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/* AF_RXRPC implementation
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 */
11
12#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13
14#include <linux/module.h>
15#include <linux/kernel.h>
16#include <linux/net.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/random.h>
20#include <linux/poll.h>
21#include <linux/proc_fs.h>
22#include <linux/key-type.h>
23#include <net/net_namespace.h>
24#include <net/sock.h>
25#include <net/af_rxrpc.h>
26#define CREATE_TRACE_POINTS
27#include "ar-internal.h"
28
29MODULE_DESCRIPTION("RxRPC network protocol");
30MODULE_AUTHOR("Red Hat, Inc.");
31MODULE_LICENSE("GPL");
32MODULE_ALIAS_NETPROTO(PF_RXRPC);
33
34unsigned int rxrpc_debug; // = RXRPC_DEBUG_KPROTO;
35module_param_named(debug, rxrpc_debug, uint, 0644);
36MODULE_PARM_DESC(debug, "RxRPC debugging mask");
37
38static struct proto rxrpc_proto;
39static const struct proto_ops rxrpc_rpc_ops;
40
41/* current debugging ID */
42atomic_t rxrpc_debug_id;
43EXPORT_SYMBOL(rxrpc_debug_id);
44
45/* count of skbs currently in use */
46atomic_t rxrpc_n_tx_skbs, rxrpc_n_rx_skbs;
47
48struct workqueue_struct *rxrpc_workqueue;
49
50static void rxrpc_sock_destructor(struct sock *);
51
52/*
53 * see if an RxRPC socket is currently writable
54 */
55static inline int rxrpc_writable(struct sock *sk)
56{
57 return refcount_read(&sk->sk_wmem_alloc) < (size_t) sk->sk_sndbuf;
58}
59
60/*
61 * wait for write bufferage to become available
62 */
63static void rxrpc_write_space(struct sock *sk)
64{
65 _enter("%p", sk);
66 rcu_read_lock();
67 if (rxrpc_writable(sk)) {
68 struct socket_wq *wq = rcu_dereference(sk->sk_wq);
69
70 if (skwq_has_sleeper(wq))
71 wake_up_interruptible(&wq->wait);
72 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
73 }
74 rcu_read_unlock();
75}
76
77/*
78 * validate an RxRPC address
79 */
80static int rxrpc_validate_address(struct rxrpc_sock *rx,
81 struct sockaddr_rxrpc *srx,
82 int len)
83{
84 unsigned int tail;
85
86 if (len < sizeof(struct sockaddr_rxrpc))
87 return -EINVAL;
88
89 if (srx->srx_family != AF_RXRPC)
90 return -EAFNOSUPPORT;
91
92 if (srx->transport_type != SOCK_DGRAM)
93 return -ESOCKTNOSUPPORT;
94
95 len -= offsetof(struct sockaddr_rxrpc, transport);
96 if (srx->transport_len < sizeof(sa_family_t) ||
97 srx->transport_len > len)
98 return -EINVAL;
99
100 if (srx->transport.family != rx->family)
101 return -EAFNOSUPPORT;
102
103 switch (srx->transport.family) {
104 case AF_INET:
105 if (srx->transport_len < sizeof(struct sockaddr_in))
106 return -EINVAL;
107 tail = offsetof(struct sockaddr_rxrpc, transport.sin.__pad);
108 break;
109
110#ifdef CONFIG_AF_RXRPC_IPV6
111 case AF_INET6:
112 if (srx->transport_len < sizeof(struct sockaddr_in6))
113 return -EINVAL;
114 tail = offsetof(struct sockaddr_rxrpc, transport) +
115 sizeof(struct sockaddr_in6);
116 break;
117#endif
118
119 default:
120 return -EAFNOSUPPORT;
121 }
122
123 if (tail < len)
124 memset((void *)srx + tail, 0, len - tail);
125 _debug("INET: %pISp", &srx->transport);
126 return 0;
127}
128
129/*
130 * bind a local address to an RxRPC socket
131 */
132static int rxrpc_bind(struct socket *sock, struct sockaddr *saddr, int len)
133{
134 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)saddr;
135 struct rxrpc_local *local;
136 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
137 u16 service_id = srx->srx_service;
138 int ret;
139
140 _enter("%p,%p,%d", rx, saddr, len);
141
142 ret = rxrpc_validate_address(rx, srx, len);
143 if (ret < 0)
144 goto error;
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 (rcu_access_pointer(local->service))
160 goto service_in_use;
161 rx->local = local;
162 rcu_assign_pointer(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
196service_in_use:
197 write_unlock(&local->services_lock);
198 rxrpc_put_local(local);
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 /* Fall through */
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 * @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 unsigned int debug_id)
289{
290 struct rxrpc_conn_parameters cp;
291 struct rxrpc_call_params p;
292 struct rxrpc_call *call;
293 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
294 int ret;
295
296 _enter(",,%x,%lx", key_serial(key), user_call_ID);
297
298 ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
299 if (ret < 0)
300 return ERR_PTR(ret);
301
302 lock_sock(&rx->sk);
303
304 if (!key)
305 key = rx->key;
306 if (key && !key->payload.data[0])
307 key = NULL; /* a no-security key */
308
309 memset(&p, 0, sizeof(p));
310 p.user_call_ID = user_call_ID;
311 p.tx_total_len = tx_total_len;
312
313 memset(&cp, 0, sizeof(cp));
314 cp.local = rx->local;
315 cp.key = key;
316 cp.security_level = rx->min_sec_level;
317 cp.exclusive = false;
318 cp.upgrade = upgrade;
319 cp.service_id = srx->srx_service;
320 call = rxrpc_new_client_call(rx, &cp, srx, &p, gfp, debug_id);
321 /* The socket has been unlocked. */
322 if (!IS_ERR(call)) {
323 call->notify_rx = notify_rx;
324 mutex_unlock(&call->user_mutex);
325 }
326
327 rxrpc_put_peer(cp.peer);
328 _leave(" = %p", call);
329 return call;
330}
331EXPORT_SYMBOL(rxrpc_kernel_begin_call);
332
333/*
334 * Dummy function used to stop the notifier talking to recvmsg().
335 */
336static void rxrpc_dummy_notify_rx(struct sock *sk, struct rxrpc_call *rxcall,
337 unsigned long call_user_ID)
338{
339}
340
341/**
342 * rxrpc_kernel_end_call - Allow a kernel service to end a call it was using
343 * @sock: The socket the call is on
344 * @call: The call to end
345 *
346 * Allow a kernel service to end a call it was using. The call must be
347 * complete before this is called (the call should be aborted if necessary).
348 */
349void rxrpc_kernel_end_call(struct socket *sock, struct rxrpc_call *call)
350{
351 _enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
352
353 mutex_lock(&call->user_mutex);
354 rxrpc_release_call(rxrpc_sk(sock->sk), call);
355
356 /* Make sure we're not going to call back into a kernel service */
357 if (call->notify_rx) {
358 spin_lock_bh(&call->notify_lock);
359 call->notify_rx = rxrpc_dummy_notify_rx;
360 spin_unlock_bh(&call->notify_lock);
361 }
362
363 mutex_unlock(&call->user_mutex);
364 rxrpc_put_call(call, rxrpc_call_put_kernel);
365}
366EXPORT_SYMBOL(rxrpc_kernel_end_call);
367
368/**
369 * rxrpc_kernel_check_life - Check to see whether a call is still alive
370 * @sock: The socket the call is on
371 * @call: The call to check
372 *
373 * Allow a kernel service to find out whether a call is still alive - ie. we're
374 * getting ACKs from the server. Returns a number representing the life state
375 * which can be compared to that returned by a previous call.
376 *
377 * If this is a client call, ping ACKs will be sent to the server to find out
378 * whether it's still responsive and whether the call is still alive on the
379 * server.
380 */
381u32 rxrpc_kernel_check_life(struct socket *sock, struct rxrpc_call *call)
382{
383 return call->acks_latest;
384}
385EXPORT_SYMBOL(rxrpc_kernel_check_life);
386
387/**
388 * rxrpc_kernel_check_call - Check a call's state
389 * @sock: The socket the call is on
390 * @call: The call to check
391 * @_compl: Where to store the completion state
392 * @_abort_code: Where to store any abort code
393 *
394 * Allow a kernel service to query the state of a call and find out the manner
395 * of its termination if it has completed. Returns -EINPROGRESS if the call is
396 * still going, 0 if the call finished successfully, -ECONNABORTED if the call
397 * was aborted and an appropriate error if the call failed in some other way.
398 */
399int rxrpc_kernel_check_call(struct socket *sock, struct rxrpc_call *call,
400 enum rxrpc_call_completion *_compl, u32 *_abort_code)
401{
402 if (call->state != RXRPC_CALL_COMPLETE)
403 return -EINPROGRESS;
404 smp_rmb();
405 *_compl = call->completion;
406 *_abort_code = call->abort_code;
407 return call->error;
408}
409EXPORT_SYMBOL(rxrpc_kernel_check_call);
410
411/**
412 * rxrpc_kernel_retry_call - Allow a kernel service to retry a call
413 * @sock: The socket the call is on
414 * @call: The call to retry
415 * @srx: The address of the peer to contact
416 * @key: The security context to use (defaults to socket setting)
417 *
418 * Allow a kernel service to try resending a client call that failed due to a
419 * network error to a new address. The Tx queue is maintained intact, thereby
420 * relieving the need to re-encrypt any request data that has already been
421 * buffered.
422 */
423int rxrpc_kernel_retry_call(struct socket *sock, struct rxrpc_call *call,
424 struct sockaddr_rxrpc *srx, struct key *key)
425{
426 struct rxrpc_conn_parameters cp;
427 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
428 int ret;
429
430 _enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
431
432 if (!key)
433 key = rx->key;
434 if (key && !key->payload.data[0])
435 key = NULL; /* a no-security key */
436
437 memset(&cp, 0, sizeof(cp));
438 cp.local = rx->local;
439 cp.key = key;
440 cp.security_level = 0;
441 cp.exclusive = false;
442 cp.service_id = srx->srx_service;
443
444 mutex_lock(&call->user_mutex);
445
446 ret = rxrpc_prepare_call_for_retry(rx, call);
447 if (ret == 0)
448 ret = rxrpc_retry_client_call(rx, call, &cp, srx, GFP_KERNEL);
449
450 mutex_unlock(&call->user_mutex);
451 rxrpc_put_peer(cp.peer);
452 _leave(" = %d", ret);
453 return ret;
454}
455EXPORT_SYMBOL(rxrpc_kernel_retry_call);
456
457/**
458 * rxrpc_kernel_new_call_notification - Get notifications of new calls
459 * @sock: The socket to intercept received messages on
460 * @notify_new_call: Function to be called when new calls appear
461 * @discard_new_call: Function to discard preallocated calls
462 *
463 * Allow a kernel service to be given notifications about new calls.
464 */
465void rxrpc_kernel_new_call_notification(
466 struct socket *sock,
467 rxrpc_notify_new_call_t notify_new_call,
468 rxrpc_discard_new_call_t discard_new_call)
469{
470 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
471
472 rx->notify_new_call = notify_new_call;
473 rx->discard_new_call = discard_new_call;
474}
475EXPORT_SYMBOL(rxrpc_kernel_new_call_notification);
476
477/*
478 * connect an RxRPC socket
479 * - this just targets it at a specific destination; no actual connection
480 * negotiation takes place
481 */
482static int rxrpc_connect(struct socket *sock, struct sockaddr *addr,
483 int addr_len, int flags)
484{
485 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
486 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
487 int ret;
488
489 _enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
490
491 ret = rxrpc_validate_address(rx, srx, addr_len);
492 if (ret < 0) {
493 _leave(" = %d [bad addr]", ret);
494 return ret;
495 }
496
497 lock_sock(&rx->sk);
498
499 ret = -EISCONN;
500 if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
501 goto error;
502
503 switch (rx->sk.sk_state) {
504 case RXRPC_UNBOUND:
505 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
506 case RXRPC_CLIENT_UNBOUND:
507 case RXRPC_CLIENT_BOUND:
508 break;
509 default:
510 ret = -EBUSY;
511 goto error;
512 }
513
514 rx->connect_srx = *srx;
515 set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
516 ret = 0;
517
518error:
519 release_sock(&rx->sk);
520 return ret;
521}
522
523/*
524 * send a message through an RxRPC socket
525 * - in a client this does a number of things:
526 * - finds/sets up a connection for the security specified (if any)
527 * - initiates a call (ID in control data)
528 * - ends the request phase of a call (if MSG_MORE is not set)
529 * - sends a call data packet
530 * - may send an abort (abort code in control data)
531 */
532static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
533{
534 struct rxrpc_local *local;
535 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
536 int ret;
537
538 _enter(",{%d},,%zu", rx->sk.sk_state, len);
539
540 if (m->msg_flags & MSG_OOB)
541 return -EOPNOTSUPP;
542
543 if (m->msg_name) {
544 ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
545 if (ret < 0) {
546 _leave(" = %d [bad addr]", ret);
547 return ret;
548 }
549 }
550
551 lock_sock(&rx->sk);
552
553 switch (rx->sk.sk_state) {
554 case RXRPC_UNBOUND:
555 rx->srx.srx_family = AF_RXRPC;
556 rx->srx.srx_service = 0;
557 rx->srx.transport_type = SOCK_DGRAM;
558 rx->srx.transport.family = rx->family;
559 switch (rx->family) {
560 case AF_INET:
561 rx->srx.transport_len = sizeof(struct sockaddr_in);
562 break;
563#ifdef CONFIG_AF_RXRPC_IPV6
564 case AF_INET6:
565 rx->srx.transport_len = sizeof(struct sockaddr_in6);
566 break;
567#endif
568 default:
569 ret = -EAFNOSUPPORT;
570 goto error_unlock;
571 }
572 local = rxrpc_lookup_local(sock_net(sock->sk), &rx->srx);
573 if (IS_ERR(local)) {
574 ret = PTR_ERR(local);
575 goto error_unlock;
576 }
577
578 rx->local = local;
579 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
580 /* Fall through */
581
582 case RXRPC_CLIENT_UNBOUND:
583 case RXRPC_CLIENT_BOUND:
584 if (!m->msg_name &&
585 test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
586 m->msg_name = &rx->connect_srx;
587 m->msg_namelen = sizeof(rx->connect_srx);
588 }
589 /* Fall through */
590 case RXRPC_SERVER_BOUND:
591 case RXRPC_SERVER_LISTENING:
592 ret = rxrpc_do_sendmsg(rx, m, len);
593 /* The socket has been unlocked */
594 goto out;
595 default:
596 ret = -EINVAL;
597 goto error_unlock;
598 }
599
600error_unlock:
601 release_sock(&rx->sk);
602out:
603 _leave(" = %d", ret);
604 return ret;
605}
606
607/*
608 * set RxRPC socket options
609 */
610static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
611 char __user *optval, unsigned int optlen)
612{
613 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
614 unsigned int min_sec_level;
615 u16 service_upgrade[2];
616 int ret;
617
618 _enter(",%d,%d,,%d", level, optname, optlen);
619
620 lock_sock(&rx->sk);
621 ret = -EOPNOTSUPP;
622
623 if (level == SOL_RXRPC) {
624 switch (optname) {
625 case RXRPC_EXCLUSIVE_CONNECTION:
626 ret = -EINVAL;
627 if (optlen != 0)
628 goto error;
629 ret = -EISCONN;
630 if (rx->sk.sk_state != RXRPC_UNBOUND)
631 goto error;
632 rx->exclusive = true;
633 goto success;
634
635 case RXRPC_SECURITY_KEY:
636 ret = -EINVAL;
637 if (rx->key)
638 goto error;
639 ret = -EISCONN;
640 if (rx->sk.sk_state != RXRPC_UNBOUND)
641 goto error;
642 ret = rxrpc_request_key(rx, optval, optlen);
643 goto error;
644
645 case RXRPC_SECURITY_KEYRING:
646 ret = -EINVAL;
647 if (rx->key)
648 goto error;
649 ret = -EISCONN;
650 if (rx->sk.sk_state != RXRPC_UNBOUND)
651 goto error;
652 ret = rxrpc_server_keyring(rx, optval, optlen);
653 goto error;
654
655 case RXRPC_MIN_SECURITY_LEVEL:
656 ret = -EINVAL;
657 if (optlen != sizeof(unsigned int))
658 goto error;
659 ret = -EISCONN;
660 if (rx->sk.sk_state != RXRPC_UNBOUND)
661 goto error;
662 ret = get_user(min_sec_level,
663 (unsigned int __user *) optval);
664 if (ret < 0)
665 goto error;
666 ret = -EINVAL;
667 if (min_sec_level > RXRPC_SECURITY_MAX)
668 goto error;
669 rx->min_sec_level = min_sec_level;
670 goto success;
671
672 case RXRPC_UPGRADEABLE_SERVICE:
673 ret = -EINVAL;
674 if (optlen != sizeof(service_upgrade) ||
675 rx->service_upgrade.from != 0)
676 goto error;
677 ret = -EISCONN;
678 if (rx->sk.sk_state != RXRPC_SERVER_BOUND2)
679 goto error;
680 ret = -EFAULT;
681 if (copy_from_user(service_upgrade, optval,
682 sizeof(service_upgrade)) != 0)
683 goto error;
684 ret = -EINVAL;
685 if ((service_upgrade[0] != rx->srx.srx_service ||
686 service_upgrade[1] != rx->second_service) &&
687 (service_upgrade[0] != rx->second_service ||
688 service_upgrade[1] != rx->srx.srx_service))
689 goto error;
690 rx->service_upgrade.from = service_upgrade[0];
691 rx->service_upgrade.to = service_upgrade[1];
692 goto success;
693
694 default:
695 break;
696 }
697 }
698
699success:
700 ret = 0;
701error:
702 release_sock(&rx->sk);
703 return ret;
704}
705
706/*
707 * Get socket options.
708 */
709static int rxrpc_getsockopt(struct socket *sock, int level, int optname,
710 char __user *optval, int __user *_optlen)
711{
712 int optlen;
713
714 if (level != SOL_RXRPC)
715 return -EOPNOTSUPP;
716
717 if (get_user(optlen, _optlen))
718 return -EFAULT;
719
720 switch (optname) {
721 case RXRPC_SUPPORTED_CMSG:
722 if (optlen < sizeof(int))
723 return -ETOOSMALL;
724 if (put_user(RXRPC__SUPPORTED - 1, (int __user *)optval) ||
725 put_user(sizeof(int), _optlen))
726 return -EFAULT;
727 return 0;
728
729 default:
730 return -EOPNOTSUPP;
731 }
732}
733
734/*
735 * permit an RxRPC socket to be polled
736 */
737static __poll_t rxrpc_poll(struct file *file, struct socket *sock,
738 poll_table *wait)
739{
740 struct sock *sk = sock->sk;
741 struct rxrpc_sock *rx = rxrpc_sk(sk);
742 __poll_t mask;
743
744 sock_poll_wait(file, sk_sleep(sk), wait);
745 mask = 0;
746
747 /* the socket is readable if there are any messages waiting on the Rx
748 * queue */
749 if (!list_empty(&rx->recvmsg_q))
750 mask |= EPOLLIN | EPOLLRDNORM;
751
752 /* the socket is writable if there is space to add new data to the
753 * socket; there is no guarantee that any particular call in progress
754 * on the socket may have space in the Tx ACK window */
755 if (rxrpc_writable(sk))
756 mask |= EPOLLOUT | EPOLLWRNORM;
757
758 return mask;
759}
760
761/*
762 * create an RxRPC socket
763 */
764static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
765 int kern)
766{
767 struct rxrpc_net *rxnet;
768 struct rxrpc_sock *rx;
769 struct sock *sk;
770
771 _enter("%p,%d", sock, protocol);
772
773 /* we support transport protocol UDP/UDP6 only */
774 if (protocol != PF_INET &&
775 IS_ENABLED(CONFIG_AF_RXRPC_IPV6) && protocol != PF_INET6)
776 return -EPROTONOSUPPORT;
777
778 if (sock->type != SOCK_DGRAM)
779 return -ESOCKTNOSUPPORT;
780
781 sock->ops = &rxrpc_rpc_ops;
782 sock->state = SS_UNCONNECTED;
783
784 sk = sk_alloc(net, PF_RXRPC, GFP_KERNEL, &rxrpc_proto, kern);
785 if (!sk)
786 return -ENOMEM;
787
788 sock_init_data(sock, sk);
789 sock_set_flag(sk, SOCK_RCU_FREE);
790 sk->sk_state = RXRPC_UNBOUND;
791 sk->sk_write_space = rxrpc_write_space;
792 sk->sk_max_ack_backlog = 0;
793 sk->sk_destruct = rxrpc_sock_destructor;
794
795 rx = rxrpc_sk(sk);
796 rx->family = protocol;
797 rx->calls = RB_ROOT;
798
799 spin_lock_init(&rx->incoming_lock);
800 INIT_LIST_HEAD(&rx->sock_calls);
801 INIT_LIST_HEAD(&rx->to_be_accepted);
802 INIT_LIST_HEAD(&rx->recvmsg_q);
803 rwlock_init(&rx->recvmsg_lock);
804 rwlock_init(&rx->call_lock);
805 memset(&rx->srx, 0, sizeof(rx->srx));
806
807 rxnet = rxrpc_net(sock_net(&rx->sk));
808 timer_reduce(&rxnet->peer_keepalive_timer, jiffies + 1);
809
810 _leave(" = 0 [%p]", rx);
811 return 0;
812}
813
814/*
815 * Kill all the calls on a socket and shut it down.
816 */
817static int rxrpc_shutdown(struct socket *sock, int flags)
818{
819 struct sock *sk = sock->sk;
820 struct rxrpc_sock *rx = rxrpc_sk(sk);
821 int ret = 0;
822
823 _enter("%p,%d", sk, flags);
824
825 if (flags != SHUT_RDWR)
826 return -EOPNOTSUPP;
827 if (sk->sk_state == RXRPC_CLOSE)
828 return -ESHUTDOWN;
829
830 lock_sock(sk);
831
832 spin_lock_bh(&sk->sk_receive_queue.lock);
833 if (sk->sk_state < RXRPC_CLOSE) {
834 sk->sk_state = RXRPC_CLOSE;
835 sk->sk_shutdown = SHUTDOWN_MASK;
836 } else {
837 ret = -ESHUTDOWN;
838 }
839 spin_unlock_bh(&sk->sk_receive_queue.lock);
840
841 rxrpc_discard_prealloc(rx);
842
843 release_sock(sk);
844 return ret;
845}
846
847/*
848 * RxRPC socket destructor
849 */
850static void rxrpc_sock_destructor(struct sock *sk)
851{
852 _enter("%p", sk);
853
854 rxrpc_purge_queue(&sk->sk_receive_queue);
855
856 WARN_ON(refcount_read(&sk->sk_wmem_alloc));
857 WARN_ON(!sk_unhashed(sk));
858 WARN_ON(sk->sk_socket);
859
860 if (!sock_flag(sk, SOCK_DEAD)) {
861 printk("Attempt to release alive rxrpc socket: %p\n", sk);
862 return;
863 }
864}
865
866/*
867 * release an RxRPC socket
868 */
869static int rxrpc_release_sock(struct sock *sk)
870{
871 struct rxrpc_sock *rx = rxrpc_sk(sk);
872 struct rxrpc_net *rxnet = rxrpc_net(sock_net(&rx->sk));
873
874 _enter("%p{%d,%d}", sk, sk->sk_state, refcount_read(&sk->sk_refcnt));
875
876 /* declare the socket closed for business */
877 sock_orphan(sk);
878 sk->sk_shutdown = SHUTDOWN_MASK;
879
880 /* We want to kill off all connections from a service socket
881 * as fast as possible because we can't share these; client
882 * sockets, on the other hand, can share an endpoint.
883 */
884 switch (sk->sk_state) {
885 case RXRPC_SERVER_BOUND:
886 case RXRPC_SERVER_BOUND2:
887 case RXRPC_SERVER_LISTENING:
888 case RXRPC_SERVER_LISTEN_DISABLED:
889 rx->local->service_closed = true;
890 break;
891 }
892
893 spin_lock_bh(&sk->sk_receive_queue.lock);
894 sk->sk_state = RXRPC_CLOSE;
895 spin_unlock_bh(&sk->sk_receive_queue.lock);
896
897 if (rx->local && rcu_access_pointer(rx->local->service) == rx) {
898 write_lock(&rx->local->services_lock);
899 rcu_assign_pointer(rx->local->service, NULL);
900 write_unlock(&rx->local->services_lock);
901 }
902
903 /* try to flush out this socket */
904 rxrpc_discard_prealloc(rx);
905 rxrpc_release_calls_on_socket(rx);
906 flush_workqueue(rxrpc_workqueue);
907 rxrpc_purge_queue(&sk->sk_receive_queue);
908 rxrpc_queue_work(&rxnet->service_conn_reaper);
909 rxrpc_queue_work(&rxnet->client_conn_reaper);
910
911 rxrpc_put_local(rx->local);
912 rx->local = NULL;
913 key_put(rx->key);
914 rx->key = NULL;
915 key_put(rx->securities);
916 rx->securities = NULL;
917 sock_put(sk);
918
919 _leave(" = 0");
920 return 0;
921}
922
923/*
924 * release an RxRPC BSD socket on close() or equivalent
925 */
926static int rxrpc_release(struct socket *sock)
927{
928 struct sock *sk = sock->sk;
929
930 _enter("%p{%p}", sock, sk);
931
932 if (!sk)
933 return 0;
934
935 sock->sk = NULL;
936
937 return rxrpc_release_sock(sk);
938}
939
940/*
941 * RxRPC network protocol
942 */
943static const struct proto_ops rxrpc_rpc_ops = {
944 .family = PF_RXRPC,
945 .owner = THIS_MODULE,
946 .release = rxrpc_release,
947 .bind = rxrpc_bind,
948 .connect = rxrpc_connect,
949 .socketpair = sock_no_socketpair,
950 .accept = sock_no_accept,
951 .getname = sock_no_getname,
952 .poll = rxrpc_poll,
953 .ioctl = sock_no_ioctl,
954 .listen = rxrpc_listen,
955 .shutdown = rxrpc_shutdown,
956 .setsockopt = rxrpc_setsockopt,
957 .getsockopt = rxrpc_getsockopt,
958 .sendmsg = rxrpc_sendmsg,
959 .recvmsg = rxrpc_recvmsg,
960 .mmap = sock_no_mmap,
961 .sendpage = sock_no_sendpage,
962};
963
964static struct proto rxrpc_proto = {
965 .name = "RXRPC",
966 .owner = THIS_MODULE,
967 .obj_size = sizeof(struct rxrpc_sock),
968 .max_header = sizeof(struct rxrpc_wire_header),
969};
970
971static const struct net_proto_family rxrpc_family_ops = {
972 .family = PF_RXRPC,
973 .create = rxrpc_create,
974 .owner = THIS_MODULE,
975};
976
977/*
978 * initialise and register the RxRPC protocol
979 */
980static int __init af_rxrpc_init(void)
981{
982 int ret = -1;
983 unsigned int tmp;
984
985 BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > FIELD_SIZEOF(struct sk_buff, cb));
986
987 get_random_bytes(&tmp, sizeof(tmp));
988 tmp &= 0x3fffffff;
989 if (tmp == 0)
990 tmp = 1;
991 idr_set_cursor(&rxrpc_client_conn_ids, tmp);
992
993 ret = -ENOMEM;
994 rxrpc_call_jar = kmem_cache_create(
995 "rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
996 SLAB_HWCACHE_ALIGN, NULL);
997 if (!rxrpc_call_jar) {
998 pr_notice("Failed to allocate call jar\n");
999 goto error_call_jar;
1000 }
1001
1002 rxrpc_workqueue = alloc_workqueue("krxrpcd", 0, 1);
1003 if (!rxrpc_workqueue) {
1004 pr_notice("Failed to allocate work queue\n");
1005 goto error_work_queue;
1006 }
1007
1008 ret = rxrpc_init_security();
1009 if (ret < 0) {
1010 pr_crit("Cannot initialise security\n");
1011 goto error_security;
1012 }
1013
1014 ret = register_pernet_subsys(&rxrpc_net_ops);
1015 if (ret)
1016 goto error_pernet;
1017
1018 ret = proto_register(&rxrpc_proto, 1);
1019 if (ret < 0) {
1020 pr_crit("Cannot register protocol\n");
1021 goto error_proto;
1022 }
1023
1024 ret = sock_register(&rxrpc_family_ops);
1025 if (ret < 0) {
1026 pr_crit("Cannot register socket family\n");
1027 goto error_sock;
1028 }
1029
1030 ret = register_key_type(&key_type_rxrpc);
1031 if (ret < 0) {
1032 pr_crit("Cannot register client key type\n");
1033 goto error_key_type;
1034 }
1035
1036 ret = register_key_type(&key_type_rxrpc_s);
1037 if (ret < 0) {
1038 pr_crit("Cannot register server key type\n");
1039 goto error_key_type_s;
1040 }
1041
1042 ret = rxrpc_sysctl_init();
1043 if (ret < 0) {
1044 pr_crit("Cannot register sysctls\n");
1045 goto error_sysctls;
1046 }
1047
1048 return 0;
1049
1050error_sysctls:
1051 unregister_key_type(&key_type_rxrpc_s);
1052error_key_type_s:
1053 unregister_key_type(&key_type_rxrpc);
1054error_key_type:
1055 sock_unregister(PF_RXRPC);
1056error_sock:
1057 proto_unregister(&rxrpc_proto);
1058error_proto:
1059 unregister_pernet_subsys(&rxrpc_net_ops);
1060error_pernet:
1061 rxrpc_exit_security();
1062error_security:
1063 destroy_workqueue(rxrpc_workqueue);
1064error_work_queue:
1065 kmem_cache_destroy(rxrpc_call_jar);
1066error_call_jar:
1067 return ret;
1068}
1069
1070/*
1071 * unregister the RxRPC protocol
1072 */
1073static void __exit af_rxrpc_exit(void)
1074{
1075 _enter("");
1076 rxrpc_sysctl_exit();
1077 unregister_key_type(&key_type_rxrpc_s);
1078 unregister_key_type(&key_type_rxrpc);
1079 sock_unregister(PF_RXRPC);
1080 proto_unregister(&rxrpc_proto);
1081 unregister_pernet_subsys(&rxrpc_net_ops);
1082 ASSERTCMP(atomic_read(&rxrpc_n_tx_skbs), ==, 0);
1083 ASSERTCMP(atomic_read(&rxrpc_n_rx_skbs), ==, 0);
1084
1085 /* Make sure the local and peer records pinned by any dying connections
1086 * are released.
1087 */
1088 rcu_barrier();
1089 rxrpc_destroy_client_conn_ids();
1090
1091 destroy_workqueue(rxrpc_workqueue);
1092 rxrpc_exit_security();
1093 kmem_cache_destroy(rxrpc_call_jar);
1094 _leave("");
1095}
1096
1097module_init(af_rxrpc_init);
1098module_exit(af_rxrpc_exit);