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->proto)
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 *rxrpc_name_to_transport(struct rxrpc_sock *rx,
231 struct sockaddr *addr,
232 int addr_len, int flags,
235 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *) addr;
236 struct rxrpc_transport *trans;
237 struct rxrpc_peer *peer;
239 _enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
241 ASSERT(rx->local != NULL);
243 if (rx->srx.transport_type != srx->transport_type)
244 return ERR_PTR(-ESOCKTNOSUPPORT);
245 if (rx->srx.transport.family != srx->transport.family)
246 return ERR_PTR(-EAFNOSUPPORT);
248 /* find a remote transport endpoint from the local one */
249 peer = rxrpc_lookup_peer(rx->local, srx, gfp);
251 return ERR_CAST(peer);
253 /* find a transport */
254 trans = rxrpc_get_transport(rx->local, peer, gfp);
255 rxrpc_put_peer(peer);
256 _leave(" = %p", trans);
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
267 * Allow a kernel service to begin a call on the nominated socket. This just
268 * sets up all the internal tracking structures and allocates connection and
269 * call IDs as appropriate. The call to be used is returned.
271 * The default socket destination address and security may be overridden by
272 * supplying @srx and @key.
274 struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
275 struct sockaddr_rxrpc *srx,
277 unsigned long user_call_ID,
280 struct rxrpc_conn_bundle *bundle;
281 struct rxrpc_transport *trans;
282 struct rxrpc_call *call;
283 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
285 _enter(",,%x,%lx", key_serial(key), user_call_ID);
289 trans = rxrpc_name_to_transport(rx, (struct sockaddr *)srx,
290 sizeof(*srx), 0, gfp);
292 call = ERR_CAST(trans);
299 if (key && !key->payload.data[0])
300 key = NULL; /* a no-security key */
302 bundle = rxrpc_get_bundle(rx, trans, key, srx->srx_service, gfp);
303 if (IS_ERR(bundle)) {
304 call = ERR_CAST(bundle);
308 call = rxrpc_new_client_call(rx, trans, bundle, user_call_ID, gfp);
309 rxrpc_put_bundle(trans, bundle);
311 rxrpc_put_transport(trans);
313 release_sock(&rx->sk);
314 _leave(" = %p", call);
317 EXPORT_SYMBOL(rxrpc_kernel_begin_call);
320 * rxrpc_kernel_end_call - Allow a kernel service to end a call it was using
321 * @call: The call to end
323 * Allow a kernel service to end a call it was using. The call must be
324 * complete before this is called (the call should be aborted if necessary).
326 void rxrpc_kernel_end_call(struct rxrpc_call *call)
328 _enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
329 rxrpc_remove_user_ID(call->socket, call);
330 rxrpc_put_call(call);
332 EXPORT_SYMBOL(rxrpc_kernel_end_call);
335 * rxrpc_kernel_intercept_rx_messages - Intercept received RxRPC messages
336 * @sock: The socket to intercept received messages on
337 * @interceptor: The function to pass the messages to
339 * Allow a kernel service to intercept messages heading for the Rx queue on an
340 * RxRPC socket. They get passed to the specified function instead.
341 * @interceptor should free the socket buffers it is given. @interceptor is
342 * called with the socket receive queue spinlock held and softirqs disabled -
343 * this ensures that the messages will be delivered in the right order.
345 void rxrpc_kernel_intercept_rx_messages(struct socket *sock,
346 rxrpc_interceptor_t interceptor)
348 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
351 rx->interceptor = interceptor;
354 EXPORT_SYMBOL(rxrpc_kernel_intercept_rx_messages);
357 * connect an RxRPC socket
358 * - this just targets it at a specific destination; no actual connection
359 * negotiation takes place
361 static int rxrpc_connect(struct socket *sock, struct sockaddr *addr,
362 int addr_len, int flags)
364 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
365 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
368 _enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
370 ret = rxrpc_validate_address(rx, srx, addr_len);
372 _leave(" = %d [bad addr]", ret);
379 if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
382 switch (rx->sk.sk_state) {
384 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
385 case RXRPC_CLIENT_UNBOUND:
386 case RXRPC_CLIENT_BOUND:
393 rx->connect_srx = *srx;
394 set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
398 release_sock(&rx->sk);
403 * send a message through an RxRPC socket
404 * - in a client this does a number of things:
405 * - finds/sets up a connection for the security specified (if any)
406 * - initiates a call (ID in control data)
407 * - ends the request phase of a call (if MSG_MORE is not set)
408 * - sends a call data packet
409 * - may send an abort (abort code in control data)
411 static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
413 struct rxrpc_local *local;
414 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
417 _enter(",{%d},,%zu", rx->sk.sk_state, len);
419 if (m->msg_flags & MSG_OOB)
423 ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
425 _leave(" = %d [bad addr]", ret);
432 switch (rx->sk.sk_state) {
434 local = rxrpc_lookup_local(&rx->srx);
436 ret = PTR_ERR(local);
441 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
444 case RXRPC_CLIENT_UNBOUND:
445 case RXRPC_CLIENT_BOUND:
447 test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
448 m->msg_name = &rx->connect_srx;
449 m->msg_namelen = sizeof(rx->connect_srx);
451 case RXRPC_SERVER_BOUND:
452 case RXRPC_SERVER_LISTENING:
453 ret = rxrpc_do_sendmsg(rx, m, len);
461 release_sock(&rx->sk);
462 _leave(" = %d", ret);
467 * set RxRPC socket options
469 static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
470 char __user *optval, unsigned int optlen)
472 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
473 unsigned int min_sec_level;
476 _enter(",%d,%d,,%d", level, optname, optlen);
481 if (level == SOL_RXRPC) {
483 case RXRPC_EXCLUSIVE_CONNECTION:
488 if (rx->sk.sk_state != RXRPC_UNBOUND)
490 set_bit(RXRPC_SOCK_EXCLUSIVE_CONN, &rx->flags);
493 case RXRPC_SECURITY_KEY:
498 if (rx->sk.sk_state != RXRPC_UNBOUND)
500 ret = rxrpc_request_key(rx, optval, optlen);
503 case RXRPC_SECURITY_KEYRING:
508 if (rx->sk.sk_state != RXRPC_UNBOUND)
510 ret = rxrpc_server_keyring(rx, optval, optlen);
513 case RXRPC_MIN_SECURITY_LEVEL:
515 if (optlen != sizeof(unsigned int))
518 if (rx->sk.sk_state != RXRPC_UNBOUND)
520 ret = get_user(min_sec_level,
521 (unsigned int __user *) optval);
525 if (min_sec_level > RXRPC_SECURITY_MAX)
527 rx->min_sec_level = min_sec_level;
538 release_sock(&rx->sk);
543 * permit an RxRPC socket to be polled
545 static unsigned int rxrpc_poll(struct file *file, struct socket *sock,
549 struct sock *sk = sock->sk;
551 sock_poll_wait(file, sk_sleep(sk), wait);
554 /* the socket is readable if there are any messages waiting on the Rx
556 if (!skb_queue_empty(&sk->sk_receive_queue))
557 mask |= POLLIN | POLLRDNORM;
559 /* the socket is writable if there is space to add new data to the
560 * socket; there is no guarantee that any particular call in progress
561 * on the socket may have space in the Tx ACK window */
562 if (rxrpc_writable(sk))
563 mask |= POLLOUT | POLLWRNORM;
569 * create an RxRPC socket
571 static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
574 struct rxrpc_sock *rx;
577 _enter("%p,%d", sock, protocol);
579 if (!net_eq(net, &init_net))
580 return -EAFNOSUPPORT;
582 /* we support transport protocol UDP/UDP6 only */
583 if (protocol != PF_INET)
584 return -EPROTONOSUPPORT;
586 if (sock->type != SOCK_DGRAM)
587 return -ESOCKTNOSUPPORT;
589 sock->ops = &rxrpc_rpc_ops;
590 sock->state = SS_UNCONNECTED;
592 sk = sk_alloc(net, PF_RXRPC, GFP_KERNEL, &rxrpc_proto, kern);
596 sock_init_data(sock, sk);
597 sk->sk_state = RXRPC_UNBOUND;
598 sk->sk_write_space = rxrpc_write_space;
599 sk->sk_max_ack_backlog = 0;
600 sk->sk_destruct = rxrpc_sock_destructor;
603 rx->proto = protocol;
606 INIT_LIST_HEAD(&rx->listen_link);
607 INIT_LIST_HEAD(&rx->secureq);
608 INIT_LIST_HEAD(&rx->acceptq);
609 rwlock_init(&rx->call_lock);
610 memset(&rx->srx, 0, sizeof(rx->srx));
612 _leave(" = 0 [%p]", rx);
617 * RxRPC socket destructor
619 static void rxrpc_sock_destructor(struct sock *sk)
623 rxrpc_purge_queue(&sk->sk_receive_queue);
625 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
626 WARN_ON(!sk_unhashed(sk));
627 WARN_ON(sk->sk_socket);
629 if (!sock_flag(sk, SOCK_DEAD)) {
630 printk("Attempt to release alive rxrpc socket: %p\n", sk);
636 * release an RxRPC socket
638 static int rxrpc_release_sock(struct sock *sk)
640 struct rxrpc_sock *rx = rxrpc_sk(sk);
642 _enter("%p{%d,%d}", sk, sk->sk_state, atomic_read(&sk->sk_refcnt));
644 /* declare the socket closed for business */
646 sk->sk_shutdown = SHUTDOWN_MASK;
648 spin_lock_bh(&sk->sk_receive_queue.lock);
649 sk->sk_state = RXRPC_CLOSE;
650 spin_unlock_bh(&sk->sk_receive_queue.lock);
652 ASSERTCMP(rx->listen_link.next, !=, LIST_POISON1);
654 if (!list_empty(&rx->listen_link)) {
655 write_lock_bh(&rx->local->services_lock);
656 list_del(&rx->listen_link);
657 write_unlock_bh(&rx->local->services_lock);
660 /* try to flush out this socket */
661 rxrpc_release_calls_on_socket(rx);
662 flush_workqueue(rxrpc_workqueue);
663 rxrpc_purge_queue(&sk->sk_receive_queue);
666 rxrpc_put_connection(rx->conn);
671 rxrpc_put_local(rx->local);
677 key_put(rx->securities);
678 rx->securities = NULL;
686 * release an RxRPC BSD socket on close() or equivalent
688 static int rxrpc_release(struct socket *sock)
690 struct sock *sk = sock->sk;
692 _enter("%p{%p}", sock, sk);
699 return rxrpc_release_sock(sk);
703 * RxRPC network protocol
705 static const struct proto_ops rxrpc_rpc_ops = {
707 .owner = THIS_MODULE,
708 .release = rxrpc_release,
710 .connect = rxrpc_connect,
711 .socketpair = sock_no_socketpair,
712 .accept = sock_no_accept,
713 .getname = sock_no_getname,
715 .ioctl = sock_no_ioctl,
716 .listen = rxrpc_listen,
717 .shutdown = sock_no_shutdown,
718 .setsockopt = rxrpc_setsockopt,
719 .getsockopt = sock_no_getsockopt,
720 .sendmsg = rxrpc_sendmsg,
721 .recvmsg = rxrpc_recvmsg,
722 .mmap = sock_no_mmap,
723 .sendpage = sock_no_sendpage,
726 static struct proto rxrpc_proto = {
728 .owner = THIS_MODULE,
729 .obj_size = sizeof(struct rxrpc_sock),
730 .max_header = sizeof(struct rxrpc_wire_header),
733 static const struct net_proto_family rxrpc_family_ops = {
735 .create = rxrpc_create,
736 .owner = THIS_MODULE,
740 * initialise and register the RxRPC protocol
742 static int __init af_rxrpc_init(void)
746 BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > FIELD_SIZEOF(struct sk_buff, cb));
748 rxrpc_epoch = get_seconds();
751 rxrpc_call_jar = kmem_cache_create(
752 "rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
753 SLAB_HWCACHE_ALIGN, NULL);
754 if (!rxrpc_call_jar) {
755 pr_notice("Failed to allocate call jar\n");
759 rxrpc_workqueue = alloc_workqueue("krxrpcd", 0, 1);
760 if (!rxrpc_workqueue) {
761 pr_notice("Failed to allocate work queue\n");
762 goto error_work_queue;
765 ret = rxrpc_init_security();
767 pr_crit("Cannot initialise security\n");
771 ret = proto_register(&rxrpc_proto, 1);
773 pr_crit("Cannot register protocol\n");
777 ret = sock_register(&rxrpc_family_ops);
779 pr_crit("Cannot register socket family\n");
783 ret = register_key_type(&key_type_rxrpc);
785 pr_crit("Cannot register client key type\n");
789 ret = register_key_type(&key_type_rxrpc_s);
791 pr_crit("Cannot register server key type\n");
792 goto error_key_type_s;
795 ret = rxrpc_sysctl_init();
797 pr_crit("Cannot register sysctls\n");
801 #ifdef CONFIG_PROC_FS
802 proc_create("rxrpc_calls", 0, init_net.proc_net, &rxrpc_call_seq_fops);
803 proc_create("rxrpc_conns", 0, init_net.proc_net,
804 &rxrpc_connection_seq_fops);
809 unregister_key_type(&key_type_rxrpc_s);
811 unregister_key_type(&key_type_rxrpc);
813 sock_unregister(PF_RXRPC);
815 proto_unregister(&rxrpc_proto);
817 destroy_workqueue(rxrpc_workqueue);
819 rxrpc_exit_security();
821 kmem_cache_destroy(rxrpc_call_jar);
827 * unregister the RxRPC protocol
829 static void __exit af_rxrpc_exit(void)
833 unregister_key_type(&key_type_rxrpc_s);
834 unregister_key_type(&key_type_rxrpc);
835 sock_unregister(PF_RXRPC);
836 proto_unregister(&rxrpc_proto);
837 rxrpc_destroy_all_calls();
838 rxrpc_destroy_all_connections();
839 rxrpc_destroy_all_transports();
841 ASSERTCMP(atomic_read(&rxrpc_n_skbs), ==, 0);
843 /* We need to flush the scheduled work twice because the local endpoint
844 * records involve a work item in their destruction as they can only be
845 * destroyed from process context. However, a connection may have a
846 * work item outstanding - and this will pin the local endpoint record
847 * until the connection goes away.
849 * Peers don't pin locals and calls pin sockets - which prevents the
850 * module from being unloaded - so we should only need two flushes.
852 _debug("flush scheduled work");
853 flush_workqueue(rxrpc_workqueue);
854 _debug("flush scheduled work 2");
855 flush_workqueue(rxrpc_workqueue);
856 _debug("synchronise RCU");
858 _debug("destroy locals");
859 rxrpc_destroy_all_locals();
861 remove_proc_entry("rxrpc_conns", init_net.proc_net);
862 remove_proc_entry("rxrpc_calls", init_net.proc_net);
863 destroy_workqueue(rxrpc_workqueue);
864 rxrpc_exit_security();
865 kmem_cache_destroy(rxrpc_call_jar);
869 module_init(af_rxrpc_init);
870 module_exit(af_rxrpc_exit);