xref: /linux/net/rxrpc/af_rxrpc.c (revision 90e63d5354951d37fa2b3b91e6f17b95d2bf9bee)
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 <linux/uio.h>
20 #include <net/net_namespace.h>
21 #include <net/sock.h>
22 #include <net/af_rxrpc.h>
23 #define CREATE_TRACE_POINTS
24 #include "ar-internal.h"
25 
26 MODULE_DESCRIPTION("RxRPC network protocol");
27 MODULE_AUTHOR("Red Hat, Inc.");
28 MODULE_LICENSE("GPL");
29 MODULE_ALIAS_NETPROTO(PF_RXRPC);
30 
31 unsigned int rxrpc_debug; // = RXRPC_DEBUG_KPROTO;
32 module_param_named(debug, rxrpc_debug, uint, 0644);
33 MODULE_PARM_DESC(debug, "RxRPC debugging mask");
34 
35 static struct proto rxrpc_proto;
36 static const struct proto_ops rxrpc_rpc_ops;
37 
38 /* current debugging ID */
39 atomic_t rxrpc_debug_id;
40 EXPORT_SYMBOL(rxrpc_debug_id);
41 
42 /* count of skbs currently in use */
43 atomic_t rxrpc_n_rx_skbs;
44 
45 struct workqueue_struct *rxrpc_workqueue;
46 
47 static void rxrpc_sock_destructor(struct sock *);
48 
49 /*
50  * see if an RxRPC socket is currently writable
51  */
52 static inline int rxrpc_writable(struct sock *sk)
53 {
54 	return refcount_read(&sk->sk_wmem_alloc) < (size_t) sk->sk_sndbuf;
55 }
56 
57 /*
58  * wait for write bufferage to become available
59  */
60 static void rxrpc_write_space(struct sock *sk)
61 {
62 	_enter("%p", sk);
63 	rcu_read_lock();
64 	if (rxrpc_writable(sk)) {
65 		struct socket_wq *wq = rcu_dereference(sk->sk_wq);
66 
67 		if (skwq_has_sleeper(wq))
68 			wake_up_interruptible(&wq->wait);
69 		sk_wake_async_rcu(sk, SOCK_WAKE_SPACE, POLL_OUT);
70 	}
71 	rcu_read_unlock();
72 }
73 
74 /*
75  * validate an RxRPC address
76  */
77 static int rxrpc_validate_address(struct rxrpc_sock *rx,
78 				  struct sockaddr_rxrpc *srx,
79 				  int len)
80 {
81 	unsigned int tail;
82 
83 	if (len < sizeof(struct sockaddr_rxrpc))
84 		return -EINVAL;
85 
86 	if (srx->srx_family != AF_RXRPC)
87 		return -EAFNOSUPPORT;
88 
89 	if (srx->transport_type != SOCK_DGRAM)
90 		return -ESOCKTNOSUPPORT;
91 
92 	len -= offsetof(struct sockaddr_rxrpc, transport);
93 	if (srx->transport_len < sizeof(sa_family_t) ||
94 	    srx->transport_len > len)
95 		return -EINVAL;
96 
97 	switch (srx->transport.family) {
98 	case AF_INET:
99 		if (rx->family != AF_INET &&
100 		    rx->family != AF_INET6)
101 			return -EAFNOSUPPORT;
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 (rx->family != AF_INET6)
110 			return -EAFNOSUPPORT;
111 		if (srx->transport_len < sizeof(struct sockaddr_in6))
112 			return -EINVAL;
113 		tail = offsetof(struct sockaddr_rxrpc, transport) +
114 			sizeof(struct sockaddr_in6);
115 		break;
116 #endif
117 
118 	default:
119 		return -EAFNOSUPPORT;
120 	}
121 
122 	if (tail < len)
123 		memset((void *)srx + tail, 0, len - tail);
124 	_debug("INET: %pISp", &srx->transport);
125 	return 0;
126 }
127 
128 /*
129  * bind a local address to an RxRPC socket
130  */
131 static int rxrpc_bind(struct socket *sock, struct sockaddr_unsized *saddr, int len)
132 {
133 	struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)saddr;
134 	struct rxrpc_local *local;
135 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
136 	u16 service_id;
137 	int ret;
138 
139 	_enter("%p,%p,%d", rx, saddr, len);
140 
141 	ret = rxrpc_validate_address(rx, srx, len);
142 	if (ret < 0)
143 		goto error;
144 	service_id = srx->srx_service;
145 
146 	lock_sock(&rx->sk);
147 
148 	switch (rx->sk.sk_state) {
149 	case RXRPC_UNBOUND:
150 		rx->srx = *srx;
151 		local = rxrpc_lookup_local(sock_net(&rx->sk), &rx->srx);
152 		if (IS_ERR(local)) {
153 			ret = PTR_ERR(local);
154 			goto error_unlock;
155 		}
156 
157 		if (service_id) {
158 			write_lock(&local->services_lock);
159 			if (local->service)
160 				goto service_in_use;
161 			rx->local = local;
162 			local->service = rx;
163 			write_unlock(&local->services_lock);
164 
165 			rx->sk.sk_state = RXRPC_SERVER_BOUND;
166 		} else {
167 			rx->local = local;
168 			rx->sk.sk_state = RXRPC_CLIENT_BOUND;
169 		}
170 		break;
171 
172 	case RXRPC_SERVER_BOUND:
173 		ret = -EINVAL;
174 		if (service_id == 0)
175 			goto error_unlock;
176 		ret = -EADDRINUSE;
177 		if (service_id == rx->srx.srx_service)
178 			goto error_unlock;
179 		ret = -EINVAL;
180 		srx->srx_service = rx->srx.srx_service;
181 		if (memcmp(srx, &rx->srx, sizeof(*srx)) != 0)
182 			goto error_unlock;
183 		rx->second_service = service_id;
184 		rx->sk.sk_state = RXRPC_SERVER_BOUND2;
185 		break;
186 
187 	default:
188 		ret = -EINVAL;
189 		goto error_unlock;
190 	}
191 
192 	release_sock(&rx->sk);
193 	_leave(" = 0");
194 	return 0;
195 
196 service_in_use:
197 	write_unlock(&local->services_lock);
198 	rxrpc_unuse_local(local, rxrpc_local_unuse_bind);
199 	rxrpc_put_local(local, rxrpc_local_put_bind);
200 	ret = -EADDRINUSE;
201 error_unlock:
202 	release_sock(&rx->sk);
203 error:
204 	_leave(" = %d", ret);
205 	return ret;
206 }
207 
208 /*
209  * set the number of pending calls permitted on a listening socket
210  */
211 static int rxrpc_listen(struct socket *sock, int backlog)
212 {
213 	struct sock *sk = sock->sk;
214 	struct rxrpc_sock *rx = rxrpc_sk(sk);
215 	unsigned int max, old;
216 	int ret;
217 
218 	_enter("%p,%d", rx, backlog);
219 
220 	lock_sock(&rx->sk);
221 
222 	switch (rx->sk.sk_state) {
223 	case RXRPC_UNBOUND:
224 		ret = -EADDRNOTAVAIL;
225 		break;
226 	case RXRPC_SERVER_BOUND:
227 	case RXRPC_SERVER_BOUND2:
228 		ASSERT(rx->local != NULL);
229 		max = READ_ONCE(rxrpc_max_backlog);
230 		ret = -EINVAL;
231 		if (backlog == INT_MAX)
232 			backlog = max;
233 		else if (backlog < 0 || backlog > max)
234 			break;
235 		old = sk->sk_max_ack_backlog;
236 		sk->sk_max_ack_backlog = backlog;
237 		ret = rxrpc_service_prealloc(rx, GFP_KERNEL);
238 		if (ret == 0)
239 			rx->sk.sk_state = RXRPC_SERVER_LISTENING;
240 		else
241 			sk->sk_max_ack_backlog = old;
242 		break;
243 	case RXRPC_SERVER_LISTENING:
244 		if (backlog == 0) {
245 			rx->sk.sk_state = RXRPC_SERVER_LISTEN_DISABLED;
246 			sk->sk_max_ack_backlog = 0;
247 			rxrpc_discard_prealloc(rx);
248 			ret = 0;
249 			break;
250 		}
251 		fallthrough;
252 	default:
253 		ret = -EBUSY;
254 		break;
255 	}
256 
257 	release_sock(&rx->sk);
258 	_leave(" = %d", ret);
259 	return ret;
260 }
261 
262 /**
263  * rxrpc_kernel_lookup_peer - Obtain remote transport endpoint for an address
264  * @sock: The socket through which it will be accessed
265  * @srx: The network address
266  * @gfp: Allocation flags
267  *
268  * Lookup or create a remote transport endpoint record for the specified
269  * address.
270  *
271  * Return: The peer record found with a reference or a negative error code if
272  * the address is invalid or unsupported.
273  */
274 struct rxrpc_peer *rxrpc_kernel_lookup_peer(struct socket *sock,
275 					    struct sockaddr_rxrpc *srx, gfp_t gfp)
276 {
277 	struct rxrpc_peer *peer;
278 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
279 	int ret;
280 
281 	ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
282 	if (ret < 0)
283 		return ERR_PTR(ret);
284 
285 	peer = rxrpc_lookup_peer(rx->local, srx, gfp);
286 	return peer ?: ERR_PTR(-ENOMEM);
287 }
288 EXPORT_SYMBOL(rxrpc_kernel_lookup_peer);
289 
290 /**
291  * rxrpc_kernel_get_peer - Get a reference on a peer
292  * @peer: The peer to get a reference on (may be NULL).
293  *
294  * Get a reference for a remote peer record (if not NULL).
295  *
296  * Return: The @peer argument.
297  */
298 struct rxrpc_peer *rxrpc_kernel_get_peer(struct rxrpc_peer *peer)
299 {
300 	return peer ? rxrpc_get_peer(peer, rxrpc_peer_get_application) : NULL;
301 }
302 EXPORT_SYMBOL(rxrpc_kernel_get_peer);
303 
304 /**
305  * rxrpc_kernel_put_peer - Allow a kernel app to drop a peer reference
306  * @peer: The peer to drop a ref on
307  *
308  * Drop a reference on a peer record.
309  */
310 void rxrpc_kernel_put_peer(struct rxrpc_peer *peer)
311 {
312 	rxrpc_put_peer(peer, rxrpc_peer_put_application);
313 }
314 EXPORT_SYMBOL(rxrpc_kernel_put_peer);
315 
316 /**
317  * rxrpc_kernel_begin_call - Allow a kernel service to begin a call
318  * @sock: The socket on which to make the call
319  * @peer: The peer to contact
320  * @key: The security context to use (defaults to socket setting)
321  * @user_call_ID: The ID to use
322  * @tx_total_len: Total length of data to transmit during the call (or -1)
323  * @hard_timeout: The maximum lifespan of the call in sec
324  * @gfp: The allocation constraints
325  * @notify_rx: Where to send notifications instead of socket queue
326  * @service_id: The ID of the service to contact
327  * @upgrade: Request service upgrade for call
328  * @interruptibility: The call is interruptible, or can be canceled.
329  * @debug_id: The debug ID for tracing to be assigned to the call
330  *
331  * Allow a kernel service to begin a call on the nominated socket.  This just
332  * sets up all the internal tracking structures and allocates connection and
333  * call IDs as appropriate.
334  *
335  * The default socket destination address and security may be overridden by
336  * supplying @srx and @key.
337  *
338  * Return: The new call or an error code.
339  */
340 struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
341 					   struct rxrpc_peer *peer,
342 					   struct key *key,
343 					   unsigned long user_call_ID,
344 					   s64 tx_total_len,
345 					   u32 hard_timeout,
346 					   gfp_t gfp,
347 					   rxrpc_notify_rx_t notify_rx,
348 					   u16 service_id,
349 					   bool upgrade,
350 					   enum rxrpc_interruptibility interruptibility,
351 					   unsigned int debug_id)
352 {
353 	struct rxrpc_conn_parameters cp;
354 	struct rxrpc_call_params p;
355 	struct rxrpc_call *call;
356 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
357 
358 	_enter(",,%x,%lx", key_serial(key), user_call_ID);
359 
360 	if (WARN_ON_ONCE(peer->local != rx->local))
361 		return ERR_PTR(-EIO);
362 
363 	lock_sock(&rx->sk);
364 
365 	if (!key)
366 		key = rx->key;
367 	if (key && !key->payload.data[0])
368 		key = NULL; /* a no-security key */
369 
370 	memset(&p, 0, sizeof(p));
371 	p.user_call_ID		= user_call_ID;
372 	p.tx_total_len		= tx_total_len;
373 	p.interruptibility	= interruptibility;
374 	p.kernel		= true;
375 	p.timeouts.hard		= hard_timeout;
376 
377 	memset(&cp, 0, sizeof(cp));
378 	cp.local		= rx->local;
379 	cp.peer			= peer;
380 	cp.key			= key;
381 	cp.security_level	= rx->min_sec_level;
382 	cp.exclusive		= false;
383 	cp.upgrade		= upgrade;
384 	cp.service_id		= service_id;
385 	call = rxrpc_new_client_call(rx, &cp, &p, gfp, debug_id);
386 	/* The socket has been unlocked. */
387 	if (!IS_ERR(call)) {
388 		call->notify_rx = notify_rx;
389 		mutex_unlock(&call->user_mutex);
390 	}
391 
392 	_leave(" = %p", call);
393 	return call;
394 }
395 EXPORT_SYMBOL(rxrpc_kernel_begin_call);
396 
397 /*
398  * Dummy function used to stop the notifier talking to recvmsg().
399  */
400 static void rxrpc_dummy_notify_rx(struct sock *sk, struct rxrpc_call *rxcall,
401 				  unsigned long call_user_ID)
402 {
403 }
404 
405 /**
406  * rxrpc_kernel_shutdown_call - Allow a kernel service to shut down a call it was using
407  * @sock: The socket the call is on
408  * @call: The call to end
409  *
410  * Allow a kernel service to shut down a call it was using.  The call must be
411  * complete before this is called (the call should be aborted if necessary).
412  */
413 void rxrpc_kernel_shutdown_call(struct socket *sock, struct rxrpc_call *call)
414 {
415 	_enter("%d{%d}", call->debug_id, refcount_read(&call->ref));
416 
417 	mutex_lock(&call->user_mutex);
418 	if (!test_bit(RXRPC_CALL_RELEASED, &call->flags)) {
419 		rxrpc_release_call(rxrpc_sk(sock->sk), call);
420 
421 		/* Make sure we're not going to call back into a kernel service */
422 		if (call->notify_rx) {
423 			spin_lock_irq(&call->notify_lock);
424 			call->notify_rx = rxrpc_dummy_notify_rx;
425 			spin_unlock_irq(&call->notify_lock);
426 		}
427 	}
428 	mutex_unlock(&call->user_mutex);
429 }
430 EXPORT_SYMBOL(rxrpc_kernel_shutdown_call);
431 
432 /**
433  * rxrpc_kernel_put_call - Release a reference to a call
434  * @sock: The socket the call is on
435  * @call: The call to put
436  *
437  * Drop the application's ref on an rxrpc call.
438  */
439 void rxrpc_kernel_put_call(struct socket *sock, struct rxrpc_call *call)
440 {
441 	rxrpc_put_call(call, rxrpc_call_put_kernel);
442 }
443 EXPORT_SYMBOL(rxrpc_kernel_put_call);
444 
445 /**
446  * rxrpc_kernel_check_life - Check to see whether a call is still alive
447  * @sock: The socket the call is on
448  * @call: The call to check
449  *
450  * Allow a kernel service to find out whether a call is still alive - whether
451  * it has completed successfully and all received data has been consumed.
452  *
453  * Return: %true if the call is still ongoing and %false if it has completed.
454  */
455 bool rxrpc_kernel_check_life(const struct socket *sock,
456 			     const struct rxrpc_call *call)
457 {
458 	if (!rxrpc_call_is_complete(call))
459 		return true;
460 	if (call->completion != RXRPC_CALL_SUCCEEDED)
461 		return false;
462 	return !skb_queue_empty(&call->recvmsg_queue);
463 }
464 EXPORT_SYMBOL(rxrpc_kernel_check_life);
465 
466 /**
467  * rxrpc_kernel_set_notifications - Set table of callback operations
468  * @sock: The socket to install table upon
469  * @app_ops: Callback operation table to set
470  *
471  * Allow a kernel service to set a table of event notifications on a socket.
472  */
473 void rxrpc_kernel_set_notifications(struct socket *sock,
474 				    const struct rxrpc_kernel_ops *app_ops)
475 {
476 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
477 
478 	rx->app_ops = app_ops;
479 }
480 EXPORT_SYMBOL(rxrpc_kernel_set_notifications);
481 
482 /*
483  * connect an RxRPC socket
484  * - this just targets it at a specific destination; no actual connection
485  *   negotiation takes place
486  */
487 static int rxrpc_connect(struct socket *sock, struct sockaddr_unsized *addr,
488 			 int addr_len, int flags)
489 {
490 	struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
491 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
492 	int ret;
493 
494 	_enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
495 
496 	ret = rxrpc_validate_address(rx, srx, addr_len);
497 	if (ret < 0) {
498 		_leave(" = %d [bad addr]", ret);
499 		return ret;
500 	}
501 
502 	lock_sock(&rx->sk);
503 
504 	ret = -EISCONN;
505 	if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
506 		goto error;
507 
508 	switch (rx->sk.sk_state) {
509 	case RXRPC_UNBOUND:
510 		rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
511 		break;
512 	case RXRPC_CLIENT_UNBOUND:
513 	case RXRPC_CLIENT_BOUND:
514 		break;
515 	default:
516 		ret = -EBUSY;
517 		goto error;
518 	}
519 
520 	rx->connect_srx = *srx;
521 	set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
522 	ret = 0;
523 
524 error:
525 	release_sock(&rx->sk);
526 	return ret;
527 }
528 
529 /*
530  * send a message through an RxRPC socket
531  * - in a client this does a number of things:
532  *   - finds/sets up a connection for the security specified (if any)
533  *   - initiates a call (ID in control data)
534  *   - ends the request phase of a call (if MSG_MORE is not set)
535  *   - sends a call data packet
536  *   - may send an abort (abort code in control data)
537  */
538 static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
539 {
540 	struct rxrpc_local *local;
541 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
542 	int ret;
543 
544 	_enter(",{%d},,%zu", rx->sk.sk_state, len);
545 
546 	if (m->msg_flags & MSG_OOB)
547 		return -EOPNOTSUPP;
548 
549 	if (m->msg_name) {
550 		ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
551 		if (ret < 0) {
552 			_leave(" = %d [bad addr]", ret);
553 			return ret;
554 		}
555 	}
556 
557 	lock_sock(&rx->sk);
558 
559 	switch (rx->sk.sk_state) {
560 	case RXRPC_UNBOUND:
561 	case RXRPC_CLIENT_UNBOUND:
562 		rx->srx.srx_family = AF_RXRPC;
563 		rx->srx.srx_service = 0;
564 		rx->srx.transport_type = SOCK_DGRAM;
565 		rx->srx.transport.family = rx->family;
566 		switch (rx->family) {
567 		case AF_INET:
568 			rx->srx.transport_len = sizeof(struct sockaddr_in);
569 			break;
570 #ifdef CONFIG_AF_RXRPC_IPV6
571 		case AF_INET6:
572 			rx->srx.transport_len = sizeof(struct sockaddr_in6);
573 			break;
574 #endif
575 		default:
576 			ret = -EAFNOSUPPORT;
577 			goto error_unlock;
578 		}
579 		local = rxrpc_lookup_local(sock_net(sock->sk), &rx->srx);
580 		if (IS_ERR(local)) {
581 			ret = PTR_ERR(local);
582 			goto error_unlock;
583 		}
584 
585 		rx->local = local;
586 		rx->sk.sk_state = RXRPC_CLIENT_BOUND;
587 		fallthrough;
588 
589 	case RXRPC_CLIENT_BOUND:
590 		if (!m->msg_name &&
591 		    test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
592 			m->msg_name = &rx->connect_srx;
593 			m->msg_namelen = sizeof(rx->connect_srx);
594 		}
595 		fallthrough;
596 	case RXRPC_SERVER_BOUND:
597 	case RXRPC_SERVER_LISTENING:
598 		if (m->msg_flags & MSG_OOB)
599 			ret = rxrpc_sendmsg_oob(rx, m, len);
600 		else
601 			ret = rxrpc_do_sendmsg(rx, m, len);
602 		/* The socket has been unlocked */
603 		goto out;
604 	default:
605 		ret = -EINVAL;
606 		goto error_unlock;
607 	}
608 
609 error_unlock:
610 	release_sock(&rx->sk);
611 out:
612 	_leave(" = %d", ret);
613 	return ret;
614 }
615 
616 int rxrpc_sock_set_min_security_level(struct sock *sk, unsigned int val)
617 {
618 	if (sk->sk_state != RXRPC_UNBOUND)
619 		return -EISCONN;
620 	if (val > RXRPC_SECURITY_MAX)
621 		return -EINVAL;
622 	lock_sock(sk);
623 	rxrpc_sk(sk)->min_sec_level = val;
624 	release_sock(sk);
625 	return 0;
626 }
627 EXPORT_SYMBOL(rxrpc_sock_set_min_security_level);
628 
629 /*
630  * set RxRPC socket options
631  */
632 static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
633 			    sockptr_t optval, unsigned int optlen)
634 {
635 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
636 	unsigned int min_sec_level, val;
637 	u16 service_upgrade[2];
638 	int ret;
639 
640 	_enter(",%d,%d,,%d", level, optname, optlen);
641 
642 	lock_sock(&rx->sk);
643 	ret = -EOPNOTSUPP;
644 
645 	if (level == SOL_RXRPC) {
646 		switch (optname) {
647 		case RXRPC_EXCLUSIVE_CONNECTION:
648 			ret = -EINVAL;
649 			if (optlen != 0)
650 				goto error;
651 			ret = -EISCONN;
652 			if (rx->sk.sk_state != RXRPC_UNBOUND)
653 				goto error;
654 			rx->exclusive = true;
655 			goto success;
656 
657 		case RXRPC_SECURITY_KEY:
658 			ret = -EISCONN;
659 			if (rx->sk.sk_state != RXRPC_UNBOUND)
660 				goto error;
661 			ret = rxrpc_request_key(rx, optval, optlen);
662 			goto error;
663 
664 		case RXRPC_SECURITY_KEYRING:
665 			ret = -EISCONN;
666 			if (rx->sk.sk_state != RXRPC_UNBOUND)
667 				goto error;
668 			ret = rxrpc_server_keyring(rx, optval, optlen);
669 			goto error;
670 
671 		case RXRPC_MIN_SECURITY_LEVEL:
672 			ret = -EINVAL;
673 			if (optlen != sizeof(unsigned int))
674 				goto error;
675 			ret = -EISCONN;
676 			if (rx->sk.sk_state != RXRPC_UNBOUND)
677 				goto error;
678 			ret = copy_safe_from_sockptr(&min_sec_level,
679 						     sizeof(min_sec_level),
680 						     optval, optlen);
681 			if (ret)
682 				goto error;
683 			ret = -EINVAL;
684 			if (min_sec_level > RXRPC_SECURITY_MAX)
685 				goto error;
686 			rx->min_sec_level = min_sec_level;
687 			goto success;
688 
689 		case RXRPC_UPGRADEABLE_SERVICE:
690 			ret = -EINVAL;
691 			if (optlen != sizeof(service_upgrade) ||
692 			    rx->service_upgrade.from != 0)
693 				goto error;
694 			ret = -EISCONN;
695 			if (rx->sk.sk_state != RXRPC_SERVER_BOUND2)
696 				goto error;
697 			ret = -EFAULT;
698 			if (copy_from_sockptr(service_upgrade, optval,
699 					   sizeof(service_upgrade)) != 0)
700 				goto error;
701 			ret = -EINVAL;
702 			if ((service_upgrade[0] != rx->srx.srx_service ||
703 			     service_upgrade[1] != rx->second_service) &&
704 			    (service_upgrade[0] != rx->second_service ||
705 			     service_upgrade[1] != rx->srx.srx_service))
706 				goto error;
707 			rx->service_upgrade.from = service_upgrade[0];
708 			rx->service_upgrade.to = service_upgrade[1];
709 			goto success;
710 
711 		case RXRPC_MANAGE_RESPONSE:
712 			ret = -EINVAL;
713 			if (optlen != sizeof(unsigned int))
714 				goto error;
715 			ret = -EISCONN;
716 			if (rx->sk.sk_state != RXRPC_UNBOUND)
717 				goto error;
718 			ret = copy_safe_from_sockptr(&val, sizeof(val),
719 						     optval, optlen);
720 			if (ret)
721 				goto error;
722 			ret = -EINVAL;
723 			if (val > 1)
724 				goto error;
725 			if (val)
726 				set_bit(RXRPC_SOCK_MANAGE_RESPONSE, &rx->flags);
727 			else
728 				clear_bit(RXRPC_SOCK_MANAGE_RESPONSE, &rx->flags);
729 			goto success;
730 
731 		default:
732 			break;
733 		}
734 	}
735 
736 success:
737 	ret = 0;
738 error:
739 	release_sock(&rx->sk);
740 	return ret;
741 }
742 
743 /*
744  * Get socket options.
745  */
746 static int rxrpc_getsockopt(struct socket *sock, int level, int optname,
747 			    sockopt_t *opt)
748 {
749 	int optlen, val;
750 
751 	if (level != SOL_RXRPC)
752 		return -EOPNOTSUPP;
753 
754 	optlen = opt->optlen;
755 
756 	switch (optname) {
757 	case RXRPC_SUPPORTED_CMSG:
758 		if (optlen < sizeof(int))
759 			return -ETOOSMALL;
760 		val = RXRPC__SUPPORTED - 1;
761 		if (copy_to_iter(&val, sizeof(val), &opt->iter_out) !=
762 		    sizeof(val))
763 			return -EFAULT;
764 		opt->optlen = sizeof(val);
765 		return 0;
766 
767 	default:
768 		return -EOPNOTSUPP;
769 	}
770 }
771 
772 /*
773  * permit an RxRPC socket to be polled
774  */
775 static __poll_t rxrpc_poll(struct file *file, struct socket *sock,
776 			       poll_table *wait)
777 {
778 	struct sock *sk = sock->sk;
779 	struct rxrpc_sock *rx = rxrpc_sk(sk);
780 	__poll_t mask;
781 
782 	sock_poll_wait(file, sock, wait);
783 	mask = 0;
784 
785 	/* the socket is readable if there are any messages waiting on the Rx
786 	 * queue */
787 	if (!list_empty(&rx->recvmsg_q))
788 		mask |= EPOLLIN | EPOLLRDNORM;
789 
790 	/* the socket is writable if there is space to add new data to the
791 	 * socket; there is no guarantee that any particular call in progress
792 	 * on the socket may have space in the Tx ACK window */
793 	if (rxrpc_writable(sk))
794 		mask |= EPOLLOUT | EPOLLWRNORM;
795 
796 	return mask;
797 }
798 
799 /*
800  * create an RxRPC socket
801  */
802 static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
803 			int kern)
804 {
805 	struct rxrpc_net *rxnet;
806 	struct rxrpc_sock *rx;
807 	struct sock *sk;
808 
809 	_enter("%p,%d", sock, protocol);
810 
811 	/* we support transport protocol UDP/UDP6 only */
812 	if (protocol != PF_INET &&
813 	    IS_ENABLED(CONFIG_AF_RXRPC_IPV6) && protocol != PF_INET6)
814 		return -EPROTONOSUPPORT;
815 
816 	if (sock->type != SOCK_DGRAM)
817 		return -ESOCKTNOSUPPORT;
818 
819 	sock->ops = &rxrpc_rpc_ops;
820 	sock->state = SS_UNCONNECTED;
821 
822 	sk = sk_alloc(net, PF_RXRPC, GFP_KERNEL, &rxrpc_proto, kern);
823 	if (!sk)
824 		return -ENOMEM;
825 
826 	sock_init_data(sock, sk);
827 	sock_set_flag(sk, SOCK_RCU_FREE);
828 	sk->sk_state		= RXRPC_UNBOUND;
829 	sk->sk_write_space	= rxrpc_write_space;
830 	sk->sk_max_ack_backlog	= 0;
831 	sk->sk_destruct		= rxrpc_sock_destructor;
832 
833 	rx = rxrpc_sk(sk);
834 	rx->family = protocol;
835 	rx->calls = RB_ROOT;
836 
837 	spin_lock_init(&rx->incoming_lock);
838 	skb_queue_head_init(&rx->recvmsg_oobq);
839 	rx->pending_oobq = RB_ROOT;
840 	INIT_LIST_HEAD(&rx->sock_calls);
841 	INIT_LIST_HEAD(&rx->to_be_accepted);
842 	INIT_LIST_HEAD(&rx->recvmsg_q);
843 	spin_lock_init(&rx->recvmsg_lock);
844 	rwlock_init(&rx->call_lock);
845 	memset(&rx->srx, 0, sizeof(rx->srx));
846 
847 	rxnet = rxrpc_net(sock_net(&rx->sk));
848 	timer_reduce(&rxnet->peer_keepalive_timer, jiffies + 1);
849 
850 	_leave(" = 0 [%p]", rx);
851 	return 0;
852 }
853 
854 /*
855  * Kill all the calls on a socket and shut it down.
856  */
857 static int rxrpc_shutdown(struct socket *sock, int flags)
858 {
859 	struct sock *sk = sock->sk;
860 	struct rxrpc_sock *rx = rxrpc_sk(sk);
861 	int ret = 0;
862 
863 	_enter("%p,%d", sk, flags);
864 
865 	if (flags != SHUT_RDWR)
866 		return -EOPNOTSUPP;
867 	if (sk->sk_state == RXRPC_CLOSE)
868 		return -ESHUTDOWN;
869 
870 	lock_sock(sk);
871 
872 	if (sk->sk_state < RXRPC_CLOSE) {
873 		spin_lock_irq(&rx->recvmsg_lock);
874 		sk->sk_state = RXRPC_CLOSE;
875 		sk->sk_shutdown = SHUTDOWN_MASK;
876 		spin_unlock_irq(&rx->recvmsg_lock);
877 	} else {
878 		ret = -ESHUTDOWN;
879 	}
880 
881 	rxrpc_discard_prealloc(rx);
882 
883 	release_sock(sk);
884 	return ret;
885 }
886 
887 /*
888  * Purge the out-of-band queue.
889  */
890 static void rxrpc_purge_oob_queue(struct sock *sk)
891 {
892 	struct rxrpc_sock *rx = rxrpc_sk(sk);
893 	struct sk_buff *skb;
894 
895 	while ((skb = skb_dequeue(&rx->recvmsg_oobq)))
896 		rxrpc_kernel_free_oob(skb);
897 	while (!RB_EMPTY_ROOT(&rx->pending_oobq)) {
898 		skb = rb_entry(rx->pending_oobq.rb_node, struct sk_buff, rbnode);
899 		rb_erase(&skb->rbnode, &rx->pending_oobq);
900 		rxrpc_kernel_free_oob(skb);
901 	}
902 }
903 
904 /*
905  * RxRPC socket destructor
906  */
907 static void rxrpc_sock_destructor(struct sock *sk)
908 {
909 	_enter("%p", sk);
910 
911 	rxrpc_purge_oob_queue(sk);
912 	rxrpc_purge_queue(&sk->sk_receive_queue);
913 
914 	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
915 	WARN_ON(!sk_unhashed(sk));
916 	WARN_ON(sk->sk_socket);
917 
918 	if (!sock_flag(sk, SOCK_DEAD)) {
919 		printk("Attempt to release alive rxrpc socket: %p\n", sk);
920 		return;
921 	}
922 }
923 
924 /*
925  * release an RxRPC socket
926  */
927 static int rxrpc_release_sock(struct sock *sk)
928 {
929 	struct rxrpc_sock *rx = rxrpc_sk(sk);
930 
931 	_enter("%p{%d,%d}", sk, sk->sk_state, refcount_read(&sk->sk_refcnt));
932 
933 	/* declare the socket closed for business */
934 	sock_orphan(sk);
935 	sk->sk_shutdown = SHUTDOWN_MASK;
936 
937 	/* We want to kill off all connections from a service socket
938 	 * as fast as possible because we can't share these; client
939 	 * sockets, on the other hand, can share an endpoint.
940 	 */
941 	switch (sk->sk_state) {
942 	case RXRPC_SERVER_BOUND:
943 	case RXRPC_SERVER_BOUND2:
944 	case RXRPC_SERVER_LISTENING:
945 	case RXRPC_SERVER_LISTEN_DISABLED:
946 		rx->local->service_closed = true;
947 		break;
948 	}
949 
950 	spin_lock_irq(&rx->recvmsg_lock);
951 	sk->sk_state = RXRPC_CLOSE;
952 	spin_unlock_irq(&rx->recvmsg_lock);
953 
954 	if (rx->local && rx->local->service == rx) {
955 		write_lock(&rx->local->services_lock);
956 		rx->local->service = NULL;
957 		write_unlock(&rx->local->services_lock);
958 	}
959 
960 	/* try to flush out this socket */
961 	rxrpc_discard_prealloc(rx);
962 	rxrpc_release_calls_on_socket(rx);
963 	flush_workqueue(rxrpc_workqueue);
964 	rxrpc_purge_oob_queue(sk);
965 	rxrpc_purge_queue(&sk->sk_receive_queue);
966 
967 	rxrpc_unuse_local(rx->local, rxrpc_local_unuse_release_sock);
968 	rxrpc_put_local(rx->local, rxrpc_local_put_release_sock);
969 	rx->local = NULL;
970 	key_put(rx->key);
971 	rx->key = NULL;
972 	key_put(rx->securities);
973 	rx->securities = NULL;
974 	sock_put(sk);
975 
976 	_leave(" = 0");
977 	return 0;
978 }
979 
980 /*
981  * release an RxRPC BSD socket on close() or equivalent
982  */
983 static int rxrpc_release(struct socket *sock)
984 {
985 	struct sock *sk = sock->sk;
986 
987 	_enter("%p{%p}", sock, sk);
988 
989 	if (!sk)
990 		return 0;
991 
992 	sock->sk = NULL;
993 
994 	return rxrpc_release_sock(sk);
995 }
996 
997 /*
998  * RxRPC network protocol
999  */
1000 static const struct proto_ops rxrpc_rpc_ops = {
1001 	.family		= PF_RXRPC,
1002 	.owner		= THIS_MODULE,
1003 	.release	= rxrpc_release,
1004 	.bind		= rxrpc_bind,
1005 	.connect	= rxrpc_connect,
1006 	.socketpair	= sock_no_socketpair,
1007 	.accept		= sock_no_accept,
1008 	.getname	= sock_no_getname,
1009 	.poll		= rxrpc_poll,
1010 	.ioctl		= sock_no_ioctl,
1011 	.listen		= rxrpc_listen,
1012 	.shutdown	= rxrpc_shutdown,
1013 	.setsockopt	= rxrpc_setsockopt,
1014 	.getsockopt_iter = rxrpc_getsockopt,
1015 	.sendmsg	= rxrpc_sendmsg,
1016 	.recvmsg	= rxrpc_recvmsg,
1017 	.mmap		= sock_no_mmap,
1018 };
1019 
1020 static struct proto rxrpc_proto = {
1021 	.name		= "RXRPC",
1022 	.owner		= THIS_MODULE,
1023 	.obj_size	= sizeof(struct rxrpc_sock),
1024 	.max_header	= sizeof(struct rxrpc_wire_header),
1025 };
1026 
1027 static const struct net_proto_family rxrpc_family_ops = {
1028 	.family	= PF_RXRPC,
1029 	.create = rxrpc_create,
1030 	.owner	= THIS_MODULE,
1031 };
1032 
1033 /*
1034  * initialise and register the RxRPC protocol
1035  */
1036 static int __init af_rxrpc_init(void)
1037 {
1038 	int ret = -1;
1039 
1040 	BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > sizeof_field(struct sk_buff, cb));
1041 
1042 	ret = -ENOMEM;
1043 	rxrpc_gen_version_string();
1044 	rxrpc_call_jar = kmem_cache_create(
1045 		"rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
1046 		SLAB_HWCACHE_ALIGN, NULL);
1047 	if (!rxrpc_call_jar) {
1048 		pr_notice("Failed to allocate call jar\n");
1049 		goto error_call_jar;
1050 	}
1051 
1052 	rxrpc_workqueue = alloc_ordered_workqueue("krxrpcd", WQ_HIGHPRI | WQ_MEM_RECLAIM);
1053 	if (!rxrpc_workqueue) {
1054 		pr_notice("Failed to allocate work queue\n");
1055 		goto error_work_queue;
1056 	}
1057 
1058 	ret = rxrpc_init_security();
1059 	if (ret < 0) {
1060 		pr_crit("Cannot initialise security\n");
1061 		goto error_security;
1062 	}
1063 
1064 	ret = register_pernet_device(&rxrpc_net_ops);
1065 	if (ret)
1066 		goto error_pernet;
1067 
1068 	ret = proto_register(&rxrpc_proto, 1);
1069 	if (ret < 0) {
1070 		pr_crit("Cannot register protocol\n");
1071 		goto error_proto;
1072 	}
1073 
1074 	ret = sock_register(&rxrpc_family_ops);
1075 	if (ret < 0) {
1076 		pr_crit("Cannot register socket family\n");
1077 		goto error_sock;
1078 	}
1079 
1080 	ret = register_key_type(&key_type_rxrpc);
1081 	if (ret < 0) {
1082 		pr_crit("Cannot register client key type\n");
1083 		goto error_key_type;
1084 	}
1085 
1086 	ret = register_key_type(&key_type_rxrpc_s);
1087 	if (ret < 0) {
1088 		pr_crit("Cannot register server key type\n");
1089 		goto error_key_type_s;
1090 	}
1091 
1092 	ret = rxrpc_sysctl_init();
1093 	if (ret < 0) {
1094 		pr_crit("Cannot register sysctls\n");
1095 		goto error_sysctls;
1096 	}
1097 
1098 	return 0;
1099 
1100 error_sysctls:
1101 	unregister_key_type(&key_type_rxrpc_s);
1102 error_key_type_s:
1103 	unregister_key_type(&key_type_rxrpc);
1104 error_key_type:
1105 	sock_unregister(PF_RXRPC);
1106 error_sock:
1107 	proto_unregister(&rxrpc_proto);
1108 error_proto:
1109 	unregister_pernet_device(&rxrpc_net_ops);
1110 error_pernet:
1111 	rxrpc_exit_security();
1112 error_security:
1113 	destroy_workqueue(rxrpc_workqueue);
1114 error_work_queue:
1115 	kmem_cache_destroy(rxrpc_call_jar);
1116 error_call_jar:
1117 	return ret;
1118 }
1119 
1120 /*
1121  * unregister the RxRPC protocol
1122  */
1123 static void __exit af_rxrpc_exit(void)
1124 {
1125 	_enter("");
1126 	rxrpc_sysctl_exit();
1127 	unregister_key_type(&key_type_rxrpc_s);
1128 	unregister_key_type(&key_type_rxrpc);
1129 	sock_unregister(PF_RXRPC);
1130 	proto_unregister(&rxrpc_proto);
1131 	unregister_pernet_device(&rxrpc_net_ops);
1132 	ASSERTCMP(atomic_read(&rxrpc_n_rx_skbs), ==, 0);
1133 
1134 	/* Make sure the local and peer records pinned by any dying connections
1135 	 * are released.
1136 	 */
1137 	rcu_barrier();
1138 
1139 	destroy_workqueue(rxrpc_workqueue);
1140 	rxrpc_exit_security();
1141 	kmem_cache_destroy(rxrpc_call_jar);
1142 	_leave("");
1143 }
1144 
1145 module_init(af_rxrpc_init);
1146 module_exit(af_rxrpc_exit);
1147