1 /* AF_RXRPC implementation
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
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.
12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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/poll.h>
20 #include <linux/proc_fs.h>
21 #include <linux/key-type.h>
22 #include <net/net_namespace.h>
24 #include <net/af_rxrpc.h>
25 #include "ar-internal.h"
27 MODULE_DESCRIPTION("RxRPC network protocol");
28 MODULE_AUTHOR("Red Hat, Inc.");
29 MODULE_LICENSE("GPL");
30 MODULE_ALIAS_NETPROTO(PF_RXRPC);
32 unsigned int rxrpc_debug; // = RXRPC_DEBUG_KPROTO;
33 module_param_named(debug, rxrpc_debug, uint, S_IWUSR | S_IRUGO);
34 MODULE_PARM_DESC(debug, "RxRPC debugging mask");
36 static struct proto rxrpc_proto;
37 static const struct proto_ops rxrpc_rpc_ops;
39 /* local epoch for detecting local-end reset */
42 /* current debugging ID */
43 atomic_t rxrpc_debug_id;
45 /* count of skbs currently in use */
46 atomic_t rxrpc_n_skbs;
48 struct workqueue_struct *rxrpc_workqueue;
50 static void rxrpc_sock_destructor(struct sock *);
53 * see if an RxRPC socket is currently writable
55 static inline int rxrpc_writable(struct sock *sk)
57 return atomic_read(&sk->sk_wmem_alloc) < (size_t) sk->sk_sndbuf;
61 * wait for write bufferage to become available
63 static void rxrpc_write_space(struct sock *sk)
67 if (rxrpc_writable(sk)) {
68 struct socket_wq *wq = rcu_dereference(sk->sk_wq);
70 if (skwq_has_sleeper(wq))
71 wake_up_interruptible(&wq->wait);
72 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
78 * validate an RxRPC address
80 static int rxrpc_validate_address(struct rxrpc_sock *rx,
81 struct sockaddr_rxrpc *srx,
86 if (len < sizeof(struct sockaddr_rxrpc))
89 if (srx->srx_family != AF_RXRPC)
92 if (srx->transport_type != SOCK_DGRAM)
93 return -ESOCKTNOSUPPORT;
95 len -= offsetof(struct sockaddr_rxrpc, transport);
96 if (srx->transport_len < sizeof(sa_family_t) ||
97 srx->transport_len > len)
100 if (srx->transport.family != rx->family)
101 return -EAFNOSUPPORT;
103 switch (srx->transport.family) {
105 if (srx->transport_len < sizeof(struct sockaddr_in))
107 _debug("INET: %x @ %pI4",
108 ntohs(srx->transport.sin.sin_port),
109 &srx->transport.sin.sin_addr);
110 tail = offsetof(struct sockaddr_rxrpc, transport.sin.__pad);
115 return -EAFNOSUPPORT;
119 memset((void *)srx + tail, 0, len - tail);
124 * bind a local address to an RxRPC socket
126 static int rxrpc_bind(struct socket *sock, struct sockaddr *saddr, int len)
128 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)saddr;
129 struct sock *sk = sock->sk;
130 struct rxrpc_local *local;
131 struct rxrpc_sock *rx = rxrpc_sk(sk), *prx;
134 _enter("%p,%p,%d", rx, saddr, len);
136 ret = rxrpc_validate_address(rx, srx, len);
142 if (rx->sk.sk_state != RXRPC_UNBOUND) {
147 memcpy(&rx->srx, srx, sizeof(rx->srx));
149 local = rxrpc_lookup_local(&rx->srx);
151 ret = PTR_ERR(local);
155 if (rx->srx.srx_service) {
156 write_lock_bh(&local->services_lock);
157 list_for_each_entry(prx, &local->services, listen_link) {
158 if (prx->srx.srx_service == rx->srx.srx_service)
163 list_add_tail(&rx->listen_link, &local->services);
164 write_unlock_bh(&local->services_lock);
166 rx->sk.sk_state = RXRPC_SERVER_BOUND;
169 rx->sk.sk_state = RXRPC_CLIENT_BOUND;
172 release_sock(&rx->sk);
177 write_unlock_bh(&local->services_lock);
178 rxrpc_put_local(local);
181 release_sock(&rx->sk);
183 _leave(" = %d", ret);
188 * set the number of pending calls permitted on a listening socket
190 static int rxrpc_listen(struct socket *sock, int backlog)
192 struct sock *sk = sock->sk;
193 struct rxrpc_sock *rx = rxrpc_sk(sk);
197 _enter("%p,%d", rx, backlog);
201 switch (rx->sk.sk_state) {
203 ret = -EADDRNOTAVAIL;
205 case RXRPC_SERVER_BOUND:
206 ASSERT(rx->local != NULL);
207 max = READ_ONCE(rxrpc_max_backlog);
209 if (backlog == INT_MAX)
211 else if (backlog < 0 || backlog > max)
213 sk->sk_max_ack_backlog = backlog;
214 rx->sk.sk_state = RXRPC_SERVER_LISTENING;
222 release_sock(&rx->sk);
223 _leave(" = %d", ret);
228 * find a transport by address
230 struct rxrpc_transport *
231 rxrpc_name_to_transport(struct rxrpc_conn_parameters *cp,
232 struct sockaddr *addr,
236 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *) addr;
237 struct rxrpc_transport *trans;
239 _enter("%p,%d", addr, addr_len);
241 if (cp->local->srx.transport_type != srx->transport_type)
242 return ERR_PTR(-ESOCKTNOSUPPORT);
243 if (cp->local->srx.transport.family != srx->transport.family)
244 return ERR_PTR(-EAFNOSUPPORT);
246 /* find a remote transport endpoint from the local one */
247 cp->peer = rxrpc_lookup_peer(cp->local, srx, gfp);
249 return ERR_PTR(-ENOMEM);
251 /* find a transport */
252 trans = rxrpc_get_transport(cp->local, cp->peer, gfp);
253 rxrpc_put_peer(cp->peer);
254 _leave(" = %p", trans);
259 * rxrpc_kernel_begin_call - Allow a kernel service to begin a call
260 * @sock: The socket on which to make the call
261 * @srx: The address of the peer to contact
262 * @key: The security context to use (defaults to socket setting)
263 * @user_call_ID: The ID to use
265 * Allow a kernel service to begin a call on the nominated socket. This just
266 * sets up all the internal tracking structures and allocates connection and
267 * call IDs as appropriate. The call to be used is returned.
269 * The default socket destination address and security may be overridden by
270 * supplying @srx and @key.
272 struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
273 struct sockaddr_rxrpc *srx,
275 unsigned long user_call_ID,
278 struct rxrpc_conn_parameters cp;
279 struct rxrpc_conn_bundle *bundle;
280 struct rxrpc_transport *trans;
281 struct rxrpc_call *call;
282 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
285 _enter(",,%x,%lx", key_serial(key), user_call_ID);
287 ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
295 if (key && !key->payload.data[0])
296 key = NULL; /* a no-security key */
298 memset(&cp, 0, sizeof(cp));
299 cp.local = rx->local;
301 cp.security_level = 0;
302 cp.exclusive = false;
303 cp.service_id = srx->srx_service;
305 trans = rxrpc_name_to_transport(&cp, (struct sockaddr *)srx,
308 call = ERR_CAST(trans);
312 cp.peer = trans->peer;
314 bundle = rxrpc_get_bundle(rx, trans, key, srx->srx_service, gfp);
315 if (IS_ERR(bundle)) {
316 call = ERR_CAST(bundle);
320 call = rxrpc_new_client_call(rx, &cp, trans, bundle, user_call_ID, gfp);
321 rxrpc_put_bundle(trans, bundle);
323 rxrpc_put_transport(trans);
325 release_sock(&rx->sk);
326 _leave(" = %p", call);
329 EXPORT_SYMBOL(rxrpc_kernel_begin_call);
332 * rxrpc_kernel_end_call - Allow a kernel service to end a call it was using
333 * @call: The call to end
335 * Allow a kernel service to end a call it was using. The call must be
336 * complete before this is called (the call should be aborted if necessary).
338 void rxrpc_kernel_end_call(struct rxrpc_call *call)
340 _enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
341 rxrpc_remove_user_ID(call->socket, call);
342 rxrpc_put_call(call);
344 EXPORT_SYMBOL(rxrpc_kernel_end_call);
347 * rxrpc_kernel_intercept_rx_messages - Intercept received RxRPC messages
348 * @sock: The socket to intercept received messages on
349 * @interceptor: The function to pass the messages to
351 * Allow a kernel service to intercept messages heading for the Rx queue on an
352 * RxRPC socket. They get passed to the specified function instead.
353 * @interceptor should free the socket buffers it is given. @interceptor is
354 * called with the socket receive queue spinlock held and softirqs disabled -
355 * this ensures that the messages will be delivered in the right order.
357 void rxrpc_kernel_intercept_rx_messages(struct socket *sock,
358 rxrpc_interceptor_t interceptor)
360 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
363 rx->interceptor = interceptor;
366 EXPORT_SYMBOL(rxrpc_kernel_intercept_rx_messages);
369 * connect an RxRPC socket
370 * - this just targets it at a specific destination; no actual connection
371 * negotiation takes place
373 static int rxrpc_connect(struct socket *sock, struct sockaddr *addr,
374 int addr_len, int flags)
376 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
377 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
380 _enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
382 ret = rxrpc_validate_address(rx, srx, addr_len);
384 _leave(" = %d [bad addr]", ret);
391 if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
394 switch (rx->sk.sk_state) {
396 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
397 case RXRPC_CLIENT_UNBOUND:
398 case RXRPC_CLIENT_BOUND:
405 rx->connect_srx = *srx;
406 set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
410 release_sock(&rx->sk);
415 * send a message through an RxRPC socket
416 * - in a client this does a number of things:
417 * - finds/sets up a connection for the security specified (if any)
418 * - initiates a call (ID in control data)
419 * - ends the request phase of a call (if MSG_MORE is not set)
420 * - sends a call data packet
421 * - may send an abort (abort code in control data)
423 static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
425 struct rxrpc_local *local;
426 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
429 _enter(",{%d},,%zu", rx->sk.sk_state, len);
431 if (m->msg_flags & MSG_OOB)
435 ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
437 _leave(" = %d [bad addr]", ret);
444 switch (rx->sk.sk_state) {
446 local = rxrpc_lookup_local(&rx->srx);
448 ret = PTR_ERR(local);
453 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
456 case RXRPC_CLIENT_UNBOUND:
457 case RXRPC_CLIENT_BOUND:
459 test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
460 m->msg_name = &rx->connect_srx;
461 m->msg_namelen = sizeof(rx->connect_srx);
463 case RXRPC_SERVER_BOUND:
464 case RXRPC_SERVER_LISTENING:
465 ret = rxrpc_do_sendmsg(rx, m, len);
473 release_sock(&rx->sk);
474 _leave(" = %d", ret);
479 * set RxRPC socket options
481 static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
482 char __user *optval, unsigned int optlen)
484 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
485 unsigned int min_sec_level;
488 _enter(",%d,%d,,%d", level, optname, optlen);
493 if (level == SOL_RXRPC) {
495 case RXRPC_EXCLUSIVE_CONNECTION:
500 if (rx->sk.sk_state != RXRPC_UNBOUND)
502 rx->exclusive = true;
505 case RXRPC_SECURITY_KEY:
510 if (rx->sk.sk_state != RXRPC_UNBOUND)
512 ret = rxrpc_request_key(rx, optval, optlen);
515 case RXRPC_SECURITY_KEYRING:
520 if (rx->sk.sk_state != RXRPC_UNBOUND)
522 ret = rxrpc_server_keyring(rx, optval, optlen);
525 case RXRPC_MIN_SECURITY_LEVEL:
527 if (optlen != sizeof(unsigned int))
530 if (rx->sk.sk_state != RXRPC_UNBOUND)
532 ret = get_user(min_sec_level,
533 (unsigned int __user *) optval);
537 if (min_sec_level > RXRPC_SECURITY_MAX)
539 rx->min_sec_level = min_sec_level;
550 release_sock(&rx->sk);
555 * permit an RxRPC socket to be polled
557 static unsigned int rxrpc_poll(struct file *file, struct socket *sock,
561 struct sock *sk = sock->sk;
563 sock_poll_wait(file, sk_sleep(sk), wait);
566 /* the socket is readable if there are any messages waiting on the Rx
568 if (!skb_queue_empty(&sk->sk_receive_queue))
569 mask |= POLLIN | POLLRDNORM;
571 /* the socket is writable if there is space to add new data to the
572 * socket; there is no guarantee that any particular call in progress
573 * on the socket may have space in the Tx ACK window */
574 if (rxrpc_writable(sk))
575 mask |= POLLOUT | POLLWRNORM;
581 * create an RxRPC socket
583 static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
586 struct rxrpc_sock *rx;
589 _enter("%p,%d", sock, protocol);
591 if (!net_eq(net, &init_net))
592 return -EAFNOSUPPORT;
594 /* we support transport protocol UDP/UDP6 only */
595 if (protocol != PF_INET)
596 return -EPROTONOSUPPORT;
598 if (sock->type != SOCK_DGRAM)
599 return -ESOCKTNOSUPPORT;
601 sock->ops = &rxrpc_rpc_ops;
602 sock->state = SS_UNCONNECTED;
604 sk = sk_alloc(net, PF_RXRPC, GFP_KERNEL, &rxrpc_proto, kern);
608 sock_init_data(sock, sk);
609 sk->sk_state = RXRPC_UNBOUND;
610 sk->sk_write_space = rxrpc_write_space;
611 sk->sk_max_ack_backlog = 0;
612 sk->sk_destruct = rxrpc_sock_destructor;
615 rx->family = protocol;
618 INIT_LIST_HEAD(&rx->listen_link);
619 INIT_LIST_HEAD(&rx->secureq);
620 INIT_LIST_HEAD(&rx->acceptq);
621 rwlock_init(&rx->call_lock);
622 memset(&rx->srx, 0, sizeof(rx->srx));
624 _leave(" = 0 [%p]", rx);
629 * RxRPC socket destructor
631 static void rxrpc_sock_destructor(struct sock *sk)
635 rxrpc_purge_queue(&sk->sk_receive_queue);
637 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
638 WARN_ON(!sk_unhashed(sk));
639 WARN_ON(sk->sk_socket);
641 if (!sock_flag(sk, SOCK_DEAD)) {
642 printk("Attempt to release alive rxrpc socket: %p\n", sk);
648 * release an RxRPC socket
650 static int rxrpc_release_sock(struct sock *sk)
652 struct rxrpc_sock *rx = rxrpc_sk(sk);
654 _enter("%p{%d,%d}", sk, sk->sk_state, atomic_read(&sk->sk_refcnt));
656 /* declare the socket closed for business */
658 sk->sk_shutdown = SHUTDOWN_MASK;
660 spin_lock_bh(&sk->sk_receive_queue.lock);
661 sk->sk_state = RXRPC_CLOSE;
662 spin_unlock_bh(&sk->sk_receive_queue.lock);
664 ASSERTCMP(rx->listen_link.next, !=, LIST_POISON1);
666 if (!list_empty(&rx->listen_link)) {
667 write_lock_bh(&rx->local->services_lock);
668 list_del(&rx->listen_link);
669 write_unlock_bh(&rx->local->services_lock);
672 /* try to flush out this socket */
673 rxrpc_release_calls_on_socket(rx);
674 flush_workqueue(rxrpc_workqueue);
675 rxrpc_purge_queue(&sk->sk_receive_queue);
678 rxrpc_put_local(rx->local);
684 key_put(rx->securities);
685 rx->securities = NULL;
693 * release an RxRPC BSD socket on close() or equivalent
695 static int rxrpc_release(struct socket *sock)
697 struct sock *sk = sock->sk;
699 _enter("%p{%p}", sock, sk);
706 return rxrpc_release_sock(sk);
710 * RxRPC network protocol
712 static const struct proto_ops rxrpc_rpc_ops = {
714 .owner = THIS_MODULE,
715 .release = rxrpc_release,
717 .connect = rxrpc_connect,
718 .socketpair = sock_no_socketpair,
719 .accept = sock_no_accept,
720 .getname = sock_no_getname,
722 .ioctl = sock_no_ioctl,
723 .listen = rxrpc_listen,
724 .shutdown = sock_no_shutdown,
725 .setsockopt = rxrpc_setsockopt,
726 .getsockopt = sock_no_getsockopt,
727 .sendmsg = rxrpc_sendmsg,
728 .recvmsg = rxrpc_recvmsg,
729 .mmap = sock_no_mmap,
730 .sendpage = sock_no_sendpage,
733 static struct proto rxrpc_proto = {
735 .owner = THIS_MODULE,
736 .obj_size = sizeof(struct rxrpc_sock),
737 .max_header = sizeof(struct rxrpc_wire_header),
740 static const struct net_proto_family rxrpc_family_ops = {
742 .create = rxrpc_create,
743 .owner = THIS_MODULE,
747 * initialise and register the RxRPC protocol
749 static int __init af_rxrpc_init(void)
753 BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > FIELD_SIZEOF(struct sk_buff, cb));
755 rxrpc_epoch = get_seconds();
758 rxrpc_call_jar = kmem_cache_create(
759 "rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
760 SLAB_HWCACHE_ALIGN, NULL);
761 if (!rxrpc_call_jar) {
762 pr_notice("Failed to allocate call jar\n");
766 rxrpc_workqueue = alloc_workqueue("krxrpcd", 0, 1);
767 if (!rxrpc_workqueue) {
768 pr_notice("Failed to allocate work queue\n");
769 goto error_work_queue;
772 ret = rxrpc_init_security();
774 pr_crit("Cannot initialise security\n");
778 ret = proto_register(&rxrpc_proto, 1);
780 pr_crit("Cannot register protocol\n");
784 ret = sock_register(&rxrpc_family_ops);
786 pr_crit("Cannot register socket family\n");
790 ret = register_key_type(&key_type_rxrpc);
792 pr_crit("Cannot register client key type\n");
796 ret = register_key_type(&key_type_rxrpc_s);
798 pr_crit("Cannot register server key type\n");
799 goto error_key_type_s;
802 ret = rxrpc_sysctl_init();
804 pr_crit("Cannot register sysctls\n");
808 #ifdef CONFIG_PROC_FS
809 proc_create("rxrpc_calls", 0, init_net.proc_net, &rxrpc_call_seq_fops);
810 proc_create("rxrpc_conns", 0, init_net.proc_net,
811 &rxrpc_connection_seq_fops);
816 unregister_key_type(&key_type_rxrpc_s);
818 unregister_key_type(&key_type_rxrpc);
820 sock_unregister(PF_RXRPC);
822 proto_unregister(&rxrpc_proto);
824 destroy_workqueue(rxrpc_workqueue);
826 rxrpc_exit_security();
828 kmem_cache_destroy(rxrpc_call_jar);
834 * unregister the RxRPC protocol
836 static void __exit af_rxrpc_exit(void)
840 unregister_key_type(&key_type_rxrpc_s);
841 unregister_key_type(&key_type_rxrpc);
842 sock_unregister(PF_RXRPC);
843 proto_unregister(&rxrpc_proto);
844 rxrpc_destroy_all_calls();
845 rxrpc_destroy_all_connections();
846 rxrpc_destroy_all_transports();
848 ASSERTCMP(atomic_read(&rxrpc_n_skbs), ==, 0);
850 /* We need to flush the scheduled work twice because the local endpoint
851 * records involve a work item in their destruction as they can only be
852 * destroyed from process context. However, a connection may have a
853 * work item outstanding - and this will pin the local endpoint record
854 * until the connection goes away.
856 * Peers don't pin locals and calls pin sockets - which prevents the
857 * module from being unloaded - so we should only need two flushes.
859 _debug("flush scheduled work");
860 flush_workqueue(rxrpc_workqueue);
861 _debug("flush scheduled work 2");
862 flush_workqueue(rxrpc_workqueue);
863 _debug("synchronise RCU");
865 _debug("destroy locals");
866 ASSERT(idr_is_empty(&rxrpc_client_conn_ids));
867 idr_destroy(&rxrpc_client_conn_ids);
868 rxrpc_destroy_all_locals();
870 remove_proc_entry("rxrpc_conns", init_net.proc_net);
871 remove_proc_entry("rxrpc_calls", init_net.proc_net);
872 destroy_workqueue(rxrpc_workqueue);
873 rxrpc_exit_security();
874 kmem_cache_destroy(rxrpc_call_jar);
878 module_init(af_rxrpc_init);
879 module_exit(af_rxrpc_exit);