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