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 /* local epoch for detecting local-end reset */
44 /* current debugging ID */
45 atomic_t rxrpc_debug_id;
47 /* count of skbs currently in use */
48 atomic_t rxrpc_n_tx_skbs, rxrpc_n_rx_skbs;
50 struct workqueue_struct *rxrpc_workqueue;
52 static void rxrpc_sock_destructor(struct sock *);
55 * see if an RxRPC socket is currently writable
57 static inline int rxrpc_writable(struct sock *sk)
59 return atomic_read(&sk->sk_wmem_alloc) < (size_t) sk->sk_sndbuf;
63 * wait for write bufferage to become available
65 static void rxrpc_write_space(struct sock *sk)
69 if (rxrpc_writable(sk)) {
70 struct socket_wq *wq = rcu_dereference(sk->sk_wq);
72 if (skwq_has_sleeper(wq))
73 wake_up_interruptible(&wq->wait);
74 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
80 * validate an RxRPC address
82 static int rxrpc_validate_address(struct rxrpc_sock *rx,
83 struct sockaddr_rxrpc *srx,
88 if (len < sizeof(struct sockaddr_rxrpc))
91 if (srx->srx_family != AF_RXRPC)
94 if (srx->transport_type != SOCK_DGRAM)
95 return -ESOCKTNOSUPPORT;
97 len -= offsetof(struct sockaddr_rxrpc, transport);
98 if (srx->transport_len < sizeof(sa_family_t) ||
99 srx->transport_len > len)
102 if (srx->transport.family != rx->family)
103 return -EAFNOSUPPORT;
105 switch (srx->transport.family) {
107 if (srx->transport_len < sizeof(struct sockaddr_in))
109 tail = offsetof(struct sockaddr_rxrpc, transport.sin.__pad);
112 #ifdef CONFIG_AF_RXRPC_IPV6
114 if (srx->transport_len < sizeof(struct sockaddr_in6))
116 tail = offsetof(struct sockaddr_rxrpc, transport) +
117 sizeof(struct sockaddr_in6);
122 return -EAFNOSUPPORT;
126 memset((void *)srx + tail, 0, len - tail);
127 _debug("INET: %pISp", &srx->transport);
132 * bind a local address to an RxRPC socket
134 static int rxrpc_bind(struct socket *sock, struct sockaddr *saddr, int len)
136 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)saddr;
137 struct sock *sk = sock->sk;
138 struct rxrpc_local *local;
139 struct rxrpc_sock *rx = rxrpc_sk(sk), *prx;
142 _enter("%p,%p,%d", rx, saddr, len);
144 ret = rxrpc_validate_address(rx, srx, len);
150 if (rx->sk.sk_state != RXRPC_UNBOUND) {
155 memcpy(&rx->srx, srx, sizeof(rx->srx));
157 local = rxrpc_lookup_local(&rx->srx);
159 ret = PTR_ERR(local);
163 if (rx->srx.srx_service) {
164 write_lock(&local->services_lock);
165 hlist_for_each_entry(prx, &local->services, listen_link) {
166 if (prx->srx.srx_service == rx->srx.srx_service)
171 hlist_add_head_rcu(&rx->listen_link, &local->services);
172 write_unlock(&local->services_lock);
174 rx->sk.sk_state = RXRPC_SERVER_BOUND;
177 rx->sk.sk_state = RXRPC_CLIENT_BOUND;
180 release_sock(&rx->sk);
185 write_unlock(&local->services_lock);
186 rxrpc_put_local(local);
189 release_sock(&rx->sk);
191 _leave(" = %d", ret);
196 * set the number of pending calls permitted on a listening socket
198 static int rxrpc_listen(struct socket *sock, int backlog)
200 struct sock *sk = sock->sk;
201 struct rxrpc_sock *rx = rxrpc_sk(sk);
202 unsigned int max, old;
205 _enter("%p,%d", rx, backlog);
209 switch (rx->sk.sk_state) {
211 ret = -EADDRNOTAVAIL;
213 case RXRPC_SERVER_BOUND:
214 ASSERT(rx->local != NULL);
215 max = READ_ONCE(rxrpc_max_backlog);
217 if (backlog == INT_MAX)
219 else if (backlog < 0 || backlog > max)
221 old = sk->sk_max_ack_backlog;
222 sk->sk_max_ack_backlog = backlog;
223 ret = rxrpc_service_prealloc(rx, GFP_KERNEL);
225 rx->sk.sk_state = RXRPC_SERVER_LISTENING;
227 sk->sk_max_ack_backlog = old;
234 release_sock(&rx->sk);
235 _leave(" = %d", ret);
240 * rxrpc_kernel_begin_call - Allow a kernel service to begin a call
241 * @sock: The socket on which to make the call
242 * @srx: The address of the peer to contact
243 * @key: The security context to use (defaults to socket setting)
244 * @user_call_ID: The ID to use
245 * @gfp: The allocation constraints
246 * @notify_rx: Where to send notifications instead of socket queue
248 * Allow a kernel service to begin a call on the nominated socket. This just
249 * sets up all the internal tracking structures and allocates connection and
250 * call IDs as appropriate. The call to be used is returned.
252 * The default socket destination address and security may be overridden by
253 * supplying @srx and @key.
255 struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
256 struct sockaddr_rxrpc *srx,
258 unsigned long user_call_ID,
260 rxrpc_notify_rx_t notify_rx)
262 struct rxrpc_conn_parameters cp;
263 struct rxrpc_call *call;
264 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
267 _enter(",,%x,%lx", key_serial(key), user_call_ID);
269 ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
277 if (key && !key->payload.data[0])
278 key = NULL; /* a no-security key */
280 memset(&cp, 0, sizeof(cp));
281 cp.local = rx->local;
283 cp.security_level = 0;
284 cp.exclusive = false;
285 cp.service_id = srx->srx_service;
286 call = rxrpc_new_client_call(rx, &cp, srx, user_call_ID, gfp);
288 call->notify_rx = notify_rx;
290 release_sock(&rx->sk);
291 _leave(" = %p", call);
294 EXPORT_SYMBOL(rxrpc_kernel_begin_call);
297 * rxrpc_kernel_end_call - Allow a kernel service to end a call it was using
298 * @sock: The socket the call is on
299 * @call: The call to end
301 * Allow a kernel service to end a call it was using. The call must be
302 * complete before this is called (the call should be aborted if necessary).
304 void rxrpc_kernel_end_call(struct socket *sock, struct rxrpc_call *call)
306 _enter("%d{%d}", call->debug_id, atomic_read(&call->usage));
307 rxrpc_release_call(rxrpc_sk(sock->sk), call);
308 rxrpc_put_call(call, rxrpc_call_put_kernel);
310 EXPORT_SYMBOL(rxrpc_kernel_end_call);
313 * rxrpc_kernel_new_call_notification - Get notifications of new calls
314 * @sock: The socket to intercept received messages on
315 * @notify_new_call: Function to be called when new calls appear
316 * @discard_new_call: Function to discard preallocated calls
318 * Allow a kernel service to be given notifications about new calls.
320 void rxrpc_kernel_new_call_notification(
322 rxrpc_notify_new_call_t notify_new_call,
323 rxrpc_discard_new_call_t discard_new_call)
325 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
327 rx->notify_new_call = notify_new_call;
328 rx->discard_new_call = discard_new_call;
330 EXPORT_SYMBOL(rxrpc_kernel_new_call_notification);
333 * connect an RxRPC socket
334 * - this just targets it at a specific destination; no actual connection
335 * negotiation takes place
337 static int rxrpc_connect(struct socket *sock, struct sockaddr *addr,
338 int addr_len, int flags)
340 struct sockaddr_rxrpc *srx = (struct sockaddr_rxrpc *)addr;
341 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
344 _enter("%p,%p,%d,%d", rx, addr, addr_len, flags);
346 ret = rxrpc_validate_address(rx, srx, addr_len);
348 _leave(" = %d [bad addr]", ret);
355 if (test_bit(RXRPC_SOCK_CONNECTED, &rx->flags))
358 switch (rx->sk.sk_state) {
360 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
361 case RXRPC_CLIENT_UNBOUND:
362 case RXRPC_CLIENT_BOUND:
369 rx->connect_srx = *srx;
370 set_bit(RXRPC_SOCK_CONNECTED, &rx->flags);
374 release_sock(&rx->sk);
379 * send a message through an RxRPC socket
380 * - in a client this does a number of things:
381 * - finds/sets up a connection for the security specified (if any)
382 * - initiates a call (ID in control data)
383 * - ends the request phase of a call (if MSG_MORE is not set)
384 * - sends a call data packet
385 * - may send an abort (abort code in control data)
387 static int rxrpc_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
389 struct rxrpc_local *local;
390 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
393 _enter(",{%d},,%zu", rx->sk.sk_state, len);
395 if (m->msg_flags & MSG_OOB)
399 ret = rxrpc_validate_address(rx, m->msg_name, m->msg_namelen);
401 _leave(" = %d [bad addr]", ret);
408 switch (rx->sk.sk_state) {
410 rx->srx.srx_family = AF_RXRPC;
411 rx->srx.srx_service = 0;
412 rx->srx.transport_type = SOCK_DGRAM;
413 rx->srx.transport.family = rx->family;
414 switch (rx->family) {
416 rx->srx.transport_len = sizeof(struct sockaddr_in);
418 #ifdef CONFIG_AF_RXRPC_IPV6
420 rx->srx.transport_len = sizeof(struct sockaddr_in6);
427 local = rxrpc_lookup_local(&rx->srx);
429 ret = PTR_ERR(local);
434 rx->sk.sk_state = RXRPC_CLIENT_UNBOUND;
437 case RXRPC_CLIENT_UNBOUND:
438 case RXRPC_CLIENT_BOUND:
440 test_bit(RXRPC_SOCK_CONNECTED, &rx->flags)) {
441 m->msg_name = &rx->connect_srx;
442 m->msg_namelen = sizeof(rx->connect_srx);
444 case RXRPC_SERVER_BOUND:
445 case RXRPC_SERVER_LISTENING:
446 ret = rxrpc_do_sendmsg(rx, m, len);
454 release_sock(&rx->sk);
455 _leave(" = %d", ret);
460 * set RxRPC socket options
462 static int rxrpc_setsockopt(struct socket *sock, int level, int optname,
463 char __user *optval, unsigned int optlen)
465 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
466 unsigned int min_sec_level;
469 _enter(",%d,%d,,%d", level, optname, optlen);
474 if (level == SOL_RXRPC) {
476 case RXRPC_EXCLUSIVE_CONNECTION:
481 if (rx->sk.sk_state != RXRPC_UNBOUND)
483 rx->exclusive = true;
486 case RXRPC_SECURITY_KEY:
491 if (rx->sk.sk_state != RXRPC_UNBOUND)
493 ret = rxrpc_request_key(rx, optval, optlen);
496 case RXRPC_SECURITY_KEYRING:
501 if (rx->sk.sk_state != RXRPC_UNBOUND)
503 ret = rxrpc_server_keyring(rx, optval, optlen);
506 case RXRPC_MIN_SECURITY_LEVEL:
508 if (optlen != sizeof(unsigned int))
511 if (rx->sk.sk_state != RXRPC_UNBOUND)
513 ret = get_user(min_sec_level,
514 (unsigned int __user *) optval);
518 if (min_sec_level > RXRPC_SECURITY_MAX)
520 rx->min_sec_level = min_sec_level;
531 release_sock(&rx->sk);
536 * permit an RxRPC socket to be polled
538 static unsigned int rxrpc_poll(struct file *file, struct socket *sock,
541 struct sock *sk = sock->sk;
542 struct rxrpc_sock *rx = rxrpc_sk(sk);
545 sock_poll_wait(file, sk_sleep(sk), wait);
548 /* the socket is readable if there are any messages waiting on the Rx
550 if (!list_empty(&rx->recvmsg_q))
551 mask |= POLLIN | POLLRDNORM;
553 /* the socket is writable if there is space to add new data to the
554 * socket; there is no guarantee that any particular call in progress
555 * on the socket may have space in the Tx ACK window */
556 if (rxrpc_writable(sk))
557 mask |= POLLOUT | POLLWRNORM;
563 * create an RxRPC socket
565 static int rxrpc_create(struct net *net, struct socket *sock, int protocol,
568 struct rxrpc_sock *rx;
571 _enter("%p,%d", sock, protocol);
573 if (!net_eq(net, &init_net))
574 return -EAFNOSUPPORT;
576 /* we support transport protocol UDP/UDP6 only */
577 if (protocol != PF_INET &&
578 IS_ENABLED(CONFIG_AF_RXRPC_IPV6) && protocol != PF_INET6)
579 return -EPROTONOSUPPORT;
581 if (sock->type != SOCK_DGRAM)
582 return -ESOCKTNOSUPPORT;
584 sock->ops = &rxrpc_rpc_ops;
585 sock->state = SS_UNCONNECTED;
587 sk = sk_alloc(net, PF_RXRPC, GFP_KERNEL, &rxrpc_proto, kern);
591 sock_init_data(sock, sk);
592 sock_set_flag(sk, SOCK_RCU_FREE);
593 sk->sk_state = RXRPC_UNBOUND;
594 sk->sk_write_space = rxrpc_write_space;
595 sk->sk_max_ack_backlog = 0;
596 sk->sk_destruct = rxrpc_sock_destructor;
599 rx->family = protocol;
602 INIT_HLIST_NODE(&rx->listen_link);
603 spin_lock_init(&rx->incoming_lock);
604 INIT_LIST_HEAD(&rx->sock_calls);
605 INIT_LIST_HEAD(&rx->to_be_accepted);
606 INIT_LIST_HEAD(&rx->recvmsg_q);
607 rwlock_init(&rx->recvmsg_lock);
608 rwlock_init(&rx->call_lock);
609 memset(&rx->srx, 0, sizeof(rx->srx));
611 _leave(" = 0 [%p]", rx);
616 * Kill all the calls on a socket and shut it down.
618 static int rxrpc_shutdown(struct socket *sock, int flags)
620 struct sock *sk = sock->sk;
621 struct rxrpc_sock *rx = rxrpc_sk(sk);
624 _enter("%p,%d", sk, flags);
626 if (flags != SHUT_RDWR)
628 if (sk->sk_state == RXRPC_CLOSE)
633 spin_lock_bh(&sk->sk_receive_queue.lock);
634 if (sk->sk_state < RXRPC_CLOSE) {
635 sk->sk_state = RXRPC_CLOSE;
636 sk->sk_shutdown = SHUTDOWN_MASK;
640 spin_unlock_bh(&sk->sk_receive_queue.lock);
642 rxrpc_discard_prealloc(rx);
649 * RxRPC socket destructor
651 static void rxrpc_sock_destructor(struct sock *sk)
655 rxrpc_purge_queue(&sk->sk_receive_queue);
657 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
658 WARN_ON(!sk_unhashed(sk));
659 WARN_ON(sk->sk_socket);
661 if (!sock_flag(sk, SOCK_DEAD)) {
662 printk("Attempt to release alive rxrpc socket: %p\n", sk);
668 * release an RxRPC socket
670 static int rxrpc_release_sock(struct sock *sk)
672 struct rxrpc_sock *rx = rxrpc_sk(sk);
674 _enter("%p{%d,%d}", sk, sk->sk_state, atomic_read(&sk->sk_refcnt));
676 /* declare the socket closed for business */
678 sk->sk_shutdown = SHUTDOWN_MASK;
680 spin_lock_bh(&sk->sk_receive_queue.lock);
681 sk->sk_state = RXRPC_CLOSE;
682 spin_unlock_bh(&sk->sk_receive_queue.lock);
684 ASSERTCMP(rx->listen_link.next, !=, LIST_POISON1);
686 if (!hlist_unhashed(&rx->listen_link)) {
687 write_lock(&rx->local->services_lock);
688 hlist_del_rcu(&rx->listen_link);
689 write_unlock(&rx->local->services_lock);
692 /* try to flush out this socket */
693 rxrpc_discard_prealloc(rx);
694 rxrpc_release_calls_on_socket(rx);
695 flush_workqueue(rxrpc_workqueue);
696 rxrpc_purge_queue(&sk->sk_receive_queue);
698 rxrpc_put_local(rx->local);
702 key_put(rx->securities);
703 rx->securities = NULL;
711 * release an RxRPC BSD socket on close() or equivalent
713 static int rxrpc_release(struct socket *sock)
715 struct sock *sk = sock->sk;
717 _enter("%p{%p}", sock, sk);
724 return rxrpc_release_sock(sk);
728 * RxRPC network protocol
730 static const struct proto_ops rxrpc_rpc_ops = {
732 .owner = THIS_MODULE,
733 .release = rxrpc_release,
735 .connect = rxrpc_connect,
736 .socketpair = sock_no_socketpair,
737 .accept = sock_no_accept,
738 .getname = sock_no_getname,
740 .ioctl = sock_no_ioctl,
741 .listen = rxrpc_listen,
742 .shutdown = rxrpc_shutdown,
743 .setsockopt = rxrpc_setsockopt,
744 .getsockopt = sock_no_getsockopt,
745 .sendmsg = rxrpc_sendmsg,
746 .recvmsg = rxrpc_recvmsg,
747 .mmap = sock_no_mmap,
748 .sendpage = sock_no_sendpage,
751 static struct proto rxrpc_proto = {
753 .owner = THIS_MODULE,
754 .obj_size = sizeof(struct rxrpc_sock),
755 .max_header = sizeof(struct rxrpc_wire_header),
758 static const struct net_proto_family rxrpc_family_ops = {
760 .create = rxrpc_create,
761 .owner = THIS_MODULE,
765 * initialise and register the RxRPC protocol
767 static int __init af_rxrpc_init(void)
771 BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > FIELD_SIZEOF(struct sk_buff, cb));
773 get_random_bytes(&rxrpc_epoch, sizeof(rxrpc_epoch));
774 rxrpc_epoch |= RXRPC_RANDOM_EPOCH;
775 get_random_bytes(&rxrpc_client_conn_ids.cur,
776 sizeof(rxrpc_client_conn_ids.cur));
777 rxrpc_client_conn_ids.cur &= 0x3fffffff;
778 if (rxrpc_client_conn_ids.cur == 0)
779 rxrpc_client_conn_ids.cur = 1;
782 rxrpc_call_jar = kmem_cache_create(
783 "rxrpc_call_jar", sizeof(struct rxrpc_call), 0,
784 SLAB_HWCACHE_ALIGN, NULL);
785 if (!rxrpc_call_jar) {
786 pr_notice("Failed to allocate call jar\n");
790 rxrpc_workqueue = alloc_workqueue("krxrpcd", 0, 1);
791 if (!rxrpc_workqueue) {
792 pr_notice("Failed to allocate work queue\n");
793 goto error_work_queue;
796 ret = rxrpc_init_security();
798 pr_crit("Cannot initialise security\n");
802 ret = proto_register(&rxrpc_proto, 1);
804 pr_crit("Cannot register protocol\n");
808 ret = sock_register(&rxrpc_family_ops);
810 pr_crit("Cannot register socket family\n");
814 ret = register_key_type(&key_type_rxrpc);
816 pr_crit("Cannot register client key type\n");
820 ret = register_key_type(&key_type_rxrpc_s);
822 pr_crit("Cannot register server key type\n");
823 goto error_key_type_s;
826 ret = rxrpc_sysctl_init();
828 pr_crit("Cannot register sysctls\n");
832 #ifdef CONFIG_PROC_FS
833 proc_create("rxrpc_calls", 0, init_net.proc_net, &rxrpc_call_seq_fops);
834 proc_create("rxrpc_conns", 0, init_net.proc_net,
835 &rxrpc_connection_seq_fops);
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 rxrpc_exit_security();
850 destroy_workqueue(rxrpc_workqueue);
852 kmem_cache_destroy(rxrpc_call_jar);
858 * unregister the RxRPC protocol
860 static void __exit af_rxrpc_exit(void)
864 unregister_key_type(&key_type_rxrpc_s);
865 unregister_key_type(&key_type_rxrpc);
866 sock_unregister(PF_RXRPC);
867 proto_unregister(&rxrpc_proto);
868 rxrpc_destroy_all_calls();
869 rxrpc_destroy_all_connections();
870 ASSERTCMP(atomic_read(&rxrpc_n_tx_skbs), ==, 0);
871 ASSERTCMP(atomic_read(&rxrpc_n_rx_skbs), ==, 0);
872 rxrpc_destroy_all_locals();
874 remove_proc_entry("rxrpc_conns", init_net.proc_net);
875 remove_proc_entry("rxrpc_calls", init_net.proc_net);
876 destroy_workqueue(rxrpc_workqueue);
877 rxrpc_exit_security();
878 kmem_cache_destroy(rxrpc_call_jar);
882 module_init(af_rxrpc_init);
883 module_exit(af_rxrpc_exit);