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/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>
25 #include <net/af_rxrpc.h>
26 #define CREATE_TRACE_POINTS
27 #include "ar-internal.h"
29 MODULE_DESCRIPTION("RxRPC network protocol");
30 MODULE_AUTHOR("Red Hat, Inc.");
31 MODULE_LICENSE("GPL");
32 MODULE_ALIAS_NETPROTO(PF_RXRPC);
34 unsigned int rxrpc_debug; // = RXRPC_DEBUG_KPROTO;
35 module_param_named(debug, rxrpc_debug, uint, S_IWUSR | S_IRUGO);
36 MODULE_PARM_DESC(debug, "RxRPC debugging mask");
38 static struct proto rxrpc_proto;
39 static const struct proto_ops rxrpc_rpc_ops;
41 /* current debugging ID */
42 atomic_t rxrpc_debug_id;
44 /* count of skbs currently in use */
45 atomic_t rxrpc_n_tx_skbs, rxrpc_n_rx_skbs;
47 struct workqueue_struct *rxrpc_workqueue;
49 static void rxrpc_sock_destructor(struct sock *);
52 * see if an RxRPC socket is currently writable
54 static inline int rxrpc_writable(struct sock *sk)
56 return atomic_read(&sk->sk_wmem_alloc) < (size_t) sk->sk_sndbuf;
60 * wait for write bufferage to become available
62 static void rxrpc_write_space(struct sock *sk)
66 if (rxrpc_writable(sk)) {
67 struct socket_wq *wq = rcu_dereference(sk->sk_wq);
69 if (skwq_has_sleeper(wq))
70 wake_up_interruptible(&wq->wait);
71 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
77 * validate an RxRPC address
79 static int rxrpc_validate_address(struct rxrpc_sock *rx,
80 struct sockaddr_rxrpc *srx,
85 if (len < sizeof(struct sockaddr_rxrpc))
88 if (srx->srx_family != AF_RXRPC)
91 if (srx->transport_type != SOCK_DGRAM)
92 return -ESOCKTNOSUPPORT;
94 len -= offsetof(struct sockaddr_rxrpc, transport);
95 if (srx->transport_len < sizeof(sa_family_t) ||
96 srx->transport_len > len)
99 if (srx->transport.family != rx->family)
100 return -EAFNOSUPPORT;
102 switch (srx->transport.family) {
104 if (srx->transport_len < sizeof(struct sockaddr_in))
106 tail = offsetof(struct sockaddr_rxrpc, transport.sin.__pad);
109 #ifdef CONFIG_AF_RXRPC_IPV6
111 if (srx->transport_len < sizeof(struct sockaddr_in6))
113 tail = offsetof(struct sockaddr_rxrpc, transport) +
114 sizeof(struct sockaddr_in6);
119 return -EAFNOSUPPORT;
123 memset((void *)srx + tail, 0, len - tail);
124 _debug("INET: %pISp", &srx->transport);
129 * bind a local address to an RxRPC socket
131 static int rxrpc_bind(struct socket *sock, struct sockaddr *saddr, int len)
133 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)saddr;
134 struct rxrpc_local *local;
135 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
136 u16 service_id = srx->srx_service;
139 _enter("%p,%p,%d", rx, saddr, len);
141 ret = rxrpc_validate_address(rx, srx, len);
147 if (rx->sk.sk_state != RXRPC_UNBOUND) {
152 memcpy(&rx->srx, srx, sizeof(rx->srx));
154 local = rxrpc_lookup_local(sock_net(&rx->sk), &rx->srx);
156 ret = PTR_ERR(local);
161 write_lock(&local->services_lock);
162 if (rcu_access_pointer(local->service))
165 rcu_assign_pointer(local->service, rx);
166 write_unlock(&local->services_lock);
168 rx->sk.sk_state = RXRPC_SERVER_BOUND;
171 rx->sk.sk_state = RXRPC_CLIENT_BOUND;
174 release_sock(&rx->sk);
179 write_unlock(&local->services_lock);
180 rxrpc_put_local(local);
183 release_sock(&rx->sk);
185 _leave(" = %d", ret);
190 * set the number of pending calls permitted on a listening socket
192 static int rxrpc_listen(struct socket *sock, int backlog)
194 struct sock *sk = sock->sk;
195 struct rxrpc_sock *rx = rxrpc_sk(sk);
196 unsigned int max, old;
199 _enter("%p,%d", rx, backlog);
203 switch (rx->sk.sk_state) {
205 ret = -EADDRNOTAVAIL;
207 case RXRPC_SERVER_BOUND:
208 ASSERT(rx->local != NULL);
209 max = READ_ONCE(rxrpc_max_backlog);
211 if (backlog == INT_MAX)
213 else if (backlog < 0 || backlog > max)
215 old = sk->sk_max_ack_backlog;
216 sk->sk_max_ack_backlog = backlog;
217 ret = rxrpc_service_prealloc(rx, GFP_KERNEL);
219 rx->sk.sk_state = RXRPC_SERVER_LISTENING;
221 sk->sk_max_ack_backlog = old;
223 case RXRPC_SERVER_LISTENING:
225 rx->sk.sk_state = RXRPC_SERVER_LISTEN_DISABLED;
226 sk->sk_max_ack_backlog = 0;
227 rxrpc_discard_prealloc(rx);
236 release_sock(&rx->sk);
237 _leave(" = %d", ret);
242 * rxrpc_kernel_begin_call - Allow a kernel service to begin a call
243 * @sock: The socket on which to make the call
244 * @srx: The address of the peer to contact
245 * @key: The security context to use (defaults to socket setting)
246 * @user_call_ID: The ID to use
247 * @gfp: The allocation constraints
248 * @notify_rx: Where to send notifications instead of socket queue
250 * Allow a kernel service to begin a call on the nominated socket. This just
251 * sets up all the internal tracking structures and allocates connection and
252 * call IDs as appropriate. The call to be used is returned.
254 * The default socket destination address and security may be overridden by
255 * supplying @srx and @key.
257 struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
258 struct sockaddr_rxrpc *srx,
260 unsigned long user_call_ID,
262 rxrpc_notify_rx_t notify_rx)
264 struct rxrpc_conn_parameters cp;
265 struct rxrpc_call *call;
266 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
269 _enter(",,%x,%lx", key_serial(key), user_call_ID);
271 ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
279 if (key && !key->payload.data[0])
280 key = NULL; /* a no-security key */
282 memset(&cp, 0, sizeof(cp));
283 cp.local = rx->local;
285 cp.security_level = 0;
286 cp.exclusive = false;
287 cp.service_id = srx->srx_service;
288 call = rxrpc_new_client_call(rx, &cp, srx, user_call_ID, gfp);
289 /* The socket has been unlocked. */
291 call->notify_rx = notify_rx;
293 mutex_unlock(&call->user_mutex);
294 _leave(" = %p", call);
297 EXPORT_SYMBOL(rxrpc_kernel_begin_call);
300 * rxrpc_kernel_end_call - Allow a kernel service to end a call it was using
301 * @sock: The socket the call is on
302 * @call: The call to end
304 * Allow a kernel service to end a call it was using. The call must be
305 * complete before this is called (the call should be aborted if necessary).
307 void rxrpc_kernel_end_call(struct socket *sock, struct rxrpc_call *call)
309 _enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
311 mutex_lock(&call->user_mutex);
312 rxrpc_release_call(rxrpc_sk(sock->sk), call);
313 mutex_unlock(&call->user_mutex);
314 rxrpc_put_call(call, rxrpc_call_put_kernel);
316 EXPORT_SYMBOL(rxrpc_kernel_end_call);
319 * rxrpc_kernel_new_call_notification - Get notifications of new calls
320 * @sock: The socket to intercept received messages on
321 * @notify_new_call: Function to be called when new calls appear
322 * @discard_new_call: Function to discard preallocated calls
324 * Allow a kernel service to be given notifications about new calls.
326 void rxrpc_kernel_new_call_notification(
328 rxrpc_notify_new_call_t notify_new_call,
329 rxrpc_discard_new_call_t discard_new_call)
331 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
333 rx->notify_new_call = notify_new_call;
334 rx->discard_new_call = discard_new_call;
336 EXPORT_SYMBOL(rxrpc_kernel_new_call_notification);
339 * connect an RxRPC socket
340 * - this just targets it at a specific destination; no actual connection
341 * negotiation takes place
343 static int rxrpc_connect(struct socket *sock, struct sockaddr *addr,
344 int addr_len, int flags)
346 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
347 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
350 _enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
352 ret = rxrpc_validate_address(rx, srx, addr_len);
354 _leave(" = %d [bad addr]", ret);
361 if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
364 switch (rx->sk.sk_state) {
366 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
367 case RXRPC_CLIENT_UNBOUND:
368 case RXRPC_CLIENT_BOUND:
375 rx->connect_srx = *srx;
376 set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
380 release_sock(&rx->sk);
385 * send a message through an RxRPC socket
386 * - in a client this does a number of things:
387 * - finds/sets up a connection for the security specified (if any)
388 * - initiates a call (ID in control data)
389 * - ends the request phase of a call (if MSG_MORE is not set)
390 * - sends a call data packet
391 * - may send an abort (abort code in control data)
393 static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
395 struct rxrpc_local *local;
396 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
399 _enter(",{%d},,%zu", rx->sk.sk_state, len);
401 if (m->msg_flags & MSG_OOB)
405 ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
407 _leave(" = %d [bad addr]", ret);
414 switch (rx->sk.sk_state) {
416 rx->srx.srx_family = AF_RXRPC;
417 rx->srx.srx_service = 0;
418 rx->srx.transport_type = SOCK_DGRAM;
419 rx->srx.transport.family = rx->family;
420 switch (rx->family) {
422 rx->srx.transport_len = sizeof(struct sockaddr_in);
424 #ifdef CONFIG_AF_RXRPC_IPV6
426 rx->srx.transport_len = sizeof(struct sockaddr_in6);
433 local = rxrpc_lookup_local(sock_net(sock->sk), &rx->srx);
435 ret = PTR_ERR(local);
440 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
443 case RXRPC_CLIENT_UNBOUND:
444 case RXRPC_CLIENT_BOUND:
446 test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
447 m->msg_name = &rx->connect_srx;
448 m->msg_namelen = sizeof(rx->connect_srx);
450 case RXRPC_SERVER_BOUND:
451 case RXRPC_SERVER_LISTENING:
452 ret = rxrpc_do_sendmsg(rx, m, len);
453 /* The socket has been unlocked */
461 release_sock(&rx->sk);
463 _leave(" = %d", ret);
468 * set RxRPC socket options
470 static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
471 char __user *optval, unsigned int optlen)
473 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
474 unsigned int min_sec_level;
477 _enter(",%d,%d,,%d", level, optname, optlen);
482 if (level == SOL_RXRPC) {
484 case RXRPC_EXCLUSIVE_CONNECTION:
489 if (rx->sk.sk_state != RXRPC_UNBOUND)
491 rx->exclusive = true;
494 case RXRPC_SECURITY_KEY:
499 if (rx->sk.sk_state != RXRPC_UNBOUND)
501 ret = rxrpc_request_key(rx, optval, optlen);
504 case RXRPC_SECURITY_KEYRING:
509 if (rx->sk.sk_state != RXRPC_UNBOUND)
511 ret = rxrpc_server_keyring(rx, optval, optlen);
514 case RXRPC_MIN_SECURITY_LEVEL:
516 if (optlen != sizeof(unsigned int))
519 if (rx->sk.sk_state != RXRPC_UNBOUND)
521 ret = get_user(min_sec_level,
522 (unsigned int __user *) optval);
526 if (min_sec_level > RXRPC_SECURITY_MAX)
528 rx->min_sec_level = min_sec_level;
539 release_sock(&rx->sk);
544 * permit an RxRPC socket to be polled
546 static unsigned int rxrpc_poll(struct file *file, struct socket *sock,
549 struct sock *sk = sock->sk;
550 struct rxrpc_sock *rx = rxrpc_sk(sk);
553 sock_poll_wait(file, sk_sleep(sk), wait);
556 /* the socket is readable if there are any messages waiting on the Rx
558 if (!list_empty(&rx->recvmsg_q))
559 mask |= POLLIN | POLLRDNORM;
561 /* the socket is writable if there is space to add new data to the
562 * socket; there is no guarantee that any particular call in progress
563 * on the socket may have space in the Tx ACK window */
564 if (rxrpc_writable(sk))
565 mask |= POLLOUT | POLLWRNORM;
571 * create an RxRPC socket
573 static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
576 struct rxrpc_sock *rx;
579 _enter("%p,%d", sock, protocol);
581 /* we support transport protocol UDP/UDP6 only */
582 if (protocol != PF_INET &&
583 IS_ENABLED(CONFIG_AF_RXRPC_IPV6) && protocol != PF_INET6)
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 sock_set_flag(sk, SOCK_RCU_FREE);
598 sk->sk_state = RXRPC_UNBOUND;
599 sk->sk_write_space = rxrpc_write_space;
600 sk->sk_max_ack_backlog = 0;
601 sk->sk_destruct = rxrpc_sock_destructor;
604 rx->family = protocol;
607 spin_lock_init(&rx->incoming_lock);
608 INIT_LIST_HEAD(&rx->sock_calls);
609 INIT_LIST_HEAD(&rx->to_be_accepted);
610 INIT_LIST_HEAD(&rx->recvmsg_q);
611 rwlock_init(&rx->recvmsg_lock);
612 rwlock_init(&rx->call_lock);
613 memset(&rx->srx, 0, sizeof(rx->srx));
615 _leave(" = 0 [%p]", rx);
620 * Kill all the calls on a socket and shut it down.
622 static int rxrpc_shutdown(struct socket *sock, int flags)
624 struct sock *sk = sock->sk;
625 struct rxrpc_sock *rx = rxrpc_sk(sk);
628 _enter("%p,%d", sk, flags);
630 if (flags != SHUT_RDWR)
632 if (sk->sk_state == RXRPC_CLOSE)
637 spin_lock_bh(&sk->sk_receive_queue.lock);
638 if (sk->sk_state < RXRPC_CLOSE) {
639 sk->sk_state = RXRPC_CLOSE;
640 sk->sk_shutdown = SHUTDOWN_MASK;
644 spin_unlock_bh(&sk->sk_receive_queue.lock);
646 rxrpc_discard_prealloc(rx);
653 * RxRPC socket destructor
655 static void rxrpc_sock_destructor(struct sock *sk)
659 rxrpc_purge_queue(&sk->sk_receive_queue);
661 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
662 WARN_ON(!sk_unhashed(sk));
663 WARN_ON(sk->sk_socket);
665 if (!sock_flag(sk, SOCK_DEAD)) {
666 printk("Attempt to release alive rxrpc socket: %p\n", sk);
672 * release an RxRPC socket
674 static int rxrpc_release_sock(struct sock *sk)
676 struct rxrpc_sock *rx = rxrpc_sk(sk);
678 _enter("%p{%d,%d}", sk, sk->sk_state, atomic_read(&sk->sk_refcnt));
680 /* declare the socket closed for business */
682 sk->sk_shutdown = SHUTDOWN_MASK;
684 spin_lock_bh(&sk->sk_receive_queue.lock);
685 sk->sk_state = RXRPC_CLOSE;
686 spin_unlock_bh(&sk->sk_receive_queue.lock);
688 if (rx->local && rcu_access_pointer(rx->local->service) == rx) {
689 write_lock(&rx->local->services_lock);
690 rcu_assign_pointer(rx->local->service, NULL);
691 write_unlock(&rx->local->services_lock);
694 /* try to flush out this socket */
695 rxrpc_discard_prealloc(rx);
696 rxrpc_release_calls_on_socket(rx);
697 flush_workqueue(rxrpc_workqueue);
698 rxrpc_purge_queue(&sk->sk_receive_queue);
700 rxrpc_put_local(rx->local);
704 key_put(rx->securities);
705 rx->securities = NULL;
713 * release an RxRPC BSD socket on close() or equivalent
715 static int rxrpc_release(struct socket *sock)
717 struct sock *sk = sock->sk;
719 _enter("%p{%p}", sock, sk);
726 return rxrpc_release_sock(sk);
730 * RxRPC network protocol
732 static const struct proto_ops rxrpc_rpc_ops = {
734 .owner = THIS_MODULE,
735 .release = rxrpc_release,
737 .connect = rxrpc_connect,
738 .socketpair = sock_no_socketpair,
739 .accept = sock_no_accept,
740 .getname = sock_no_getname,
742 .ioctl = sock_no_ioctl,
743 .listen = rxrpc_listen,
744 .shutdown = rxrpc_shutdown,
745 .setsockopt = rxrpc_setsockopt,
746 .getsockopt = sock_no_getsockopt,
747 .sendmsg = rxrpc_sendmsg,
748 .recvmsg = rxrpc_recvmsg,
749 .mmap = sock_no_mmap,
750 .sendpage = sock_no_sendpage,
753 static struct proto rxrpc_proto = {
755 .owner = THIS_MODULE,
756 .obj_size = sizeof(struct rxrpc_sock),
757 .max_header = sizeof(struct rxrpc_wire_header),
760 static const struct net_proto_family rxrpc_family_ops = {
762 .create = rxrpc_create,
763 .owner = THIS_MODULE,
767 * initialise and register the RxRPC protocol
769 static int __init af_rxrpc_init(void)
774 BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > FIELD_SIZEOF(struct sk_buff, cb));
776 get_random_bytes(&tmp, sizeof(tmp));
780 idr_set_cursor(&rxrpc_client_conn_ids, tmp);
783 rxrpc_call_jar = kmem_cache_create(
784 "rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
785 SLAB_HWCACHE_ALIGN, NULL);
786 if (!rxrpc_call_jar) {
787 pr_notice("Failed to allocate call jar\n");
791 rxrpc_workqueue = alloc_workqueue("krxrpcd", 0, 1);
792 if (!rxrpc_workqueue) {
793 pr_notice("Failed to allocate work queue\n");
794 goto error_work_queue;
797 ret = rxrpc_init_security();
799 pr_crit("Cannot initialise security\n");
803 ret = register_pernet_subsys(&rxrpc_net_ops);
807 ret = proto_register(&rxrpc_proto, 1);
809 pr_crit("Cannot register protocol\n");
813 ret = sock_register(&rxrpc_family_ops);
815 pr_crit("Cannot register socket family\n");
819 ret = register_key_type(&key_type_rxrpc);
821 pr_crit("Cannot register client key type\n");
825 ret = register_key_type(&key_type_rxrpc_s);
827 pr_crit("Cannot register server key type\n");
828 goto error_key_type_s;
831 ret = rxrpc_sysctl_init();
833 pr_crit("Cannot register sysctls\n");
840 unregister_key_type(&key_type_rxrpc_s);
842 unregister_key_type(&key_type_rxrpc);
844 sock_unregister(PF_RXRPC);
846 proto_unregister(&rxrpc_proto);
848 unregister_pernet_subsys(&rxrpc_net_ops);
850 rxrpc_exit_security();
852 destroy_workqueue(rxrpc_workqueue);
854 kmem_cache_destroy(rxrpc_call_jar);
860 * unregister the RxRPC protocol
862 static void __exit af_rxrpc_exit(void)
866 unregister_key_type(&key_type_rxrpc_s);
867 unregister_key_type(&key_type_rxrpc);
868 sock_unregister(PF_RXRPC);
869 proto_unregister(&rxrpc_proto);
870 unregister_pernet_subsys(&rxrpc_net_ops);
871 ASSERTCMP(atomic_read(&rxrpc_n_tx_skbs), ==, 0);
872 ASSERTCMP(atomic_read(&rxrpc_n_rx_skbs), ==, 0);
874 /* Make sure the local and peer records pinned by any dying connections
878 rxrpc_destroy_client_conn_ids();
880 destroy_workqueue(rxrpc_workqueue);
881 rxrpc_exit_security();
882 kmem_cache_destroy(rxrpc_call_jar);
886 module_init(af_rxrpc_init);
887 module_exit(af_rxrpc_exit);