new_call_notification);
418 
419 /**
420  * rxrpc_kernel_set_max_life - Set maximum lifespan on a call
421  * @sock: The socket the call is on
422  * @call: The call to configure
423  * @hard_timeout: The maximum lifespan of the call in jiffies
424  *
425  * Set the maximum lifespan of a call.  The call will end with ETIME or
426  * ETIMEDOUT if it takes longer than this.
427  */
428 void rxrpc_kernel_set_max_life(struct socket *sock, struct rxrpc_call *call,
429 			       unsigned long hard_timeout)
430 {
431 	unsigned long now;
432 
433 	mutex_lock(&call->user_mutex);
434 
435 	now = jiffies;
436 	hard_timeout += now;
437 	WRITE_ONCE(call->expect_term_by, hard_timeout);
438 	rxrpc_reduce_call_timer(call, hard_timeout, now, rxrpc_timer_set_for_hard);
439 
440 	mutex_unlock(&call->user_mutex);
441 }
442 EXPORT_SYMBOL(rxrpc_kernel_set_max_life);
443 
444 /*
445  * connect an RxRPC socket
446  * - this just targets it at a specific destination; no actual connection
447  *   negotiation takes place
448  */
449 static int rxrpc_connect(struct socket *sock, struct sockaddr *addr,
450 			 int addr_len, int flags)
451 {
452 	struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
453 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
454 	int ret;
455 
456 	_enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
457 
458 	ret = rxrpc_validate_address(rx, srx, addr_len);
459 	if (ret < 0) {
460 		_leave(" = %d [bad addr]", ret);
461 		return ret;
462 	}
463 
464 	lock_sock(&rx->sk);
465 
466 	ret = -EISCONN;
467 	if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
468 		goto error;
469 
470 	switch (rx->sk.sk_state) {
471 	case RXRPC_UNBOUND:
472 		rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
473 	case RXRPC_CLIENT_UNBOUND:
474 	case RXRPC_CLIENT_BOUND:
475 		break;
476 	default:
477 		ret = -EBUSY;
478 		goto error;
479 	}
480 
481 	rx->connect_srx = *srx;
482 	set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
483 	ret = 0;
484 
485 error:
486 	release_sock(&rx->sk);
487 	return ret;
488 }
489 
490 /*
491  * send a message through an RxRPC socket
492  * - in a client this does a number of things:
493  *   - finds/sets up a connection for the security specified (if any)
494  *   - initiates a call (ID in control data)
495  *   - ends the request phase of a call (if MSG_MORE is not set)
496  *   - sends a call data packet
497  *   - may send an abort (abort code in control data)
498  */
499 static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
500 {
501 	struct rxrpc_local *local;
502 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
503 	int ret;
504 
505 	_enter(",{%d},,%zu", rx->sk.sk_state, len);
506 
507 	if (m->msg_flags & MSG_OOB)
508 		return -EOPNOTSUPP;
509 
510 	if (m->msg_name) {
511 		ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
512 		if (ret < 0) {
513 			_leave(" = %d [bad addr]", ret);
514 			return ret;
515 		}
516 	}
517 
518 	lock_sock(&rx->sk);
519 
520 	switch (rx->sk.sk_state) {
521 	case RXRPC_UNBOUND:
522 	case RXRPC_CLIENT_UNBOUND:
523 		rx->srx.srx_family = AF_RXRPC;
524 		rx->srx.srx_service = 0;
525 		rx->srx.transport_type = SOCK_DGRAM;
526 		rx->srx.transport.family = rx->family;
527 		switch (rx->family) {
528 		case AF_INET:
529 			rx->srx.transport_len = sizeof(struct sockaddr_in);
530 			break;
531 #ifdef CONFIG_AF_RXRPC_IPV6
532 		case AF_INET6:
533 			rx->srx.transport_len = sizeof(struct sockaddr_in6);
534 			break;
535 #endif
536 		default:
537 			ret = -EAFNOSUPPORT;
538 			goto error_unlock;
539 		}
540 		local = rxrpc_lookup_local(sock_net(sock->sk), &rx->srx);
541 		if (IS_ERR(local)) {
542 			ret = PTR_ERR(local);
543 			goto error_unlock;
544 		}
545 
546 		rx->local = local;
547 		rx->sk.sk_state = RXRPC_CLIENT_BOUND;
548 		/* Fall through */
549 
550 	case RXRPC_CLIENT_BOUND:
551 		if (!m->msg_name &&
552 		    test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
553 			m->msg_name = &rx->connect_srx;
554 			m->msg_namelen = sizeof(rx->connect_srx);
555 		}
556 		/* Fall through */
557 	case RXRPC_SERVER_BOUND:
558 	case RXRPC_SERVER_LISTENING:
559 		ret = rxrpc_do_sendmsg(rx, m, len);
560 		/* The socket has been unlocked */
561 		goto out;
562 	default:
563 		ret = -EINVAL;
564 		goto error_unlock;
565 	}
566 
567 error_unlock:
568 	release_sock(&rx->sk);
569 out:
570 	_leave(" = %d", ret);
571 	return ret;
572 }
573 
574 int rxrpc_sock_set_min_security_level(struct sock *sk, unsigned int val)
575 {
576 	if (sk->sk_state != RXRPC_UNBOUND)
577 		return -EISCONN;
578 	if (val > RXRPC_SECURITY_MAX)
579 		return -EINVAL;
580 	lock_sock(sk);
581 	rxrpc_sk(sk)->min_sec_level = val;
582 	release_sock(sk);
583 	return 0;
584 }
585 EXPORT_SYMBOL(rxrpc_sock_set_min_security_level);
586 
587 /*
588  * set RxRPC socket options
589  */
590 static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
591 			    char __user *optval, unsigned int optlen)
592 {
593 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
594 	unsigned int min_sec_level;
595 	u16 service_upgrade[2];
596 	int ret;
597 
598 	_enter(",%d,%d,,%d", level, optname, optlen);
599 
600 	lock_sock(&rx->sk);
601 	ret = -EOPNOTSUPP;
602 
603 	if (level == SOL_RXRPC) {
604 		switch (optname) {
605 		case RXRPC_EXCLUSIVE_CONNECTION:
606 			ret = -EINVAL;
607 			if (optlen != 0)
608 				goto error;
609 			ret = -EISCONN;
610 			if (rx->sk.sk_state != RXRPC_UNBOUND)
611 				goto error;
612 			rx->exclusive = true;
613 			goto success;
614 
615 		case RXRPC_SECURITY_KEY:
616 			ret = -EINVAL;
617 			if (rx->key)
618 				goto error;
619 			ret = -EISCONN;
620 			if (rx->sk.sk_state != RXRPC_UNBOUND)
621 				goto error;
622 			ret = rxrpc_request_key(rx, optval, optlen);
623 			goto error;
624 
625 		case RXRPC_SECURITY_KEYRING:
626 			ret = -EINVAL;
627 			if (rx->key)
628 				goto error;
629 			ret = -EISCONN;
630 			if (rx->sk.sk_state != RXRPC_UNBOUND)
631 				goto error;
632 			ret = rxrpc_server_keyring(rx, optval, optlen);
633 			goto error;
634 
635 		case RXRPC_MIN_SECURITY_LEVEL:
636 			ret = -EINVAL;
637 			if (optlen != sizeof(unsigned int))
638 				goto error;
639 			ret = -EISCONN;
640 			if (rx->sk.sk_state != RXRPC_UNBOUND)
641 				goto error;
642 			ret = get_user(min_sec_level,
643 				       (unsigned int __user *) optval);
644 			if (ret < 0)
645 				goto error;
646 			ret = -EINVAL;
647 			if (min_sec_level > RXRPC_SECURITY_MAX)
648 				goto error;
649 			rx->min_sec_level = min_sec_level;
650 			goto success;
651 
652 		case RXRPC_UPGRADEABLE_SERVICE:
653 			ret = -EINVAL;
654 			if (optlen != sizeof(service_upgrade) ||
655 			    rx->service_upgrade.from != 0)
656 				goto error;
657 			ret = -EISCONN;
658 			if (rx->sk.sk_state != RXRPC_SERVER_BOUND2)
659 				goto error;
660 			ret = -EFAULT;
661 			if (copy_from_user(service_upgrade, optval,
662 					   sizeof(service_upgrade)) != 0)
663 				goto error;
664 			ret = -EINVAL;
665 			if ((service_upgrade[0] != rx->srx.srx_service ||
666 			     service_upgrade[1] != rx->second_service) &&
667 			    (service_upgrade[0] != rx->second_service ||
668 			     service_upgrade[1] != rx->srx.srx_service))
669 				goto error;
670 			rx->service_upgrade.from = service_upgrade[0];
671 			rx->service_upgrade.to = service_upgrade[1];
672 			goto success;
673 
674 		default:
675 			break;
676 		}
677 	}
678 
679 success:
680 	ret = 0;
681 error:
682 	release_sock(&rx->sk);
683 	return ret;
684 }
685 
686 /*
687  * Get socket options.
688  */
689 static int rxrpc_getsockopt(struct socket *sock, int level, int optname,
690 			    char __user *optval, int __user *_optlen)
691 {
692 	int optlen;
693 
694 	if (level != SOL_RXRPC)
695 		return -EOPNOTSUPP;
696 
697 	if (get_user(optlen, _optlen))
698 		return -EFAULT;
699 
700 	switch (optname) {
701 	case RXRPC_SUPPORTED_CMSG:
702 		if (optlen < sizeof(int))
703 			return -ETOOSMALL;
704 		if (put_user(RXRPC__SUPPORTED - 1, (int __user *)optval) ||
705 		    put_user(sizeof(int), _optlen))
706 			return -EFAULT;
707 		return 0;
708 
709 	default:
710 		return -EOPNOTSUPP;
711 	}
712 }
713 
714 /*
715  * permit an RxRPC socket to be polled
716  */
717 static __poll_t rxrpc_poll(struct file *file, struct socket *sock,
718 			       poll_table *wait)
719 {
720 	struct sock *sk = sock->sk;
721 	struct rxrpc_sock *rx = rxrpc_sk(sk);
722 	__poll_t mask;
723 
724 	sock_poll_wait(file, sock, wait);
725 	mask = 0;
726 
727 	/* the socket is readable if there are any messages waiting on the Rx
728 	 * queue */
729 	if (!list_empty(&rx->recvmsg_q))
730 		mask |= EPOLLIN | EPOLLRDNORM;
731 
732 	/* the socket is writable if there is space to add new data to the
733 	 * socket; there is no guarantee that any particular call in progress
734 	 * on the socket may have space in the Tx ACK window */
735 	if (rxrpc_writable(sk))
736 		mask |= EPOLLOUT | EPOLLWRNORM;
737 
738 	return mask;
739 }
740 
741 /*
742  * create an RxRPC socket
743  */
744 static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
745 			int kern)
746 {
747 	struct rxrpc_net *rxnet;
748 	struct rxrpc_sock *rx;
749 	struct sock *sk;
750 
751 	_enter("%p,%d", sock, protocol);
752 
753 	/* we support transport protocol UDP/UDP6 only */
754 	if (protocol != PF_INET &&
755 	    IS_ENABLED(CONFIG_AF_RXRPC_IPV6) && protocol != PF_INET6)
756 		return -EPROTONOSUPPORT;
757 
758 	if (sock->type != SOCK_DGRAM)
759 		return -ESOCKTNOSUPPORT;
760 
761 	sock->ops = &rxrpc_rpc_ops;
762 	sock->state = SS_UNCONNECTED;
763 
764 	sk = sk_alloc(net, PF_RXRPC, GFP_KERNEL, &rxrpc_proto, kern);
765 	if (!sk)
766 		return -ENOMEM;
767 
768 	sock_init_data(sock, sk);
769 	sock_set_flag(sk, SOCK_RCU_FREE);
770 	sk->sk_state		= RXRPC_UNBOUND;
771 	sk->sk_write_space	= rxrpc_write_space;
772 	sk->sk_max_ack_backlog	= 0;
773 	sk->sk_destruct		= rxrpc_sock_destructor;
774 
775 	rx = rxrpc_sk(sk);
776 	rx->family = protocol;
777 	rx->calls = RB_ROOT;
778 
779 	spin_lock_init(&rx->incoming_lock);
780 	INIT_LIST_HEAD(&rx->sock_calls);
781 	INIT_LIST_HEAD(&rx->to_be_accepted);
782 	INIT_LIST_HEAD(&rx->recvmsg_q);
783 	rwlock_init(&rx->recvmsg_lock);
784 	rwlock_init(&rx->call_lock);
785 	memset(&rx->srx, 0, sizeof(rx->srx));
786 
787 	rxnet = rxrpc_net(sock_net(&rx->sk));
788 	timer_reduce(&rxnet->peer_keepalive_timer, jiffies + 1);
789 
790 	_leave(" = 0 [%p]", rx);
791 	return 0;
792 }
793 
794 /*
795  * Kill all the calls on a socket and shut it down.
796  */
797 static int rxrpc_shutdown(struct socket *sock, int flags)
798 {
799 	struct sock *sk = sock->sk;
800 	struct rxrpc_sock *rx = rxrpc_sk(sk);
801 	int ret = 0;
802 
803 	_enter("%p,%d", sk, flags);
804 
805 	if (flags != SHUT_RDWR)
806 		return -EOPNOTSUPP;
807 	if (sk->sk_state == RXRPC_CLOSE)
808 		return -ESHUTDOWN;
809 
810 	lock_sock(sk);
811 
812 	spin_lock_bh(&sk->sk_receive_queue.lock);
813 	if (sk->sk_state < RXRPC_CLOSE) {
814 		sk->sk_state = RXRPC_CLOSE;
815 		sk->sk_shutdown = SHUTDOWN_MASK;
816 	} else {
817 		ret = -ESHUTDOWN;
818 	}
819 	spin_unlock_bh(&sk->sk_receive_queue.lock);
820 
821 	rxrpc_discard_prealloc(rx);
822 
823 	release_sock(sk);
824 	return ret;
825 }
826 
827 /*
828  * RxRPC socket destructor
829  */
830 static void rxrpc_sock_destructor(struct sock *sk)
831 {
832 	_enter("%p", sk);
833 
834 	rxrpc_purge_queue(&sk->sk_receive_queue);
835 
836 	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
837 	WARN_ON(!sk_unhashed(sk));
838 	WARN_ON(sk->sk_socket);
839 
840 	if (!sock_flag(sk, SOCK_DEAD)) {
841 		printk("Attempt to release alive rxrpc socket: %p\n", sk);
842 		return;
843 	}
844 }
845 
846 /*
847  * release an RxRPC socket
848  */
849 static int rxrpc_release_sock(struct sock *sk)
850 {
851 	struct rxrpc_sock *rx = rxrpc_sk(sk);
852 
853 	_enter("%p{%d,%d}", sk, sk->sk_state, refcount_read(&sk->sk_refcnt));
854 
855 	/* declare the socket closed for business */
856 	sock_orphan(sk);
857 	sk->sk_shutdown = SHUTDOWN_MASK;
858 
859 	/* We want to kill off all connections from a service socket
860 	 * as fast as possible because we can't share these; client
861 	 * sockets, on the other hand, can share an endpoint.
862 	 */
863 	switch (sk->sk_state) {
864 	case RXRPC_SERVER_BOUND:
865 	case RXRPC_SERVER_BOUND2:
866 	case RXRPC_SERVER_LISTENING:
867 	case RXRPC_SERVER_LISTEN_DISABLED:
868 		rx->local->service_closed = true;
869 		break;
870 	}
871 
872 	spin_lock_bh(&sk->sk_receive_queue.lock);
873 	sk->sk_state = RXRPC_CLOSE;
874 	spin_unlock_bh(&sk->sk_receive_queue.lock);
875 
876 	if (rx->local && rcu_access_pointer(rx->local->service) == rx) {
877 		write_lock(&rx->local->services_lock);
878 		rcu_assign_pointer(rx->local->service, NULL);
879 		write_unlock(&rx->local->services_lock);
880 	}
881 
882 	/* try to flush out this socket */
883 	rxrpc_discard_prealloc(rx);
884 	rxrpc_release_calls_on_socket(rx);
885 	flush_workqueue(rxrpc_workqueue);
886 	rxrpc_purge_queue(&sk->sk_receive_queue);
887 
888 	rxrpc_unuse_local(rx->local);
889 	rxrpc_put_local(rx->local);
890 	rx->local = NULL;
891 	key_put(rx->key);
892 	rx->key = NULL;
893 	key_put(rx->securities);
894 	rx->securities = NULL;
895 	sock_put(sk);
896 
897 	_leave(" = 0");
898 	return 0;
899 }
900 
901 /*
902  * release an RxRPC BSD socket on close() or equivalent
903  */
904 static int rxrpc_release(struct socket *sock)
905 {
906 	struct sock *sk = sock->sk;
907 
908 	_enter("%p{%p}", sock, sk);
909 
910 	if (!sk)
911 		return 0;
912 
913 	sock->sk = NULL;
914 
915 	return rxrpc_release_sock(sk);
916 }
917 
918 /*
919  * RxRPC network protocol
920  */
921 static const struct proto_ops rxrpc_rpc_ops = {
922 	.family		= PF_RXRPC,
923 	.owner		= THIS_MODULE,
924 	.release	= rxrpc_release,
925 	.bind		= rxrpc_bind,
926 	.connect	= rxrpc_connect,
927 	.socketpair	= sock_no_socketpair,
928 	.accept		= sock_no_accept,
929 	.getname	= sock_no_getname,
930 	.poll		= rxrpc_poll,
931 	.ioctl		= sock_no_ioctl,
932 	.listen		= rxrpc_listen,
933 	.shutdown	= rxrpc_shutdown,
934 	.setsockopt	= rxrpc_setsockopt,
935 	.getsockopt	= rxrpc_getsockopt,
936 	.sendmsg	= rxrpc_sendmsg,
937 	.recvmsg	= rxrpc_recvmsg,
938 	.mmap		= sock_no_mmap,
939 	.sendpage	= sock_no_sendpage,
940 };
941 
942 static struct proto rxrpc_proto = {
943 	.name		= "RXRPC",
944 	.owner		= THIS_MODULE,
945 	.obj_size	= sizeof(struct rxrpc_sock),
946 	.max_header	= sizeof(struct rxrpc_wire_header),
947 };
948 
949 static const struct net_proto_family rxrpc_family_ops = {
950 	.family	= PF_RXRPC,
951 	.create = rxrpc_create,
952 	.owner	= THIS_MODULE,
953 };
954 
955 /*
956  * initialise and register the RxRPC protocol
957  */
958 static int __init af_rxrpc_init(void)
959 {
960 	int ret = -1;
961 	unsigned int tmp;
962 
963 	BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > sizeof_field(struct sk_buff, cb));
964 
965 	get_random_bytes(&tmp, sizeof(tmp));
966 	tmp &= 0x3fffffff;
967 	if (tmp == 0)
968 		tmp = 1;
969 	idr_set_cursor(&rxrpc_client_conn_ids, tmp);
970 
971 	ret = -ENOMEM;
972 	rxrpc_call_jar = kmem_cache_create(
973 		"rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
974 		SLAB_HWCACHE_ALIGN, NULL);
975 	if (!rxrpc_call_jar) {
976 		pr_notice("Failed to allocate call jar\n");
977 		goto error_call_jar;
978 	}
979 
980 	rxrpc_workqueue = alloc_workqueue("krxrpcd", 0, 1);
981 	if (!rxrpc_workqueue) {
982 		pr_notice("Failed to allocate work queue\n");
983 		goto error_work_queue;
984 	}
985 
986 	ret = rxrpc_init_security();
987 	if (ret < 0) {
988 		pr_crit("Cannot initialise security\n");
989 		goto error_security;
990 	}
991 
992 	ret = register_pernet_subsys(&rxrpc_net_ops);
993 	if (ret)
994 		goto error_pernet;
995 
996 	ret = proto_register(&rxrpc_proto, 1);
997 	if (ret < 0) {
998 		pr_crit("Cannot register protocol\n");
999 		goto error_proto;
1000 	}
1001 
1002 	ret = sock_register(&rxrpc_family_ops);
1003 	if (ret < 0) {
1004 		pr_crit("Cannot register socket family\n");
1005 		goto error_sock;
1006 	}
1007 
1008 	ret = register_key_type(&key_type_rxrpc);
1009 	if (ret < 0) {
1010 		pr_crit("Cannot register client key type\n");
1011 		goto error_key_type;
1012 	}
1013 
1014 	ret = register_key_type(&key_type_rxrpc_s);
1015 	if (ret < 0) {
1016 		pr_crit("Cannot register server key type\n");
1017 		goto error_key_type_s;
1018 	}
1019 
1020 	ret = rxrpc_sysctl_init();
1021 	if (ret < 0) {
1022 		pr_crit("Cannot register sysctls\n");
1023 		goto error_sysctls;
1024 	}
1025 
1026 	return 0;
1027 
1028 error_sysctls:
1029 	unregister_key_type(&key_type_rxrpc_s);
1030 error_key_type_s:
1031 	unregister_key_type(&key_type_rxrpc);
1032 error_key_type:
1033 	sock_unregister(PF_RXRPC);
1034 error_sock:
1035 	proto_unregister(&rxrpc_proto);
1036 error_proto:
1037 	unregister_pernet_subsys(&rxrpc_net_ops);
1038 error_pernet:
1039 	rxrpc_exit_security();
1040 error_security:
1041 	destroy_workqueue(rxrpc_workqueue);
1042 error_work_queue:
1043 	kmem_cache_destroy(rxrpc_call_jar);
1044 error_call_jar:
1045 	return ret;
1046 }
1047 
1048 /*
1049  * unregister the RxRPC protocol
1050  */
1051 static void __exit af_rxrpc_exit(void)
1052 {
1053 	_enter("");
1054 	rxrpc_sysctl_exit();
1055 	unregister_key_type(&key_type_rxrpc_s);
1056 	unregister_key_type(&key_type_rxrpc);
1057 	sock_unregister(PF_RXRPC);
1058 	proto_unregister(&rxrpc_proto);
1059 	unregister_pernet_subsys(&rxrpc_net_ops);
1060 	ASSERTCMP(atomic_read(&rxrpc_n_tx_skbs), ==, 0);
1061 	ASSERTCMP(atomic_read(&rxrpc_n_rx_skbs), ==, 0);
1062 
1063 	/* Make sure the local and peer records pinned by any dying connections
1064 	 * are released.
1065 	 */
1066 	rcu_barrier();
1067 	rxrpc_destroy_client_conn_ids();
1068 
1069 	destroy_workqueue(rxrpc_workqueue);
1070 	rxrpc_exit_security();
1071 	kmem_cache_destroy(rxrpc_call_jar);
1072 	_leave("");
1073 }
1074 
1075 module_init(af_rxrpc_init);
1076 module_exit(af_rxrpc_exit);
1077